Virtue Theoretic Epistemology: New Methods and Approaches 1108481213, 9781108481212

Table of contents :
Contents
List of Figures
List of Contributors
Acknowledgements
Introduction: Virtue Theoretic Epistemology • Christoph Kelp
1 Closed-Mindedness As an Intellectual Vice • Heather Battaly
2 Epistemic Virtues and Virtues with Epistemic Content • Cameron Boult, Christoph Kelp, Johanna Schnurr and Mona Simion
3 Difficulty and Knowledge • Fernando Broncano-Berrocal
4 What Is Epistemic Entitlement? Reliable Competence, Reasons, Inference, Access • Peter J. Graham
5 Knowledge-Producing Abilities • John Greco
6 Virtue Epistemology, Two Kinds of Internalism, and the Intelligibility Problem • Jonathan L. Kvanvig
7 Knowledge Is Extrinsically Apt Belief: Virtue Epistemology and the Temporal Objection • Anne Meylan
8 Explaining Knowledge • Alan Millar
9 Anti-risk Virtue Epistemology • Duncan Pritchard
10 Responsibilism within Reason • Kurt Sylvan
Index

Citation preview

VIRTUE THEORETIC EPISTEMOLOGY

Virtue epistemology is one of the most flourishing research programmes in contemporary epistemology. Its defining thesis is that properties of agents and groups are the primary focus of epistemic theorising. Within virtue epistemology two key strands can be distinguished: virtue reliabilism, which focuses on agent properties that are strongly truth-conducive, such as perceptual and inferential abilities of agents; and virtue responsibilism, which focuses on intellectual virtues in the sense of character traits of agents, such as open-mindedness and intellectual courage. This volume brings together ten new chapters on virtue epistemology, with contributions to both its key strands, written by leading authors in the field. It will advance the state of the art and provide readers with a valuable overview of what virtue epistemology has achieved.   is a senior lecturer in philosophy at the University of Glasgow. He is the author of Good Thinking: A Knowledge First Virtue Epistemology () and winner of the  Young Epistemologist Prize.   holds the McDevitt Chair in Philosophy at Georgetown University. His publications include Achieving Knowledge: A Virtue-Theoretic Account of Epistemic Normativity (Cambridge, ) and The Transmission of Knowledge (Cambridge, forthcoming).

VIRTUE THEORETIC EPISTEMOLOGY New Methods and Approaches       CHRISTOPH KELP University of Glasgow

JOHN GRECO Georgetown University

University Printing House, Cambridge  , United Kingdom One Liberty Plaza, th Floor, New York,  , USA  Williamstown Road, Port Melbourne,  , Australia –, rd Floor, Plot , Splendor Forum, Jasola District Centre, New Delhi – , India  Anson Road, #–/, Singapore  Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning, and research at the highest international levels of excellence. www.cambridge.org Information on this title: www.cambridge.org/ : ./ © Cambridge University Press  This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published  Printed in the United Kingdom by TJ International Ltd, Padstow Cornwall A catalogue record for this publication is available from the British Library. Library of Congress Cataloging-in-Publication Data : Greco, John, editor. | Kelp, Christoph, editor. : Virtue theoretic epistemology : new methods and approaches / edited by John Greco, Christoph Kelp. : Cambridge ; New York, NY : Cambridge University Press, . | Includes bibliographical references and index. :   (print) |   (ebook) |   (hardback) |   (paperback) |   (ebook) : : Virtue epistemology. :   .  (print) |   (ebook) |  –dc LC record available at https://lccn.loc.gov/ LC ebook record available at https://lccn.loc.gov/  ---- Hardback Cambridge University Press has no responsibility for the persistence or accuracy of URLs for external or third-party internet websites referred to in this publication and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.

Contents

List of Figures List of Contributors Acknowledgements

page vii viii ix

Introduction: Virtue Theoretic Epistemology



Christoph Kelp



Closed-Mindedness As an Intellectual Vice



Heather Battaly



Epistemic Virtues and Virtues with Epistemic Content



Cameron Boult, Christoph Kelp, Johanna Schnurr and Mona Simion





Difficulty and Knowledge Fernando Broncano-Berrocal



What Is Epistemic Entitlement? Reliable Competence, Reasons, Inference, Access



Peter J. Graham





Knowledge-Producing Abilities John Greco



Virtue Epistemology, Two Kinds of Internalism, and the Intelligibility Problem



Jonathan L. Kvanvig



Knowledge Is Extrinsically Apt Belief: Virtue Epistemology and the Temporal Objection



Anne Meylan





Explaining Knowledge Alan Millar

v

Contents

vi 

Anti-risk Virtue Epistemology



Duncan Pritchard

 Responsibilism within Reason



Kurt Sylvan

Index



Figures

. . . . .

A decision-tree and the scope of entitlement A decision-tree and inference The environments on Earth and Twin Earth Epistemic Twin Earth Metaphysical picture of intrinsic and extrinsic properties

vii

page     

Contributors

Heather Battaly, Professor of Philosophy, University of Connecticut Cameron Boult, Assistant Professor of Philosophy, Brandon University Fernando Broncano-Berrocal, Talent Attraction Fellow, Autonomous University of Madrid Peter J. Graham, Professor of Philosophy, University of California, Riverside John Greco, Robert and Catherine McDevitt Chair in Philosophy, Georgetown University Christoph Kelp, Senior Lecturer in Philosophy, University of Glasgow Jonathan L. Kvanvig, Professor of Philosophy, Washington University, St Louis Anne Meylan, Assistant Professor of Philosophy, University of Zurich Alan Millar, Emeritus Professor of Philosophy, University of Stirling Duncan Pritchard, Distinguished Professor of Philosophy, University of California, Irvine and Chair in Epistemology, University of Edinburgh Johanna Schnurr, Lecturer in Philosophy, Jesus College, Oxford Mona Simion, Lecturer in Philosophy, University of Glasgow Kurt Sylvan, Associate Professor in Philosophy, University of Southampton

viii

Acknowledgements

We would like to thank our contributors to the volume for their good work and patience throughout the process. We would also like to thank Hal Churchman and Stephanie Taylor at Cambridge University Press for their work and support for the project, as well as Yassar Abdulnasser for his work on production and Elizabeth Stone for her work on copy-editing. Finally, many thanks to Haicheng Zhao for creating the index.

ix

Introduction Virtue Theoretic Epistemology Christoph Kelp

There are three leading theories of normativity: teleology, deontology and virtue theory. All three types of normative theory countenance values, norms and virtues. What they disagree on is the order of explanation. Teleology takes values to be the fundamental normative kind and explains norms and virtues in terms of them. Deontology takes norms to be the fundamental normative kind and explains values and virtues in terms of them. And, finally, virtue theory takes virtues to be the fundamental normative kind and explains norms and values in terms of them. To get a better feel for how this is meant to work, let’s look at some dummy versions of teleological, deontological and virtue theories of normativity. According to our dummy teleological theory, one ought to ϕ if and only if ϕ-ing maximises value. And, similarly, S is a virtuous person if and only if S is disposed to maximise value. What makes this view into a distinctively teleological theory of normativity is that the direction of explanation goes from right to left with the result that norms and virtues are analysed in terms of value. What we ought to do and what makes for a virtuous person are explained in terms of value. Our dummy deontological account has it that X has value if and only if one ought to favour X. And S is a virtuous person if and only if S has a disposition to do as S ought. This dummy view is deontological in virtue of the direction of explanation, which once again goes from right to left. What’s of value and what makes for a virtuous person are explained in terms of what one ought to do. Finally, and most importantly for present purposes, let’s look at a dummy virtue theory of normativity. According to our view, X is valuable if and only if X would characteristically be favoured by a virtuous person, and one ought to ϕ if and only if ϕ-ing is what a virtuous person would characteristically do. What’s important here is again the way in which the direction of explanation proceeds. What is of value and what one ought to 



 

do is analysed in terms of virtue. This makes our dummy view a distinctively virtue theoretic approach to normativity. As characterised so far, the debate between the three theories of normativity is extremely abstract. Things get a little more concrete once we apply these views to more familiar normative domains. The kind of domains I have in mind here include ethics, aesthetics and, most importantly for present purposes, epistemology. It is hard to deny that most of the research has been done in normative ethics. While the debates with teleological and deontological approaches to ethics are of course important, it is worth mentioning that there are a variety of important strands of virtue ethics on the market. According to the perhaps most popular form of virtue ethics, and the one I will be focusing on here, the idea of eudaimonia takes centre stage. Eudaimonia is often translated as flourishing, happiness or well-being. What is important to note, however, is that it is itself a value-laden notion. It is the kind of well-being that is worth having (Hursthouse and Pettigrove ). Consider, by way of illustration, the following case: eating lots of chocolate might give you lots of pleasure. Say that you spend your entire life doing nothing but eating chocolate. At the end of your life you look back with great satisfaction. You take yourself to have led a happy life. Even if in your chocolate-eating life you achieved a form of happiness, you did not achieve eudaimonia. The reason for this is that the happiness that derives from only ever eating chocolate is not a form of well-being worth having, at least not for a normal adult human being. Now, the thought is that moral virtues play the central part in normative ethics that they do because of their relation to eudaimonia. Of course, the thought cannot be that eudaimonia is identified as a value that’s independent of the moral virtues and that moral virtues are understood in terms of their relation to eudaimonia. After all, this would turn the view right into a teleological normative ethics. Rather, the idea is that the moral virtues are constitutive of and essential to eudaimonia. This means that we cannot fully understand what eudaimonia is without understanding the moral virtues. In this way, virtues play the foundational role in ethics that virtue ethics takes them to play. For champions of this kind of virtue ethics, one central question will be which moral virtues are constitutive of and essential to eudaimonia. While 

Classical defences of teleological approaches to moral normativity include Bentham ; Mill ; Sidgwick . For classical defences of deontological approaches, see e.g. Kant ; Ross  and for virtue ethics, e.g. Aristotle ; Plato .

Introduction: Virtue Theoretic Epistemology



there is of course ample room for debate, it is hard to deny that there are a number of character traits that are widely recognised as virtues, including charity, courage, generosity, honesty, justice and kindness, to name but a few. The morally perfectly virtuous agent possesses all the moral virtues to the highest degree. Less than morally perfectly virtuous agents are virtuous to the extent that they approximate the morally perfectly virtuous agent. With these points in play, we can see how our virtue ethics might account for moral norms and values, at least on the dummy version outlined above. According to one straightforward way of implementing the view, we get the result that one morally ought to ϕ if and only if ϕ-ing is what a morally virtuous agent would characteristically do. To take a more concrete example, you morally ought not to lie if and only if this is what a morally virtuous person would characteristically do. Since lying is not something that a morally virtuous person would do (as the virtue of honesty would characteristically lead them in this direction), you morally ought not to lie. Crucially, what we find here is the direction of explanation that is key to virtue theories of morality. What explains why you morally ought not to lie is not, for instance, that it would violate a moral norm, as deontologists would have us think, but that it is something a morally virtuous person would not do. Moreover, X is morally valuable if and only if X would characteristically be favoured by a morally virtuous agent. For instance, organisations such as Doctors without Borders and UNICEF have moral value if and only if they are the kind of entity that a morally virtuous agent would characteristically favour. Since they are the kind of organisation that a virtuous person would characteristically favour (as the virtue of charity would characteristically lead them to do so), these organisations have moral value. Again, what’s important here is the direction of explanation. It is not the fact that these organisations bring about a lot of moral value in the world that explains why they are morally good. Rather, it is the fact that they are likely to be favoured by a morally virtuous agent. Let us now move on to virtue epistemology. If what we have just described captures the contours of virtue ethics at least roughly, it is tempting to think that the contours of virtue epistemology can be captured in much the same way. For instance, it is tempting to think that virtue epistemology will adopt a virtue-based account of epistemic normativity. It is also tempting to think that a virtue epistemology (will, or at least might) hold that intellectual virtues, like moral virtues, play the central part in normative epistemology they do because, just like moral virtues, they are partly constitutive of eudaimonia. After all, it is plausible that eudaimonia is constituted not only by moral virtues but also by intellectual virtues.



 

Again, there is ample room for discussion, but in the intellectual sphere as well, there is a wide range of character traits that are widely recognised as virtues, including attentiveness, curiosity, intellectual courage, intellectual humility, open-mindedness and intellectual thoroughness. Unfortunately, there is reason not to adopt this straightforward way of characterising virtue epistemology. Or, to be more precise, there is reason not to do so if we want all paradigm cases of virtue epistemologies to be classified as virtue epistemologies. To see this, note that it is customary to distinguish between two different kinds of virtue epistemology: reliabilist and responsibilist. The trouble is that reliabilist virtue epistemologies will simply not fit with the general theory of normativity that takes virtues to be fundamental. This is because reliabilist virtue epistemologies are very clear that at least some (perhaps all) epistemic virtues are virtues at least in part because they are conducive to producing some independently specified epistemic good or other. In other words, they offer a distinctively teleological account of epistemic virtues. And, of course, it is hard to see how this could still be compatible with the approach to general normativity that is distinctive of virtue theories of normativity and the approach to moral normativity that we found in virtue ethics. Again, an example may help to make this point clearer. Consider a disposition that takes the look of a certain object as trigger and beliefs about the presence of chanterelles as manifestations. Reliabilist virtue epistemologists will typically allow for this disposition to count as an epistemic virtue. Crucially, the reason why they do so is that the disposition reliably produces true beliefs rather than false ones, at least in suitably favourable conditions. But, of course, if this is how we explain why certain dispositions count as epistemic virtues, we are effectively offering a teleological account of these epistemic virtues. It is hard to see how this can be squared with the approach to normativity distinctive of virtue theory according to which virtues play the foundational part. 



Among the most influential defences of virtue epistemology are Greco ; Pritchard, Millar, and Haddock ; and Sosa  on the reliabilist side and Code ; Montmarquet ; and Zagzebski  on the responsibilist side. It may also be worth noting that these virtues cannot be accounted for except in teleological terms. To see this, note that the very same disposition may not count as an epistemic virtue for agents who inhabit environments in which too many things that have the relevant look are not chanterelles (e.g. environments in which jack-o’-lanterns abound) and, as a result, the disposition does not produce true beliefs about the presence of chanterelles reliably enough. What comes to light is that whether one and the same disposition is an epistemic virtue may turn only on a difference in reliability with which it produces true belief. At the same time, it is hard to see how it can be the difference in the degree of reliability that determines whether one and the same disposition counts as an epistemic virtue unless we adopt a teleological account of epistemic normativity.

Introduction: Virtue Theoretic Epistemology



It might be thought that the take-home lesson is that reliabilist virtue epistemology is not virtue epistemology proper and that virtue responsibilism is the real thing. On the face of it, this may sound promising. And it gets even more promising once we appreciate the fact that virtue responsibilists are in practice often much closer to virtue ethicists than virtue reliabilists. Reliabilist virtue epistemologists are for the most part interested in offering accounts of the nature and value of core phenomena in epistemology such as knowledge and justified belief. Virtue responsibilists, in contrast, are often less interested in the analysis of knowledge and justified belief. What they often focus on instead is the nature of the wide range of intellectual virtues that we find among humans, including the likes of curiosity, open-mindedness and intellectual humility. It is much less clear that these virtues must be understood in terms of the good epistemic consequences they tend to produce. Rather, an account in terms of their relation to eudaimonia, perhaps along much the same lines as virtue ethicists envisage for the moral virtues, may still seem quite promising. Does this mean that, as true virtue epistemologists, we will have to turn to virtue responsibilism? The answer to this question is no. This is because there is independent reason to think that a virtue theory of normativity isn’t all that plausible for the epistemic domain in the first place. To see why not, note that on a virtue theory of normativity, virtues will essentially have positive normative status and not possessing a given virtue will necessarily reflect negatively on a given moral agent (or at least show that there is room for improvement). Note also that this is fairly plausible in the case of the moral virtues. That is to say, it is fairly plausible that, for instance, kindness necessarily has positive moral status and that not possessing this virtue will reflect badly on a moral agent. The trouble is that the same does not appear to be true in the case of epistemic agents. Perhaps the easiest way to see this is by considering infallible and omniscient epistemic agents. Infallible omniscient agents may fail to possess a range of core responsibilist virtues without this reflecting badly on them qua epistemic agents, nor even meaning that they would be improved qua epistemic agents if they had these virtues. Most importantly, they may not be curious, open-minded, intellectually humble or intellectually courageous. And yet, for all that, they may not be any worse qua epistemic agents, nor even better if they did have them. If this is right, there are intellectual virtues, even paradigm ones, which do not enjoy positive epistemic status essentially. And that does not fit well with a virtue theory of epistemic normativity.



 

By way of further support for the thesis that a virtue theory of normativity isn’t all that plausible for the epistemic domain, consider an epistemic phenomenon that is widely believed to enjoy positive epistemic status essentially, to wit, knowledge. Given that knowledge enjoys positive epistemic status essentially, we may expect a virtue theory of epistemic normativity to explain why this is. The trouble is that the prospects for a satisfactory account are rather dim here. Knowledge is not a virtue in itself. And an account in terms only of the kinds of virtues that the responsibilist countenances doesn’t seem promising either. You could be entirely intellectually non-virtuous (in the responsibilist sense) and still acquire perceptual knowledge, say that there is a chanterelle before you, at least in suitably favourable environments. Since your perceptual knowledge will still have positive epistemic status, an explanation in terms of responsibilist virtues seems unpromising. What comes to light, then, is that a virtue theory of epistemic normativity seems unpromising. This may be a surprising result to find defended in the introduction to a volume on virtue epistemology. On the upside, it once again levels the playing field between the two most prominent kinds of virtue epistemology, reliabilism and responsibilism. The concern that only virtue responsibilism is a virtue epistemology in the proper sense because only virtue responsibilism can accord virtues the fundamental normative role required by a genuine virtue theory of epistemic normativity can be laid to rest. What’s more, there is another way of interpreting the central thesis of virtue theoretic approaches to epistemology, which is plausibly more promising. Virtue theories of normativity venture to analyse norms and values in terms virtues. For instance, according to our dummy view X is valuable if and only if X would characteristically be favoured by a virtuous person, and one ought to ϕ if and only if ϕ-ing is what a virtuous person would characteristically do. Even if we abandon an ambitious virtue theory of epistemic normativity like our dummy view, we might still hold out hope for a more modest research programme in epistemology. Specifically, even if values and norms in the epistemic domain cannot be accounted for only in terms of epistemic virtues, it may be that what it is for someone to believe or do something well epistemically is to be accounted for in terms of virtues. That is to say, it may be that believing or doing something well epistemically is believing or acting from virtue. And, mutatis mutandis, for a wide range of positive epistemic statuses for epistemically assessable ϕ-ing. To mark the distinction between the ambitious and the modest programme, I henceforth reserve the term ‘virtue epistemology’ for the

Introduction: Virtue Theoretic Epistemology



ambitious programme. This is in recognition of the fact that the programme that is designated by the term ‘virtue ethics’ in moral philosophy is of the ambitious variety. In addition, I will use the label ‘virtue theoretic epistemology’ for the modest programme. My proposal is that the bulk of research that has been carried out in recent epistemology is really best understood as falling within the purview of the modest programme. Given this terminological distinction, it is thus most aptly characterised as contributing to virtue theoretic epistemology. It is not hard to see that a lot of the work that has been done in reliabilist virtue theoretic epistemology is indeed best understood along these lines. What has crystallised especially in the last decade or so is just how central a general theory of the normativity of performances is to virtue theoretic epistemology in the reliabilist camp. According to this theory, performances with an aim in general can be assessed along three normative dimensions: success, competence and aptness (e.g. Sosa ). Thus, for any performance with an aim, we can ask whether the performance attains its aim, i.e. whether it is successful. We can also ask whether the performance is produced by the exercise of an ability (or virtue) on the part of the performing agent, i.e. whether it is competent. And finally, we can ask whether the performance is successful because competent, i.e. whether it is apt. Crucially for present purposes, each of these dimensions specifies a way in which the agent may be performing well. An example may help to illustrate these ideas. Consider a free throw in basketball. A free throw in basketball is a performance with an aim. The aim is for the ball to go into the basket. Suppose you take a free throw. By the theory of the normativity of performances, there are three normative properties in terms of which we may assess your free throw. We may ask whether your free throw was successful. Given that its aim is for the ball to go into the basket, your free throw is successful if and only if it gets the ball into the basket. We may also ask whether your free throw was competent, i.e. whether it was produced by the exercise of an ability on your part to make free throws. Finally, we may ask whether your free throw was apt, i.e. whether it was successful because competent. This will be the case if and only if your free throw got the ball into the basket because of the exercise of an ability on your part to make free throws. It is important to note that success, competence and aptness are different normative kinds. True, there are relations between them. Most importantly, aptness by definition entails success and competence. Recall that an apt performance is a performance that it successful because competent. Given that this is so, any apt performance must also be successful and



 

competent. For instance, an apt free throw is one that finds the basket because of the exercise of an ability to make free throws. It is easy to see that an apt free throw must also find the basket and that it must be produced by an exercise of an ability to make free throws. At the same time, the three also come apart in important ways. For instance, it is possible for a performance to be successful but not competent and vice versa. For instance, you may be drunk beyond comprehension, completely unaware that you are to take a free throw and just randomly throw the ball that you are being handed. As it so happens, the ball finds the basket. Your free throw was successful but not competent. And, conversely, it may be that you are taking a perfectly competent shot that would have gone right in, had it not been for the interference of a disruptive fan who threw an apple at your ball, thus bringing it off course. Finally, a performance can be competent and successful but not apt. For instance, you may take a competent shot that would have gone right in had it not been deflected by the interference of a gust of wind, which brings it off course. At the same time, unbeknownst to you, a helper with a wind machine brings your shot back on target and it does go in after all. In this case, your shot is both competent and successful but not apt. One key idea in reliabilist virtue theoretic epistemology is that beliefs are performances with an aim. In particular, they have a distinctively epistemic aim. Given that this is so, the theory of the normativity of performances applies to the case of belief. That is to say, we can ask whether a belief is successful, i.e. whether it attains its epistemic aim. We can ask whether it is competent, i.e. whether it is produced by the exercise of an ability to attain the epistemic aim (henceforth also epistemic ability or virtue). And we can ask whether it is apt, i.e. whether it attains its epistemic aim because of the exercise of an epistemic ability. In this way, we get three normative properties of belief. And since these three normative properties specify three ways in which one may be performing well, the result that we get here is that these three normative properties specify three ways in which one may be believing well epistemically. Finally, another key idea is to identify normative properties of beliefs as performances with an epistemic aim with familiar epistemic properties. According to the standard view, the epistemic aim of belief is truth. As a result, the standard view supports an identification of successful belief with true belief. It also offers more information on what epistemic abilities are, i.e. abilities to form true beliefs. Thereby, it also tells us more about what competent and apt belief amount to. Finally, and crucially, the standard view identifies competent belief with justified belief and apt belief with

Introduction: Virtue Theoretic Epistemology



knowledge. The result that we get, then, is that familiar epistemic properties are simply instances of more general normative properties of performances with an aim. Moreover, it comes to light that true belief, justified belief and knowledge also correspond to ways in which one may believe well epistemically. And, of course, this means that the relevant forms of virtue reliabilism are brought straight under the umbrella of what we call virtue theoretic epistemology. What about virtue responsibilism? How does this important strand of virtue theoretic epistemology relate to the present idea that to believe or do something well epistemically is to believe or act from epistemic virtue? There are a number of ways in which virtue responsibilism can be squared with this idea. This is unsurprising, given that virtue responsibilism is a much more multifarious research project than virtue reliabilism. For instance, one important question that virtue responsibilists are divided on is whether they should even aim to offer accounts of familiar epistemic properties such as knowledge and/or justified belief. There are those who think that the answer is yes. For them, it will be very natural to adopt the idea that believing well epistemically is to be analysed in terms of epistemic virtues. After all, knowledge and justified belief continue to be paradigm cases of believing well epistemically. But even those who return a negative verdict need not therefore reject the core idea of virtue theoretic epistemology. To see why not, note that there is fairly widespread consensus among virtue responsibilists that intellectual character virtues play a key normative role in enquiry (Code ; Hookway ). In particular, virtues feature in norms of enquiry. For instance, we ought to enquire in a way that is open-minded, intellectually courageous and so on. Given that this is so, it is only plausible that to have conducted an enquiry well will be unpacked in terms of virtues. And, of course, this fits nicely with our core conception of virtue theoretic epistemology. There is one complication here, however. Enquiry is an activity with a constitutive aim. One way to characterise this aim in a fairly neutral manner is by saying that enquiry aims at settling the question at issue (Kelp ). Now, while many would sign up to this neutral characterisation of the aim of enquiry, there is a lively debate among epistemologists about how to characterise the aim of enquiry in a more substantive manner. The leading candidates here are that enquiry aims at knowledge (e.g. Kelp ; Millar ; Williamson ) or else at true belief (e.g. Kvanvig ; Lynch ). Crucially, if enquiry has a constitutive aim, it is plausible that virtues of enquiry are unpacked along instrumentalist lines, in terms of their conduciveness to the attainment of enquiry’s



 

constitutive aim. Note that this way of thinking has precedents outside the epistemic domain. For instance, being tall is a virtue in basketball players. The reason why this is so is arguably that it is conducive to attaining the constitutive aim of basketball, i.e. winning the game, for instance by generating more successful blocks. What’s more, this instrumentalist construal of virtues of enquiry is independently attractive. After all, it is tempting to think that the reason why open-mindedness, intellectual courage and so on are virtues of enquiry is that they are conducive to attaining the aim of enquiry. At this stage, one may wonder where the complication lies. Isn’t all this good and well and perfectly acceptable for virtue responsibilists? In principle, it is. At the same time, note that this way of thinking about character virtues makes them appear very close to reliabilist virtues. And the trouble is that one of the most important divides among virtue responsibilists is between those who think that character virtues feature a reliability condition (e.g. Zagzebski ) and those who think that this would be a mistake (e.g. Montmarquet ). That is to say, there is a significant number of virtue responsibilists who think that character traits such as open-mindedness and intellectual courage can be virtues even though they don’t produce beliefs with a reliable truth to falsity (knowledge to ignorance) ratio. There is a bit of wiggle room for responsibilists here. It could be that character traits such as open-mindedness and intellectual courage aren’t virtues because they produce true belief (knowledge) reliably. In fact, they might not produce beliefs at all. Rather, the reason why they are virtues is that they are productive in generating opportunities to form true beliefs or knowledge (e.g. Sosa ). Even so, we imagine that many virtue responsibilists would be unhappy to sign up to this construal of responsibilist virtues. This is because they think that the already instrumentalist account of these virtues, according to which they are virtues because conducive to attaining the goal of enquiry, is mistaken. The question remains as to whether our construal of virtue theoretic epistemology can accommodate virtue responsibilists of this stripe. Here, in outline, is one attractive attempt at pulling this off, which circles back once more to eudaimonia or well-being. Here are the key thoughts: there is such a thing as epistemic well-being and it is multiply realisable. In particular, what epistemic well-being amounts to for fallible agents like us humans who are also very far from being omniscient might be quite different from what it amounts to for infallible omniscient agents,

Introduction: Virtue Theoretic Epistemology



say. Even if core responsibilist epistemic virtues are not essential to all forms of epistemic well-being – in particular, they aren’t essential to epistemic well-being in infallible omniscient agents – they are essential to epistemic flourishing in agents like us humans. In this way, the positive epistemic status of core responsibilist virtues can be explained and it can be done in a way that does not commit its champions to an instrumentalist account of character virtues. Finally, recall that, according to our proposed characterisation of virtue theoretic epistemology, to ϕ well epistemically is to be accounted for in terms of epistemic virtues. This is of course entirely compatible with the present proposal. In particular, returning once more to the idea that core responsibilist virtues are virtues of enquiry, we can now allow for the idea that if one enquires in a way that manifests these virtues, one enquires well in virtue of that, at least prima facie and pro tanto. What comes to light, then, is that our construal of virtue theoretic epistemology is broad enough to accommodate even those versions of responsibilism that explicitly resist an instrumentalist account of core responsibilist virtues. This volume brings together a number of contributions on virtue theoretic epistemology, both from the reliabilist and the responsibilist camps. The remainder of this Introduction gives a brief overview of the contributions to this volume. Heather Battaly’s ‘Closemindedness as an Intellectual Vice’ proposes an account of closed-mindedness as an unwillingness or inability to engage seriously with relevant intellectual options. It distinguishes between three different kinds of intellectual vice – effects vice, responsibilist vice and personalist vice – and argues that closed-mindedness can take each of these forms. In ‘Epistemic Virtues and Virtues with Epistemic Content’ Cameron Boult, Christoph Kelp, Johanna Schnurr, and Mona Simion take a closer look at the relationship between core character virtues in responsibilist virtue epistemology, such as curiosity, open-mindedness and intellectual humility on the one hand, and other types of virtue, such as moral or 

A note of caution. It is worth bearing in mind that this way of allowing non-instrumentalist versions of virtue epistemology to fall under the concept of virtue theoretic epistemology should not be understood as resuscitating the project of giving a virtue theory of epistemic normativity, if only one that is restricted to humans. The reason for this is that there remains good reason for thinking that there are things that have positive epistemic status essentially, such as knowledge and justified belief, that are not amenable for analysis in terms of core responsibilist virtues. Accordingly, the best we can hope for is a kind of pluralist view about epistemic normativity, which recognises a variety of things that have positive epistemic status essentially, including knowledge and epistemic flourishing.



 

prudential virtue on the other. They draw a distinction between epistemic virtues and virtues with epistemic content and use this distinction to argue that there is a significant set of virtues that are widely thought to be epistemic virtues which, on closer inspection, turn out to be moral virtues with epistemic content rather than genuine epistemic virtues. Fernando Broncano-Berrocal’s ‘Difficulty and Knowledge’ argues that virtue reliabilism provides a promising framework for accounting for the relationship between difficulty and knowledge. To this end, BroncanoBerrocal connects the virtue reliabilist framework with recent work on the notions of achievement and difficulty and uses the results to develop a novel account of knowledge, according to which knowledge is a special kind of challenge with varying degrees of difficulty. Tyler Burge’s influential work on epistemology takes centre stage in Peter Graham’s ‘What Is Epistemic Entitlement? Reliable Competences, Reasons, Inference, Access’. Graham argues that Burge’s position qualifies as a kind of virtue reliabilism, and provides a lucid guide to and elaboration of Burge’s often challenging views. John Greco’s ‘Knowledge-Producing Abilities’ investigates a range of accounts regarding the kind of ability that is necessary for knowledge. Greco considers the most promising views on the market in view of questions regarding the value of knowledge. He asks in what sense knowledge is valuable, and in what sense of ‘ability’ it is that success-from-ability gives knowledge its value? In the light of these questions, Greco defends a specific kind of virtue epistemology, which analyses knowledge-producing abilities as relative to environments and as entailing a safety condition. Jonathan Kvanvig’s ‘Virtue Epistemology, Two Kinds of Internalism, and the Intelligibility Problem’ introduces Wilfred Sellars’s difficult intelligibility problem; that is, the problem of offering an account of the relation between experience and belief that isn’t merely causal but sensemaking. Kvanvig argues that this problem is a hard one to solve for both internalist and externalist epistemologies, including propositionalist versions of internalism that were designed to solve it, and that virtue epistemology may well be the most attractive way forward on the externalist side. In ‘Knowledge Is Extrinsically Apt Belief: Virtue Epistemology and the Temporal Objection’, Anne Meylan defends the virtue epistemological account of knowledge as apt belief against an important objection. According to this objection, knowledge cannot be apt belief because beliefs are states and aptness is a property of actions rather than states. To address it, Meylan introduces a distinction between intrinsic and extrinsic normative properties and argues that the objection will work if and only if aptness is

Introduction: Virtue Theoretic Epistemology



taken to be an intrinsic normative property. She proposes an improved version of the virtue epistemological account of knowledge according to which knowledge is apt belief, where aptness is unambiguously construed as an extrinsic normative property. In ‘Explaining Knowledge’, Alan Millar develops an alternative to the traditional attempt of reductively analysing knowledge. Millar’s proposal is that we can shed light on knowledge in general by providing substantive accounts of the diverse ways in which subjects can be in cognitive contact with facts. In this regard, Millar focuses on perceptual knowledge and on knowledge from a perceived indicator. He argues that how one acquires knowledge can be relevant to explaining why what one has acquired counts as knowledge, and he offers a knowledge-first account of the relation between knowledge and justified belief. The nature of knowledge is also the topic of Duncan Pritchard’s ‘Antirisk Virtue Epistemology’, which examines how anti-risk epistemology fits into a wider virtue theoretic account of knowledge – one that replaces antiluck virtue epistemology with anti-risk virtue epistemology. He explains how the two views differ by examining some of the subtle but epistemologically significant differences between the notions of luck and risk, and he argues that a distinctively anti-risk virtue epistemology is preferable. In ‘Responsibilism within Reason’, Kurt Sylvan develops a novel responsibilist account of knowledge and justified belief. He argues that standard objections to attempts to analyse knowledge and justified belief in terms of character virtues do not strike at the heart of the responsibilist project. According to Kantian responsibilism, which is Sylvan’s own preferred version of the view, epistemically virtuous thought is thought that manifests respect for truth. And because manifesting certain reasonssensitive dispositions is necessary and sufficient for respecting truth, Kantian responsibilism takes epistemic virtues to coincide substantively with reasons-sensitive dispositions. REF ERE NCE S Aristotle. . The Nicomachean Ethics. Edited by T. Irwin. Indianapolis, IN: Hackett. Bentham, J. . An Introduction to the Principles of Morals and Legislation. Garden City, NY: Doubleday. Broome, J. . Weighing Goods. Oxford: Blackwell. Code, L. . Epistemic Responsibility. Hanover, NH: University Press of New England and Brown University Press. Greco, J. . Achieving Knowledge. Cambridge, UK: Cambridge University Press.



 

Hooker, B. . Ideal Code. Oxford: Clarendon Press. Hookway, C. . ‘How to Be a Virtue Epistemologist’, in DePaul, M. (ed.), Intellectual Virtue: Perspectives from Ethics and Epistemology. Oxford: Clarendon, –. Hursthouse, R., and G. Pettigrove. . ‘Virtue Ethics’, in Zalta, E. N. (ed.), The Stanford Encyclopedia of Philosophy. https://plato.stanford.edu/archives/ win/entries/ethics-virtue/. Accessed January , . Kant, I. . Groundwork of the Metaphysics of Morals. New York: Harper and Row. Kelp, C. . ‘Two for the Knowledge Goal of Inquiry’, American Philosophical Quarterly : –. . ‘Inquiry and the Transmission of Knowledge’, Philosophy and Phenomenological Research : –. Kvanvig, J. . The Value of Knowledge and the Pursuit of Understanding. Cambridge, UK: Cambridge University Press. Lynch, M. . True to Life. Why Truth Matters. Cambridge, MA: MIT Press. Mill, J. S. . Utilitarianism. New York: Oxford University Press. Millar, A. . ‘Why Knowledge Matters’, Proceedings of the Aristotelian Society, Suppl. Vol. : –. Montmarquet, J. . Epistemic Virtue and Doxastic Responsibility. Lanham, MD: Rowman and Littlefield. Plato. . The Republic. Cambridge, UK: Cambridge University Press. Pritchard, D., Millar, A. and Haddock, A. (eds). . The Nature and Value of Knowledge. Oxford: Oxford University Press. Ross, W. D. . The Right and the Good. Oxford: Oxford University Press. Sen, A. . ‘Utilitarianism and Welfarism’, Journal of Philosophy : –. Sidgwick, H. . The Method of Ethics. London: Macmillan. Singer, P. . Practical Ethics. Cambridge, UK: Cambridge University Press. Sosa, E. . Judgment and Agency. Oxford: Oxford University Press. Williamson, T. . Knowledge and Its Limits. Oxford: Oxford University Press. Zagzebski, L. . Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge, UK: Cambridge University Press.

 

Closed-Mindedness As an Intellectual Vice Heather Battaly

Mike believes that poor Americans are simply lazy. Like many upper-class Americans of his generation, Mike grew up believing that the United States is a land of opportunity where people would not be poor if they worked hard (Pimpare ). He has stuck with these beliefs throughout his adult life and is consistently unwilling to engage seriously with ideas or evidence to the contrary: he summarily dismisses any competing ideas that cross his path, without evaluating their merits. Accordingly, when the conversation turns to low wages as a cause of poverty, Mike deems it nonsense and shuts down, closing himself off. When he sees an article arguing that lack of opportunity contributes to poverty, he thinks it silly and scrolls past it. Mike recognizes that such ideas compete with his own and rejects them because they seem implausible. In short, Mike is closedminded, at least with respect to the causes of poverty in the United States. What is closed-mindedness, and if it is an intellectual vice, what makes it so? I intend this chapter to contribute to the developing field of vice epistemology, the goals of which include analyzing qualities – for instance, closed-mindedness or intellectual arrogance – that are likely to be intellectual vices, and explaining what makes such qualities intellectually vicious when they are (Battaly a; Cassam ). I adopt a working analysis of closed-mindedness as an unwillingness or inability to engage seriously with relevant intellectual options. Mike has one familiar species of closedmindedness: he is dogmatic. Specifically, he is unwilling to engage seriously with relevant alternatives to a belief he already holds. Though I will argue that the disposition of closed-mindedness is an intellectual vice in standard cases, including Mike’s, I do not assume that the disposition is always an intellectual vice. Instead, I treat the analysis of the disposition, 

This chapter is the second in a series of three papers. I defend these analyses of closed-mindedness and dogmatism in the first paper of the series (Battaly a). The third (Battaly b) argues that closed-mindedness can be an intellectual virtue.





 

and its status as an intellectual vice, as separate questions. The bulk of the chapter distinguishes between three different kinds of intellectual vice – effects vice, responsibilist vice, and personalist vice – and contends that closed-mindedness can take each of these forms.

. Working Analyses of Closed-Mindedness and Dogmatism Mike is unwilling to engage seriously with alternatives to his belief that poor Americans are simply lazy. Let’s use Mike’s case, which is a paradigm of both closed-mindedness and dogmatism, to home in on the key features of each of these two dispositions. Dismissing relevant alternatives to a belief – as Mike does – is one way to be both closed-minded and dogmatic. But it isn’t the only way to be closed-minded or even the only way to be dogmatic. Closed-mindedness is much broader than this: it is an unwillingness or inability to engage (or engage seriously) with relevant intellectual options. Even dogmatism, which is a subset of closed-mindedness, is a bit broader: it is an unwillingness to engage (or engage seriously) with relevant alternatives to a belief one already holds. To help us home in on these features of closed-mindedness and dogmatism, let’s explore Mike’s case in greater detail. We begin with four features of his case, none of which is necessary for closed-mindedness, and two of which – () and () – are also unnecessary for dogmatism. ()



Having beliefs about the topic. Mike already believes that poor Americans are lazy, and in dismissing relevant alternatives to this belief, he is being closed-minded. But closed-mindedness doesn’t require already having beliefs about a given topic. Consider Michelle who has no beliefs about the causes of poverty in the United States and is being confronted with evidence about this topic for the very first time. Even though she lacks any beliefs about the topic, Michelle can still arrive at an initial belief about the causes of American poverty by conducting a closed-minded inquiry. She may ignore or be oblivious to evidence pointing to structural causes of poverty (perhaps, because of an implicit bias that operates below the level of belief ). Even when agents don’t already have beliefs about a topic, they can be closedminded in the ways that they conduct inquiries and in the ways that they arrive at their initial beliefs. In contrast, dogmatism does require having extant beliefs about the given topic. It requires an object – a belief or dogma (as it were) – about which the agent is dogmatic. In

The analysis in this section is further defended in Battaly a.

Closed-Mindedness As an Intellectual Vice



addition to being closed-minded, Mike satisfies this further condition for dogmatism, whereas Michelle is closed-minded but not dogmatic. () The locus of ideas and evidence. Mike is closed-minded with respect to ideas and evidence that compete with a belief he already holds. Michelle is closed-minded in the way that she handles ideas and evidence in the process of arriving at a belief – she ignores relevant evidence. But Michelle’s case raises a broader point: one can also be closed-minded in the ways that one conducts inquiries more generally. For instance, one can be closed-minded with respect to which questions one asks, which sources one consults, and which methods one uses. Accordingly, the locus of closed-mindedness isn’t restricted to ideas and evidence. Its locus will include ideas and evidence, but will also include other intellectual options, such as relevant questions, sources, and methods. The locus of dogmatism isn’t restricted to ideas and evidence either. Consider dogmatic Michaela who, after acquiring her belief that poor Americans are lazy, willfully ‘puts her fingers in her ears.’ Michaela closes herself off to any further evidence that might come in, but also to any further sources, questions, or methods that might be relevant to the topic. () Engaging with intellectual options. In dismissing intellectual options that cross his path, Mike engages with those options, at least insofar as he recognizes and rejects them; for example, he recognizes that ‘lack of opportunity contributes to poverty’ competes with his beliefs, and rejects this option because it seems implausible to him. Granted, Mike doesn’t engage seriously with these options – he doesn’t evaluate the merits of competing ideas or the arguments for them. His engagement is entirely superficial. But closed-mindedness and dogmatism don’t even require this much, since one can be closed-minded or dogmatic by failing (in various ways) to engage with intellectual options. For starters, (a) one might ignore (rather than dismiss) intellectual options; broadly speaking, upon recognizing that an option competes with one’s beliefs, one might direct one’s attention elsewhere, refusing to further engage with it or render any judgment about its truth-value. Alternatively, (b) one might be oblivious to intellectual options – one might fail to recognize or notice them in the first place. For instance, one might be so invested in ‘the American dream’ and its concomitant claims about poverty and laziness that one fails to notice defeaters – one sees the 

Thanks to Sarah Wright.



  working poor not as industrious but as demanding ‘entitlements’ and not working hard enough. In the words of Robert Roberts and Jay Wood, oblivious agents are “perceptually rigid”: their “perceptual acuity is stuck within certain categories, outside of which [they are] blind or deaf” (: ). Agents can also be oblivious to relevant sources. For example, an agent might fail to see poor people as credible sources about poverty, perhaps because the agent thinks they will lie in order to protect their ‘government handouts’. On a broader scale, being oblivious to relevant sources can take the form of testimonial injustice – when an agent systematically fails to see women and people of color as credible (Fricker ). Agents can also fail to engage with intellectual options by (c) failing to seek out or generate them. Consider a conservative voter who fails to seek out moderate or progressive sources, who fails to look beyond their own echo chamber – who only looks for sources and evidence that confirm their conservative views about poverty. In failing to engage with relevant intellectual options, the above agents are being closed-minded, and where they fail to engage with relevant alternatives to extant beliefs, they are also being dogmatic.

()



Unwillingness. Mike is unwilling to engage seriously with relevant intellectual options. This willful refusal to engage (seriously) is arguably required for dogmatism. But it isn’t required for the broader category of closed-mindedness. One can be closed-minded by being unable, albeit willing, to engage seriously with intellectual options. Adapting an example from Wayne Riggs (), consider Oblivia, who believes that poor Americans are simply lazy. Unlike Mike, Oblivia is perfectly willing to engage seriously with intellectual options to this belief; but she is also systematically oblivious to relevant options and thus fails to engage with them. She fails to notice that poor Americans are industrious, fails to see poor Americans as credible sources about their own poverty, and so forth. She moves through the world unable to perceive relevant options. She may have passively inherited this impaired perception from her surrounding society (Fricker : ), or it may have been actively indoctrinated in her. Whatever its cause, Oblivia’s impaired perception makes her closed-minded. Importantly, our environments can make us closed-minded even when we don’t want to be so. There are two further points to note. First, people who are unable to engage with intellectual options because of bad luck in their environments or their constitutions

See also Mills .

Closed-Mindedness As an Intellectual Vice



are closed-minded but aren’t blameworthy (in the standard voluntarist sense) for coming to possess closed-mindedness. Second, as an inability, closed-mindedness can be an environmentally produced impairment, and can even result from a hard-wired impairment. Accordingly, closedmindedness need not be a character trait. This means that the disposition of closed-mindedness can be a character trait, or an environmentally produced impairment, or the result of a hardwired impairment. Some readers will find this worrisome, and so it warrants further explanation. The disposition of closed-mindedness will be an epistemic character trait when the agent’s unwillingness or inability to engage (seriously) with relevant intellectual options is grounded in their stable epistemic motives and values. This idea is Aristotelian in spirit. It assumes that an epistemic character trait must express who the agent is as a thinker and what they care about and value epistemically. Roughly, to say that closed-mindedness is an epistemic character trait in agent A is to say that closed-mindedness is grounded in A’s conception of epistemic value and A’s epistemic motivations, where the latter (A’s motivations) are informed by the former (A’s conception of value) and both are relatively stable (Battaly b). Accordingly, if A’s unwillingness or inability to engage with relevant options is grounded in A’s motivation to, for example, preserve her current beliefs and A’s commitment to, for instance, stability in her belief system, then closed-mindedness is an epistemic character trait in A. But if A’s closed-mindedness is not grounded in motivations that are informed by her conception of epistemic value, or if A lacks a conception of epistemic value, then closed-mindedness won’t be an epistemic character trait in A. To illustrate, the disposition of closedmindedness won’t be an epistemic character trait in A when A’s environment makes them testimonially unjust against their will and despite their good epistemic motives and values. Miranda Fricker contends that testimonial injustice can take the form of a perceptual prejudice that operates below the level of belief and motive and without the agent’s permission (: ). Due to the prejudiced perception unwittingly inherited from their environment, someone who is, at the level of belief and motive, a card-carrying feminist can still fail to see women as credible (Fricker : ). On my view, the card-carrying feminist is closed-minded – they are oblivious to relevant sources; yet in them, closed-mindedness is an environmentally produced impairment or bias, not an epistemic character trait. Closed-mindedness can also result from hard-wired impairments; such as when a cognitive impairment prevents agent A from engaging



 

with relevant options. Here, closed-mindedness is clearly not a character trait. It does not express A’s epistemic values or motives in any way. The upshot of all of this is that closed-mindedness is an unwillingness or inability to engage (seriously) with relevant intellectual options. Dogmatism is a subset of closed-mindedness: it is an unwillingness to engage (seriously) with relevant alternatives to a belief one already holds. There are three addenda. First, though the examples above feature beliefs that are false, one can also be closed-minded and dogmatic with respect to beliefs that are true. Second, an agent might have a domain-specific disposition to be closed-minded but lack a general disposition to be closed-minded. It is also possible for an agent to perform a closed-minded action as a one-off – to do what a closed-minded person would do – while lacking either a domain-specific disposition or a general disposition to be closed-minded. Third, the analyses above define closed-mindedness and dogmatism in terms of relevant intellectual options. Relevancy conditions clearly warrant exploration in their own right. I make a start on that project in “Closedmindedness and Dogmatism” (Battaly a), which sketches some candidates for relevancy conditions, including internalist and externalist candidates, and flags some concerns. This is tricky terrain, about which it would be premature to draw conclusions. However, the good news is that if we are internalists or externalists about epistemic relevancy, intellectual options like ‘+=,’ ‘The Holocaust never happened,’ and ‘The earth is flat’ will be epistemically irrelevant in ordinary environments – since these options are in fact false, and we (and our communities) believe reliably, and with good reason, that they are false. This means that in ordinary environments we won’t be closed-minded in ignoring these options, because they won’t be epistemically relevant.  





For further defense of the point that closed-mindedness can be an inability to engage with relevant options, see Battaly a. Thanks to Paul Bloomfield. I assume that there is no moral or pragmatic encroachment on conditions of epistemic relevancy. But I allow for the possibility that moral and pragmatic concerns might trump epistemic concerns. One might have moral or pragmatic reasons to engage with an agent who is arguing for an epistemically irrelevant claim. When one refuses to so engage, one isn’t closed-minded, but one might be callous or uncivil. On the distinction between ordinary and hostile epistemic environments, see Battaly b. George Orwell’s  describes a hostile environment, as does Mike Judge’s film Idiocracy. Some of us in the contemporary United States can still find (relatively) ordinary environments to occupy, though this will be much harder for some agents than others. It is possible for a single environment to be hostile for some agents (e.g. members of non-dominant groups) but not others (e.g. members of dominant groups). It is also possible for a single agent to move through different epistemic environments, some of which are hostile and others of which are (relatively) ordinary. This follows if we assume that either internalist or externalist conditions on relevancy are successful. They might not be. Thanks to Chris Kelp for suggesting that relevancy might involve modal features, and to John Greco for suggesting that it might involve social salience.

Closed-Mindedness As an Intellectual Vice



. Intellectual Vices: A Primer Thus far, I have adopted working analyses of the dispositions of closedmindedness and dogmatism. These analyses do not presuppose that closed-mindedness and dogmatism are intellectual vices. To figure out whether, when, and why closed-mindedness and dogmatism are intellectually vicious, we first need an account of intellectual vice. Intellectual vices are, roughly, cognitive dispositions that make us bad thinkers. Arguably, there is more than one way for cognitive dispositions to make us bad thinkers and more than one kind of intellectual vice. For starters, cognitive dispositions might produce a preponderance of bad epistemic effects (including, but not limited to, false beliefs), or prevent us from achieving good epistemic effects (e.g. true beliefs, knowledge, or understanding). Such dispositions are effects vices. Dispositions can be effects vices whether or not they involve bad epistemic character. Accordingly, we can explore whether closed-mindedness sometimes (perhaps, often) takes the form of an effects vice. We can ask: what bad epistemic effects might closed-mindedness produce? Specifically, what bad epistemic effects might it produce for the closed-minded agent, for other agents, and for the epistemic environment? The production of bad epistemic effects is one way for a cognitive disposition to be an intellectual vice. But it isn’t the only way. There are other ways for cognitive dispositions to make us bad thinkers, and other kinds of intellectual vice. A disposition could also (or instead) be a responsibilist vice; roughly, it could involve bad epistemic character – bad epistemic motives and values – for which the agent is blameworthy. Accordingly, we can explore whether closed-mindedness sometimes takes the form of a responsibilist vice. We can ask: when closed-mindedness is a responsibilist vice, which bad epistemic motives and values are associated with it? Are these bad motives and values unique to closed-mindedness, or can they be shared with other responsibilist vices? In what way is the closed-minded agent blameworthy for these bad motives and values? The distinction between effects vices and responsibilist vices should be relatively familiar to virtue epistemologists, since it corresponds to that between reliabilist and responsibilist virtues. Arguably, there is a third kind of intellectual vice: it involves bad epistemic character – bad epistemic motives and values – for which the agent is not blameworthy (in the 

On reliabilist virtues, see Sosa  and Greco ; on responsibilist virtues, see Baehr  and Zagzebski .



 

voluntarist sense). These are personalist vices. Personalist vices and virtues tell us who the agent is as a person or (since we are here restricting ourselves to the domain of the intellect) as a thinker. As with responsibilist vices, they must be bad epistemic character traits of the agent – they require bad epistemic motives and values. But, as with effects vices, the agent need not be blameworthy for possessing them. Thus, we can explore whether closed-mindedness can take the form of a personalist vice, and whether it sometimes, or even paradigmatically, takes this form in the indoctrinated. In Section ., I will argue that closed-mindedness can take each of these forms: it can be an effects vice, a responsibilist vice, or a personalist vice. Indeed, when it is a responsibilist or personalist vice, it is also reasonable to assume that it will (contingently) be an effects vice.

. The Distinction between Effects Vices, Responsibilist Vices, and Personalist Vices In order to see that closed-mindedness can take each of these forms of vice, we need further details about their distinguishing features. How are effects vices, responsibilist vices, and personalist vices different from one another? Let’s begin with effects vices. As with all intellectual vices, effects vices are cognitive dispositions that make us bad thinkers. Roughly, effects vices make us bad thinkers by producing bad epistemic effects. More accurately, they are cognitive dispositions that either consistently produce a preponderance of bad epistemic effects, or consistently fail to produce a preponderance of good epistemic effects. If we understand vices to be the contraries of virtues – such that one could simultaneously fail to have either – then effects vices will produce a preponderance of bad epistemic effects. Whereas, if we understand vices to be the contradictories of virtues – such that whenever one fails to have a virtue, one thereby has a vice – then effects vices will fail to produce a preponderance of good epistemic effects. Either way, there are five features that (together) mark out effects vices. First, a disposition won’t count as an effects vice unless it consistently produces a preponderance of bad epistemic effects, or consistently fails to   

Personalist vices are defended in Battaly b; Battaly and Slote ; Slote and Battaly . The distinctions in this section are further analyzed in Battaly b. Battaly : –. For criticism of the inversion thesis, see Crerar forthcoming.

Closed-Mindedness As an Intellectual Vice



produce a preponderance of good epistemic effects. In short, (F) producing bad epistemic effects (or failing to produce good ones) is required. This is because effects vices are instrumentally bad, and necessarily so. Consequently, an agent who has bad epistemic motives, but who has the good luck of living in an angel-world in which they happen to produce good epistemic effects (true beliefs, etc.) does not have effects vices, though they may well have responsibilist or personalist vices (see below). They do not have effects vices in that world because they do not produce bad epistemic effects. Second, (F) producing bad epistemic effects is also sufficient for a cognitive disposition to be an effects vice. Any cognitive disposition – be it a hard-wired impairment, an environmentally produced impairment, or an acquired character trait – that consistently produces a preponderance of bad epistemic effects (or fails to produce a preponderance of good epistemic effects) will be an effects vice. It will be an effects vice even when it isn’t intrinsically bad. Consequently, an agent who has intrinsically good epistemic motives, but consistently produces bad epistemic effects (false beliefs, etc.), perhaps because they are in a demon-world, has effects vices. Third, ~(F) though effects vices can be acquired dispositions – acquired character traits, acquired skills, environmentally produced impairments – they need not be (hence the ~). The acquired character trait of epistemic malevolence is (presumably) an effects vice. But, so is the hard-wired faculty of / vision. Accordingly, fourth, ~(F) an agent need not be blameworthy for their effects vices, since possession of them may be beyond the agent’s control. Relatedly, fifth, ~(F) effects vices need not be personal. ‘Personal’ dispositions, as I am using that term here, are character traits (at least partly) constituted by an agent’s stable motives and values (see Section .). Effects vices can be sub-personal. The effects vice of / vision is a case in point. Let’s compare responsibilist vices. Here, intellectual vices are epistemic character traits, over which the agent has some control and for which the agent is responsible (blameworthy). The five features that mark out responsibilist vices contradict the features of effects vices. First, ~(F) responsibilist vices do not require the production of bad epistemic effects. Granted, we can expect responsibilist vices to typically produce a preponderance of bad epistemic effects; the point is that they don’t conceptually require this.  

Baehr . Epistemic malevolence is also a responsibilist or personalist vice. Responsibilists might argue amongst themselves about this point. Zagzebski () thinks that intellectual virtues conceptually require the production of a preponderance of good epistemic effects (reliability). So, she may also think that intellectual vices conceptually require the production of a preponderance of bad epistemic effects. In contrast, Baehr’s internalist analysis of the responsibilist



 

This is because responsibilist vices need not be instrumentally bad; they must be intrinsically bad. Consequently, the agent in the angel-world – who produces good epistemic effects – can still have responsibilist vices that are grounded in their intrinsically bad epistemic motives (e.g. caring solely about preserving their beliefs). Second, ~(F) producing bad epistemic effects will be insufficient for responsibilist vice. This follows from each of its three remaining features. Third, (F) responsibilist vices are restricted to acquired dispositions; hard-wired impairments that produce bad epistemic effects are excluded. Fourth, relatedly, (F) responsibilist vices must be dispositions for which the agent is blameworthy, but one need not be blameworthy for producing bad epistemic effects. An agent can produce bad epistemic effects due to bad luck – due to constitutive or environmental factors that are beyond their control. Crucially, responsibilist vices require components over which the agent is thought to have some control – namely, their bad epistemic motives and values. Responsibilists think agents are blameworthy for coming to possess bad epistemic motives and values because they are blameworthy for failing to perform the requisite voluntary actions to avoid them, and/or blameworthy for performing the voluntary actions that produce them. Ergo, (F) responsibilist vices must be personal – they must be epistemic character traits that are (at least partly) constituted by the agent’s stable epistemic motives and values. The production of bad epistemic effects need not be personal – it may reflect the environment (e.g. in the demon-world) rather than the agent’s epistemic character. Personalist vices share three features with responsibilist vices, and two features with effects vices. Like responsibilist vices, (F) personalist vices must be personal – they, too, must be epistemic character traits that are (at least partly) constituted by the agent’s stable epistemic motives and values. Additionally, ~(F) neither responsibilist nor personalist vices conceptually require the production of bad epistemic effects. Personalist vices need not be instrumentally bad, but they must be intrinsically





vice of epistemic malevolence () allows for the possibility that this vice does not produce bad epistemic effects. Caring solely about preserving their own beliefs is an intrinsically bad motive, and remains so even if the agent possesses knowledge and is in an epistemically hostile environment – such as one that Orwell’s Ministry of truth has thoroughly polluted with falsehoods. Such an agent who is solely motivated to preserve their own beliefs, and is thus closed-minded in that hostile environment, will have an effects virtue (see Battaly b for defense of this point). But a trait can simultaneously be an effects virtue and a responsibilist or personalist vice (depending on the origin of the intrinsically bad motivation). Like responsibilists, personalists might argue amongst themselves about this point.

Closed-Mindedness As an Intellectual Vice



bad – the epistemic motives and values that they have as constituents will be intrinsically bad. Likewise, ~(F) the production of bad epistemic effects will be insufficient for personalist vice. Importantly, like effects vices, ~(F) personalist vices need not be dispositions for whose possession the agent is blameworthy (in the voluntarist sense). The agent need not have control over coming to possess them. Nor, technically, must personalist vices be acquired ~(F), though we can assume that they are acquired in humans. Human beings are not born with epistemic character traits or with stable epistemic motives and values – we acquire these over time. In short, personalist vices are personal but not blameworthy (in the voluntarist sense). They are full-blown epistemic character traits with all of the features that category entails. In this way, they are unlike / vision. Personalist vices are also unlike / vision insofar as they are acquired – at least in humans – whereas / vision is hard-wired. But personalist vices are like / vision in that agents do not have control over coming to possess them; in the case of personalist vices, they do not have control over coming to acquire them. In this vein, the concluding Section . on closed-mindedness as a personalist vice focuses on indoctrination.

.

Closed-Mindedness as an Effects Vice

We can now use the concept-mapping above to help us determine whether, when, and why closed-mindedness is an intellectual vice. Is closed-mindedness an effects vice? Recall from Section . that as an unwillingness or inability to engage (seriously) with relevant intellectual options, closed-mindedness can, but need not, be an acquired character trait. Accordingly, closed-mindedness can satisfy ~(F), ~(F), and ~(F) above. The key determinate, however, is (F). Whether or not closedmindedness is an acquired character trait, it would be an effects vice just as long as it produced a preponderance of bad epistemic effects (or failed to produce a preponderance of good epistemic effects). This is where we will focus our attention. I next enumerate some of the bad epistemic effects that closed-mindedness and dogmatism can produce, and some of the good epistemic effects that they can obstruct. I suggest that in standard cases, like Mike’s, it is reasonable to think that closed-mindedness will 

Davidson’s swampman can have personalist vices (Davidson ). Stewie Griffin – the ‘evil baby’ on the FOX television series Family Guy – also seems to have personalist vices.



 

produce more bad epistemic effects than good ones, rendering it an effects vice via (F). I begin with some bad epistemic effects that closedmindedness can produce for the closed-minded agent himself. (EV) Sustaining and strengthening false beliefs. The first thing to note is that closed-mindedness enables the agent who possesses it to sustain false beliefs that they already have. To illustrate, recall Mike. In closing himself off from relevant intellectual options, Mike is able to sustain his false belief that poor Americans are lazy. Similarly, in failing to look for sources outside our own epistemic bubbles, we may be sustaining false beliefs that we already have. Indeed, if Christopher Thi Nguyen () is correct, we may even be strengthening our confidence in those false beliefs. Nguyen argues that epistemic bubbles exhibit bootstrapped corroboration, whereby one’s beliefs are repeatedly, but not independently, corroborated by the posts and reposts of one’s ‘friends’ and trusted sources. All of this agreement and corroboration can lead us to mistakenly inflate our confidence in our beliefs. This misplaced confidence will be epistemically bad, whether our beliefs are true or false. (EV) Blocking true beliefs and knowledge. Second, closedmindedness can prevent the agent from acquiring true beliefs and knowledge. Mike’s dogmatism prevents him from acquiring true beliefs about the multifaceted causes of American poverty. The same holds for Michaela’s dogmatism (she ‘puts her fingers in her ears’). But, importantly, closedmindedness can prevent agents from acquiring true beliefs and knowledge even when they aren’t dogmatic and haven’t already made up their minds. Recall Michelle, whose first foray into the topic of American poverty prevented her from acquiring true beliefs about its structural causes. Or, consider an agent who gets all of their news from Facebook, and who fails to look for any relevant sources outside this feed (Nguyen ). If their feed exhibits serious gaps in news coverage, then their closed-minded failure to seek out alternatives will prevent this agent from gaining true beliefs about events that are absent from their feed. There is a further point to note. The tendency of closed-mindedness to obstruct the agent’s acquisition of true beliefs and knowledge will be especially important if we conceive of effects vices as contradictories of effects virtues – in other words, as failures to produce a preponderance of good epistemic effects. (EV) Acquiring new false beliefs. Third, closed-mindedness can lead the agent to acquire new false beliefs. For instance, we can assume that Michelle’s closed-mindedness leads her to acquire a new false belief about the cause of American poverty. Closed-mindedness can also compound and expand an agent’s extant system of false beliefs. It can lead agents to

Closed-Mindedness As an Intellectual Vice



doxastically double-down. Consider the belief system of Samuel A. Cartwright (–), an antebellum white American doctor and a published academic, who practiced medicine in Alabama, Mississippi, and Louisiana. Cartwright’s writings fall in the category of ‘scientific racism.’ He believed that slaves lacked agency. When confronted with contrary evidence – for example, their attempts to escape slavery and to resist their oppressors – Cartwright did what a dogmatic person would do. He did not judge escape attempts and resistance to be agential. Instead, he judged escape attempts to be manifestations of a mental disorder peculiar to slaves – “drapetomania” – irrational, racist nonsense, which he invented out of whole cloth. He made up a different disease – “dysaethesia aethiopica” – to explain the attempts of slaves to resist their oppressors, and in so doing, explicitly argued that although the behavior of slaves “appears as if intentional,” it is instead pathological (Cartwright ). These ‘diseases’ were utter nonsense. In Cartwright, we see that dogmatism with respect to a false belief can facilitate the production of a supporting system of false beliefs. One might object that Cartwright is being open-minded insofar as he is generating alternative explanations of the evidence. There are three points to note. First, it is correct that Cartwright’s dogmatism isn’t producing a supporting system of false beliefs all by itself. It is working in combination with other dispositions. But, second, Cartwright isn’t generating relevant options – drapetomania is not relevant. And so, the additional disposition he is exhibiting – perhaps some sort of racist fanciful thinking – isn’t open-mindedness. Finally, Cartwright’s unwillingness to revise his belief that slaves lack agency makes him dogmatic, and it does do even if he is willing to engage seriously with relevant alternatives. (EV) Epistemic opportunity costs. Fourth, closed-minded agents might pursue questions, projects, and inquiries that are epistemically irrelevant. For instance, a dogmatic racist might set out to explore the causes and cures of drapetomania, just as a dogmatic climate denier might set out to study the environmental benefits of burning coal. Granted,   

Thanks to Jennifer Saul for raising this worry about conspiracy theories. For further defense, see Battaly a. The Western Fuels Association explored (and endorsed) the idea that burning coal releases carbon dioxide, which encourages the photosynthesis of plants, leading to increased crop production (Oreskes and Conway : ). Texas Congressman Lamar Smith has made similar arguments. http://www.houstonpress.com/news/lamar-smith-explains-in-article-why-higher-carbondioxide-levels-are-a-good-thing-. These claims are debunked here: http://www.factcheck .org///co-friend-foe-agriculture/



 

dogmatism alone won’t be sufficient for the pursuit of irrelevant inquiries, since one can be dogmatic due to laziness and incuriosity – not all dogmatic agents will pursue irrelevant inquires. But those who do are wasting their epistemic resources and amassing epistemic opportunity costs – they could have been pursuing relevant inquiries instead. The dogmatic-because-incurious will likewise miss opportunities to pursue relevant inquiries, even though they don’t actively pursue irrelevant ones. This point will be especially important if effects vices are construed as contradictories of effects virtues. The above examines some of the bad epistemic effects closedmindedness can produce for the agent who possesses it. Let’s now turn to some of the bad epistemic effects that closed-mindedness can produce for other agents and for the epistemic environment. (EV) Credibility deficit. In being closed-minded with respect to sources, one might fail to see another agent as a source of knowledge when they are. In Fricker’s words, one might assign her a “credibility deficit” (: ). As a one-off, the harm done by a single instance of such closedmindedness may be relatively ephemeral. But when it is fueled by identity prejudice, such closed-mindedness will systematically track women and people of color across domains, taking the form of testimonial injustice. Fricker rightly argues that when an agent’s credibility is thus systematically overlooked, she is “wronged in her capacity as a giver of knowledge” (: ). The closed-minded-qua-testimonially unjust person will fail to see the agent as a source of knowledge: when the agent happens to cross his path, he won’t notice her; and on the rare occasion when he goes looking for credible sources, he will pass right over her. In my view, these are at least bad epistemic effects for the agent in question, whether or not they are also bad moral effects. (EV) Obstructing intellectual virtue and facilitating intellectual vice. Closed-mindedness that takes the form of testimonial injustice can have further bad effects. Women and people of color, whose credibility is repeatedly denied or overlooked, may come to doubt their own intellectual strengths (Fricker : ). This, in turn, can impede their ability to develop virtues such as intellectual courage and pride, and can facilitate their development of vices such as cowardice and servility. To illustrate, in losing confidence, these agents may under-own their intellectual strengths, thus eroding or impeding the virtue of intellectual pride and facilitating the vice of intellectual servility (Whitcomb et al. ). These, too, are bad effects that are decidedly epistemic, whether or not they are also moral.

Closed-Mindedness As an Intellectual Vice



(EV) Epistemic exclusion. Closed-mindedness that takes the form of testimonial injustice can also result in the exclusion of women and people of color from educational institutions. Closed-minded agents may fail to see women and people of color as educable and fail to see them as candidates for formal education and other epistemic resources. Closedminded policymakers may thus exclude women and people of color from schools and universities, and in so doing obstruct the ability of these agents to acquire true beliefs, knowledge, and intellectual virtues. (EV) Credibility Excess. The closed-minded agent who fails to look for sources outside his own epistemic bubble can assign too much credibility to his ‘friends.’ He may see them as sources of knowledge when they are not. As Nguyen () contends, agents in epistemic bubbles and echo chambers can overestimate the credibility levels of their trusted sources, sometimes resulting in runaway credibility levels. But, in assigning other agents too much credibility, does one really do them any harm? José Medina and Miranda Fricker both think one does: credibility excess can cause the agent to develop intellectual vices. Medina contends that the bad epistemic effects of credibility excess can be swift. In his words, during a sufficiently long and complex conversation, “we can perceive the speaker becoming arrogant and dogmatic” due to “the disproportionate authority he has been given”; that is, “he can be perceived to become a bully in the very course of the interaction” (: ). Fricker argues that the bad epistemic effects will be cumulative. She suggests that repeated instances of credibility excess can eventually cause agents to acquire the vice of intellectual arrogance (: ). Many of the bad epistemic effects enumerated above will be familiar from Fricker’s work on testimonial injustice. Yet, closed-mindedness need not take the form of testimonial injustice to produce bad epistemic effects for other agents. (EV) Epistemic Corruption. For instance, closed-minded agents who have the power to set the intellectual agenda for others – school boards, educators, journalists, governmental agencies – can (intentionally or unintentionally) create conditions that facilitate closed-mindedness (and additional vices) in others. They can be closed-minded in deciding which topics and methods to put in the curriculum (is critical thinking included?), which stories to print (are op-eds included?), and which proposals to fund (do they only fund proposals exploring the potential benefits  

Here, too, closed-mindedness will be working in combination with other dispositions. For Fricker, cases of cumulative credibility excess are at best peripheral cases of testimonial injustice. The central cases are those of credibility deficit.



 

of carbon dioxide?). In the words of Ian Kidd, their closed-mindedness can be “epistemically corrupting,” it can “encourage the development and exercise of epistemic vices” (Kidd ). (EV) Epistemic Pollution and Obfuscation. Finally, closedmindedness can lead to the intentional or unintentional pollution of the epistemic environment. Closed-minded agents who pursue irrelevant inquiries and do so sincerely – for instance, ‘true believers’ like Samuel A. Cartwright, can unintentionally disseminate falsehoods (about ‘drapetomania’) in their environments. Closed-minded agents who repost the claims of their ‘friends’ without seeking independent corroboration, can likewise populate their feeds with unwitting falsehoods. By inadvertently polluting their epistemic environments with false claims, these closedminded agents may also be obfuscating truths and knowledge, making them harder to find. Arguably, closed-mindedness can also lead to intentional pollution and obfuscation. Let’s assume that intentionally polluting the environment with claims one knows to be false, and intentionally concealing claims one knows to be true, involve dishonesty and deceit rather than closed-mindedness. Even so, dogmatic ‘true believers’ can still intentionally hide or erase competing ideas (that they incorrectly believe to be false) in an effort to prevent others from believing them. We see this repeatedly in book-burnings (in Nazi Germany, Mao’s Cultural Revolution), and also in the Environmental Protection Agency’s  decision to delete or move information about anthropogenic climate change from its main webpage into its archive (Friedman ). Dogmatic ‘true believers’ can also intentionally populate the environment with diversions – options that will divert agents who would otherwise endorse competing views. They may do this in a sincere effort to disseminate what they think are truths, to ‘control the message,’ or to manufacture doubt. For instance, they may publicize studies on the role of natural variability in climate change (Oreskes and Conway ). This gets us started on the plethora of bad epistemic effects that the disposition of closed-mindedness can produce. Recall (F), which tells us that effects vices are dispositions that produce a preponderance of bad epistemic effects, or fail to produce a preponderance of good epistemic effects. Does the disposition of closed-mindedness satisfy (F)? In ordinary   

 The below is defended in Battaly b. Ryan () addresses epistemic pollution. Especially when their ‘friends’ turn out to be Russian state operatives (Cloud ). When closed-mindedness leads to epistemic pollution and obfuscation, it is working in combination with other dispositions.

Closed-Mindedness As an Intellectual Vice



epistemic environments like ours, it is reasonable to think that it does. As a failure to seek out sources beyond our bubbles, closed-mindedness leads to misplaced confidence and credibility excess. As testimonial injustice, it erodes the intellectual virtues of other agents, facilitating intellectual vice. As dogmatism about a belief that is false, it results in the maintenance, strengthening, and compounding of false beliefs. Furthermore, in all of these forms, it obstructs the acquisition of knowledge. That is a heap of bad epistemic effects. Even at our most conservative, we can conclude that closed-mindedness sometimes, perhaps often, fails to produce a preponderance of good epistemic effects. The disposition of closed-mindedness also satisfies the conditions for an effects vice when the belief one is being closed-minded about is true. Suppose I believe that Trump will not be reelected in , and that this belief is true but doesn’t constitute knowledge (my belief is unjustified). In ignoring relevant options – I refuse to engage with relevant evidence to the contrary and sources outside my bubble – I am ignoring options that might lead me away from the truth. My closed-mindedness thus enables me to sustain a true belief. But, in ignoring these relevant options, I am not engaging with them seriously – I am not evaluating them on their merits. Accordingly, I may be sustaining a true belief that Trump won’t be reelected while blocking my ability to gain related epistemic goods, such as justified belief or knowledge that he won’t be reelected, or an understanding of why he won’t be reelected. Here, too, closed-mindedness may hinder the acquisition of knowledge. And, of course, it may do this while simultaneously producing many of the other bad effects mentioned above, including misplaced confidence. Here, too, we can at least conclude that closed-mindedness sometimes, perhaps often, fails to produce a preponderance of good epistemic effects. In sum, in ordinary epistemic environments, closed-mindedness is an effects vice.

. Closed-Mindedness as a Responsibilist Vice Closed-mindedness also sometimes takes the form of a responsibilist vice. To be a responsibilist vice, closed-mindedness must first be personal (F), that is, it must be an epistemic character trait, and one that is intrinsically 



Battaly b argues that in hostile epistemic environments, closed-mindedness doesn’t satisfy (F), and thus isn’t an intellectual vice in hostile environments; indeed, closed-mindedness is an intellectual virtue in hostile environments. See Callan and Arena : . I treat the case of closed-mindedness about knowledge in Battaly b.



 

bad (though not necessarily instrumentally bad ~(F)). It must also be a trait for which the agent is blameworthy (F). Let’s examine these features in turn. (F) Personal character trait. For starters, to be an epistemic character trait of any kind (bad or good), closed-mindedness must reflect who the agent is as a thinker and what he or she cares about and values epistemically. Reiterating the neo-Aristotelian point in Section ., the agent’s closed-mindedness must be grounded in his or her conception of epistemic value and his or her epistemic motivations, where the latter are informed by the former and both are relatively stable. In other words, he or she must be unwilling or unable to engage with relevant intellectual options because of his or her (stable) epistemic motives and values. This means that the unwitting closed-mindedness of the card-carrying feminist is not a responsibilist vice, since it is “flatly inconsistent” with his or her epistemic motives and values (Fricker : ). His or her closed-mindedness is not an epistemic character trait. Something similar holds for hard-wired impairments that result in closed-mindedness. To be a responsibilist vice, an agent’s closed-mindedness must be driven by his or her entrenched epistemic motives and values. ~(F) Intrinsically bad motives and values. Those entrenched epistemic motives and values must also be intrinsically bad. Recall that responsibilist vices need not be instrumentally bad – though responsibilist vices typically produce a preponderance of bad epistemic effects, they need not do so. They would be bad even if they didn’t produce bad effects. This is due to their intrinsic badness, which they get from the intrinsically bad motives and values that are their constituents. In other words, for an agent’s closed-mindedness to be a responsibilist vice, it must be driven by entrenched epistemic motives and values that are intrinsically bad. The agent must be unwilling or unable to engage seriously with relevant intellectual options because of their bad epistemic motives and values. Now, which intrinsically bad epistemic motives and values are associated with closed-mindedness? Are these bad epistemic motives and values unique to closed-mindedness, or are they shared with other responsibilist vices? For starters, an agent might be unwilling or unable to engage with relevant options because they consistently place too much value on, and care too much about, protecting and preserving their own worldview, while placing too little value on, and caring too little about, truth and independent thought. In short, they may prioritize stability and conservatism in their thinking over the value of knowledge. Presumably, epistemic

Closed-Mindedness As an Intellectual Vice



values and motives like these are de re bad, even when the agent pursues them because they think they are good. Indeed, valuing and rationally desiring epistemic stability and conservatism can even lead the closedminded person to falsely believe that they are doing the right thing in dismissing options. Because they falsely believe that stability and conservatism are good, the agent may wrongly judge a wide range of relevant options to be irrelevant, thinking it appropriate to dismiss them. They may even falsely believe that seriously engaging with these options – which are, from their perspective, irrelevant – would be naïve or fanciful rather than open-minded. In this manner, the vice of closed-mindedness can evade detection: it can be stealthy (Cassam : ) and “unconscious of itself” (Aristotle NE.b). As José Medina puts it, the agent can be “meta-blind”: blind to relevant options, and blind to the fact that they are closed-minded (: ). Closed-mindedness may have taken this form in Erik Jaentsch, a Nazi psychologist, whose  work The Antitype identified “open-mindedness as a symptom of psychological morbidity” (Kruglanski : ) and viewed “consistency, stability, firmness, and decision-making confidence” as “hallmarks of normalcy and good mental health” (Kruglanski : ). Jaentsch’s blindness ran so deep that he even characterized epistemic consistency, stability, and firmness as examples of “cognitive clarity” (Kruglanski: ). In Arie Kruglanski’s words: “For Jaentsch, the antitype represented a morbid psychological disposition and a menace to the purity of the German culture . . . His own conception of cognitive clarity as the epitomy [sic] of sanity was much in accord with the Nazi ideology” (: ). Jaentsch valued and desired firmness and confidence in belief, eschewing flexibility and openness. The entrenched epistemic values and motives described above – such as stability, protecting and preserving one’s beliefs – may strike us as paradigmatically closed-minded. But these aren’t the only epistemic values and motives that can ground the disposition of closed-mindedness. An agent might be unwilling or unable to engage with relevant options because they are epistemically lazy, incurious, or arrogant. They might even be unwilling or unable to engage with options because they have a de dicto motive to pursue the truth, but falsely believes that the truth is something they already have and can only get from their own worldview. In other words, there are many bad motives and values that can drive closed-mindedness. Mike (above) could have any of these entrenched motives (and others besides). Further, these motives and values need not be unique to closedmindedness and can be shared with other responsibilist vices. This means



 

that we can’t individuate closed-mindedness from other vices by appealing to motives. So, what does individuate closed-mindedness from other responsibilist vices? The answer is that whatever one’s bad epistemic motives and values may be, they must cause one to be unwilling or unable to engage seriously with relevant intellectual options. The above tells us that the intrinsically bad epistemic motives and values of a closed-minded agent can be de dicto good. There is no restriction to deliberately pursuing what one believes to be epistemically bad. Closedminded agents need not be epistemic villains. They can be motivated to get what they think are truths and knowledge, and, more generally, to pursue what they think is epistemically good. This means that contra Quassim Cassam (: ), closed-minded conspiracy theorists who are motivated to get what they think is knowledge, but who falsely believe that knowledge can only be gotten from their trusted ‘friends,’ still have intrinsically bad epistemic motives. Their motives are de re bad, even though they are not de dicto bad. Therefore, we shouldn’t exclude them from responsibilist vice on these grounds. (F) Blameworthiness. For closed-mindedness to take the form of a responsibilist vice, the agent must also be (at least partly) blameworthy for coming to possess it. For responsibilists, blame for coming to possess a trait requires (some degree of ) control over its development. Following Aristotle, it requires the agent to be blameworthy for failing to perform the voluntary actions that were needed to avoid the trait, and/or blameworthy for performing the voluntary actions that led to the trait. To illustrate, the agent will be responsible for their closed-mindedness if they, for instance, voluntarily chose to spend all their time attending Klan rallies and immersing themselves in the worldview of white nationalists, while 









See Adams , which analyzes the moral vice of malice in terms of the agent’s intentional opposition to what he or she correctly believes is good and their intentional pursuit of what they correctly believe is bad. For Adams, the motives of the malicious agent are both de re bad and de dicto bad. See Baehr : –, which argues that epistemic malevolence requires de dicto, though not de re, opposition to the epistemic good. It requires opposition to what one regards as epistemically good, whether or not one is correct. Thanks to Jason Baehr for discussion. See Aristotle NE.a and NE.a–, which argues that morally vicious people falsely believe that the actions they perform are good, and are motivated to perform them for that reason. For Aristotle, the motives of the morally vicious agent are de re, but not de dicto, bad. Cassam () rejects pluralism about intellectual vice, and rejects motivational (responsibilist) analyses of intellectual vice. He opts for a consequentialist analysis (a version of effects vice), whereby vices obstruct inquiry. Responsibilists endorse Aristotle’s picture of the acquisition of habits via the performance of like activities.

Closed-Mindedness As an Intellectual Vice



knowing that these activities were likely to erode their commitment to the truth. Or, less dramatically, an agent might have consistently failed to make the effort to look beyond their own epistemic bubble, knowing that this would contribute to the further entrenchment of their beliefs and could make them even less motivated to engage with alternatives in the future. Or, an agent might have repeatedly avoided lines of inquiry that would have challenged their views, knowing that this could contribute to a habit of favoring belief-stability and undervaluing independent thought. We can imagine warning our friends and ourselves off of these actions, and we can blame our friends and ourselves for performing them, given that we (and they) know better (or should have known better). As Jonathan Jacobs puts the point, the agent can sometimes “foresee . . . where a pattern of reaction, reasoning, and choice would or might lead” (: ). Indeed, agents might even be explicitly warned by teachers, friends, and family members about “where they are headed as a consequence of their actions” (Jacobs : ). Responsibilism successfully describes some cases of closed-mindedness, especially for agents who are raised in hospitable epistemic environments. In such environments, agents sometimes do have a degree of control over, and responsibility for, the epistemic character traits they acquire. To illustrate, should any students at the Intellectual Virtues Academy (IVA) come to possess closed-mindedness, they would be (partly) blameworthy. Teachers and students at IVA explicitly discuss the trait of openmindedness and its development, and students have opportunities to practice it and reflect on their progress. But we don’t need formal training in the intellectual virtues to be blameworthy. When it is working well, the U.S. public education system might render us capable of knowing which actions are likely to lead to closed-mindedness, and provide us with opportunities to practice open-mindedness. Accordingly, agents who are privileged and educated, like the fictional Mike above, will also be (partly) blameworthy for becoming closed-minded. Mike Enoch (– ), who seems to have been privileged and educated before he became a spokesperson for white supremacism, may be a real-life example of someone who is blameworthy for becoming closed-minded (Marantz ). The New Yorker reports that Enoch went to a diverse public high school in suburban New Jersey. His father is a retired English  

IVA is a public charter school in Long Beach, California. http://www.ivalongbeach.org/about/ about-iva, Accessed: December , . Thanks to Tracy Llanera for suggesting this example.



 

professor, and his grandfather reportedly “helped drive the K.K.K out of North Dakota” (Marantz ). Enoch is reported never to have completed a college degree, though he learned to code and became a programmer. According to the Southern Poverty Law Center (SPLC), Enoch endorsed Marxism, liberalism, and libertarianism before becoming a white supremacist (in his mid-s). The SPLC describes Enoch’s ideas as progressing “from the libertarian fringe to race realism and eventually extreme anti-Semitism, which [Enoch] partially ascribes [to] reading Kevin MacDonald’s Culture of Critique.” Arguably, Enoch is someone who knew (or should have known) that his actions were likely to lead to closed-mindedness, and though he could have acted differently, he did them anyway. We can note that the responsibilist vice of closed-mindedness will also often be an effects vice. For example, in white supremacists (who should have known better), the responsibilist vice of closed-mindedness produces a preponderance of bad epistemic effects. Recall that this is sufficient for rendering a disposition an effects vice and that effects vices can be epistemic character traits.

. Closed-Mindedness as a Personalist Vice Closed-mindedness can take the form of a responsibilist vice: some of us are blameworthy for acquiring the character trait of closed-mindedness. But there is at least one group of paradigmatically closed-minded people who are not blameworthy (in the voluntarist sense) for becoming closedminded – the indoctrinated. Consider children raised by Hitler’s Jugend or by ISIS. It isn’t just their epistemic behavior that is indoctrinated – they also acquire intrinsically bad epistemic motives and values. Their environments indoctrinate them to care about and value stability, conformity, and obedience in their thinking, and to judge open-minded thought as dangerous and weak. They emerge with integrated dispositions of action, motivation, and value: they ignore and suppress intellectual options, and do so because of the evaluative commitments and motives they have acquired. In short, they emerge with a personal character trait. In this manner, the closed-mindedness of the indoctrinated satisfies ~(F) and (F). But the indoctrinated are not blameworthy ~(F) for becoming closed-minded. At least, they aren’t blameworthy in 

https://www.splcenter.org/fighting-hate/extremist-files/individual/michael-enoch-peinovich, Accessed: November , .

Closed-Mindedness As an Intellectual Vice



the standard voluntarist sense, where blame requires control. As Robert Adams suggests, children raised by the Hitler Jugend “were victim[s] of [their] education” (: ). They acquired closed-mindedness (and a host of other reprehensible character traits) involuntarily as products of their environment. The Jugend graduate can’t be blamed (in the voluntarist sense) for coming to possess closed-mindedness, and so their closed-mindedness is not a responsibilist vice. Yet, it is still an intellectual vice of some kind. Undoubtedly, it produces a preponderance of bad epistemic effects, most notably for others, but also for the graduate themselves. Thus, we can assume that it is an effects vice. But the point is that it would still be an intellectual vice even if didn’t produce bad epistemic effects. It is an intellectual vice insofar as it is an epistemic character trait that is grounded in the graduate’s intrinsically bad motives and values. It shouldn’t matter that they acquired this disposition via indoctrination and can’t be blamed for coming to possess it. In defense of this point, suppose we come across two people, A and B, with identical dispositions. Both consistently dismiss and suppress relevant intellectual options – both are closed-minded. Both consistently value stability and conformity in thought, and judge open-minded thinking to be weak and dangerous. Moreover, both are consistently motivated to dismiss and suppress intellectual options in pursuit of these ends. I submit that in both A and B, closed-mindedness is an intellectual vice. We don’t need to know anything about their development to make this determination. Suppose we later discovered that A’s closed-mindedness was the result of indoctrination (the Jugend graduate), whereas B’s was the result of conscious neglect (perhaps, Mike Enoch as described above). What we should not do in this situation is withdraw our judgment that A has an intellectual vice. After all, A consistently dismisses and suppresses intellectual options because of their intrinsically bad epistemic values and motives. Furthermore, A’s psychology is just like B’s: both are executing their evaluative plans. Thus, if B has the vice of closed-mindedness, so should A. The provenance of their closed-mindedness shouldn’t matter. Consequently, closed-mindedness is still an intellectual vice in the indoctrinated, even though the indoctrinated aren’t blameworthy (accountable) for coming to possess it. But its viciousness isn’t fully captured by either of our previous conceptions of intellectual vice. In taking the closed-mindedness of the 

For further defense of this point, see Battaly a: .



 

indoctrinated to be a mere effects vice, we would leave out an important part of what makes it vicious – the intrinsically bad motives and values that drive it. We need a third conception: personalist vice. This conception is a via media between effects vice and responsibilist vice. As we saw above, it shares much with the latter. It requires intellectual vices to be personal traits of epistemic character (F) – to be grounded in the stable epistemic values and motives of the agent. It also requires those epistemic values and motives to be intrinsically bad ~(F). But, like effects vice, it rejects control and blameworthiness as a necessary condition ~(F): a disposition can count as a personalist vice even though the agent lacks control over, and isn’t blameworthy (in the voluntarist sense) for, its possession. I submit that in the indoctrinated, closed-mindedness is a personalist vice. Their closed-mindedness is (at least partly) intrinsically bad, due to the intrinsic badness of the epistemic motives and values that drive it. Bad epistemic effects aren’t the only thing that make it bad – it would be bad even if it didn’t produce bad epistemic effects. As Adams puts a related point: “It matters . . . what we are for and what we are against, even if we do not have the power to do much for or against it, and even if it was not by trying that we came to be for or against it” (Adams : ). Personalist vice accounts for this intrinsic badness, effects vice does not. Nor are the indoctrinated blameworthy (accountable) for becoming closed-minded. Personalist vice accounts for this, responsibilist vice does not. Rest assured, the indoctrinated won’t be let off too easy. Even if they aren’t blameworthy in the voluntarist sense, which requires control, they may well be blameworthy in a non-voluntarist sense. If non-voluntarist analyses of moral responsibility are viable, there will still be plenty of blame to go around. In closing, I have argued that in standard cases, including Mike’s, closed-mindedness is an intellectual vice. Closed-mindedness can take the form of an effects vice, a responsibilist vice, or a personalist vice. Indeed, it is reasonable to assume that closed-mindedness typically takes  





Battaly b; Battaly and Slote ; Slote and Battaly . Thanks to Michael Slote and Chris Kelp for helping me see the importance of personalist vice. Producing bad epistemic effects will be neither conceptually necessary nor conceptually sufficient for personalist vice, though personalist vices will often produce bad epistemic effects. Personalist vice satisfies ~(F) and ~(F). It may even be a personalist vice in people who haven’t been indoctrinated, if we have less responsibility for the possession of our character traits than we think. It is worth exploring whether responsibilist vices are the norm, and personalist vices the exception, or whether the reverse is true. See Sher , . See, for instance, Sher , , Talbert , and Watson .

Closed-Mindedness As an Intellectual Vice



one or more of these forms in ordinary epistemic environments. Is closed-mindedness always an intellectual vice? I suspect that even if closed-mindedness is typically an intellectual vice of one or more of the above kinds, it is not always an intellectual vice. There may even be conditions in which closed-mindedness is an intellectual virtue. REF ERE NCE S Adams, R. . “Involuntary Sins,” Philosophical Review (): –. . A Theory of Virtue. Oxford: Oxford University Press. Aristotle. . Nicomachean Ethics. In J. Barnes (ed.), The Complete Works of Aristotle. Princeton, NJ: Princeton University Press. Baehr, J. . “Epistemic Malevolence,” in H. Battaly (ed.), Virtue and Vice, Moral and Epistemic. Malden, MA: Wiley-Blackwell, –. . The Inquiring Mind. New York: Oxford University Press. Battaly, H. . “Varieties of Epistemic Vice,” in J. Matheson and R. Vitz (eds), The Ethics of Belief. Oxford: Oxford University Press, –. a. “Developing Virtue and Rehabilitating Vice,” Journal of Moral Education (): –. b. “Epistemic Virtue and Vice: Reliabilism, Responsibilism, and Personalism,” in C. Mi, M. Slote, and E. Sosa (eds), Moral and Intellectual Virtues in Western and Chinese Philosophy. New York: Routledge, –. a. “Testimonial Injustice, Epistemic Vice, and Vice Epistemology,” in I. J. Kidd, G. Polhaus, J. Medina (eds), The Routledge Handbook of Epistemic Injustice. New York: Routledge, –. b. “Intellectual Perseverance,” Journal of Moral Philosophy (): –. a. “Closed-mindedness and Dogmatism,” Episteme (): –. b. “Can Closed-mindedness be an Intellectual Virtue?” Royal Institute of Philosophy Supplements : –. Battaly, H. and M. Slote. . “Virtue Epistemology and Virtue Ethics,” in L. Besser-Jones and M. Slote (eds), Routledge Companion to Virtue Ethics. New York: Routledge, –. Callan, E. and D. Arena. . “Indoctrination,” in H. Siegel (ed.), The Oxford Handbook of Philosophy of Education. Oxford: Oxford University Press, –. 

See Battaly b. Thanks to Scott Aiken, Teresa Allen, Paul Bloomfield, Quassim Cassam, Jeremy Fantl, Branden Fitelson, Trystan Goetze, Heidi Grasswick, John Greco, Adam Green, Daniel Howard-Snyder, Frances Howard-Snyder, Hud Hudson, Ian J. Kidd, Tracy Llanera, Michael Lynch, Christopher Thi Nguyen, Josh Rasmussen, Clifford Roth, Matt Stichter, Alessandra Tanesini, Lynne Tirrell, Rico Vitz, Ryan Wasserman, Lani Watson, Sarah Wright, Dennis Whitcomb, David Woodruff, Zhu Xu, and audiences at the New York-China Epistemology Conference (Fordham ), the  Central APA, the Epistemic Harms and Wrong Conference (Sheffield ), Azusa Pacific University, Vanderbilt University, Haverford College, and the University of Connecticut. I am especially grateful to Chris Kelp for a terrific set of comments on this chapter.



 

Cartwright, S. . “Diseases and Peculiarities of the Negro Race,” De Bow’s Review, vol. . www.pbs.org/wgbh/aia/part/ht.html. Accessed: October , . Cassam, Q. . “Stealthy Vices,” Social Epistemology Review and Reply Collective : –. . “Vice Epistemology,” The Monist : –. Cloud, D. S. . “Facebook tells Congress that  Million Americans may have seen Russia-linked Ads,” Los Angeles Times. www.latimes.com/nation/ la-na-russia-tech--story.html. Accessed: December , . Crerar, C. . “Motivational Approaches to Intellectual Vice,” Australasian Journal of Philosophy (): –. Davidson, D. . “Knowing One’s Own Mind,” Proceedings and Addresses of the American Philosophical Association : –. Fricker, M. . Epistemic Injustice. Oxford: Oxford University Press. Friedman, L. . “EPA Scrubs a Climate Website of ‘Climate Change’.” New York Times. www.nytimes.com////climate/epa-climate-change.html. Accessed: December , . Greco, J. . Achieving Knowledge. Cambridge, UK: Cambridge University Press. Jacobs, J. . Choosing Character. Ithaca, NY: Cornell University Press. Judge, M. . Idiocracy (film). Kidd, I. J. . “Epistemic Corruption and Education,” Episteme. https://doi .org/./epi... Accessed: December , . Kruglanski, A. W. . The Psychology of Closed-Mindedness. New York: Taylor and Francis. Marantz, A. . “Birth of a White Supremacist.” New Yorker. www.newyorker .com/magazine////birth-of-a-white-supremacist. Accessed: November , . Medina, J. . The Epistemology of Resistance. Oxford: Oxford University Press. Mills, C. S. . “White Ignorance,” in R. N. Proctor and L. Schiebinger (eds), Agnotology. Stanford, CA: Stanford University Press, –. Nguyen, C. T. . “Escape the Echo Chamber,” Aeon. https://aeon.co/essays/ why-its-as-hard-to-escape-an-echo-chamber-as-it-is-to-flee-a-cult. Accessed: May , . Oreskes, N. and E. M. Conway. . Merchants of Doubt. New York: Bloomsbury. Orwell, G. . . New York: Harcourt Brace. Pimpare, S. . “Laziness isn’t why people are poor. And iPhones aren’t why they lack health care.” Washington Post. www.washingtonpost.com/postevery thing/wp////laziness-isnt-why-people-are-poor-and-iphones-arentwhy-they-lack-health-care/?utm_term=.fbfa. Accessed: October , . Riggs, W. . “Open-mindedness, Insight, and Understanding,” in J. Baehr (ed.), Intellectual Virtues and Education. New York: Routledge, –. Roberts, R. and W. J. Wood. . Intellectual Virtues. New York: Oxford University Press.

Closed-Mindedness As an Intellectual Vice



Ryan, S. . “Epistemic Environmentalism,” Journal of Philosophical Research (): –. Sher, G. . In Praise of Blame. Oxford: Oxford University Press. . Who Knew? Responsibility without Awareness. Oxford: Oxford University Press. Slote, M. and H. Battaly. . “Sentimentalist Virtue Epistemology: The Challenge of Personalism,” in N. Snow (ed.), The Oxford Handbook of Virtue. New York: Oxford University Press, –. Sosa, E. . A Virtue Epistemology. New York: Oxford University Press. Talbert, M. . Moral Responsibility. Cambridge, UK: Polity. Watson, G. . “Two Faces of Responsibility,” in G. Watson, Agency and Answerability. Oxford: Oxford University Press, –. Whitcomb, D., H. Battaly, J. Baehr, and D. Howard-Snyder. . “Intellectual Humility: Owning Our Limitations,” Philosophy and Phenomenological Research (): –. Zagzebski, L. . Virtues of the Mind, Cambridge, UK: Cambridge University Press.

 

Epistemic Virtues and Virtues with Epistemic Content Cameron Boult, Christoph Kelp, Johanna Schnurr and Mona Simion

.

Introduction

A prominent project in responsibilist virtue epistemology is to develop “maps” or “perspicuous representations” of intellectual virtues, such as intellectual humility, intellectual courage, open-mindedness, and curiosity. The basic idea is to develop empirically grounded characterizations of epistemically admirable or praiseworthy character traits and to use these characterizations as guidance in a kind of regulative epistemology, sometimes with the aim of informing education theory. To take a few examples, Jason Baehr characterizes the “open-minded” person as someone who is “willing . . . to transcend a default cognitive standpoint in order to . . . take seriously a distinct cognitive standpoint” (: ). Heather Battaly considers the “epistemically humble” person as someone who is “disposed to recognize her own fallibility, and to recognize and value the epistemic abilities of others” (Battaly : ). James Montmarquet describes the “epistemically courageous” person as someone who has the disposition to “persevere in the face of opposition from others (until one is convinced that one is mistaken)” (Montmarquet : ). Roberts and Wood characterize the “epistemically autonomous” person as someone who has the “proper ability to think for herself and not be . . . improperly dependent on or influenced by others” (Roberts and Wood : ). In many ways, this work is an exciting new development in epistemology. For example, it highlights possible avenues for widening the traditional epistemological project. It also reveals some of the complex ways in which the disciplines of epistemology and ethics seem to overlap. However, regarding the latter point and, in particular, apropos the question of whether, or to what extent, a given intellectual virtue counts as distinctively epistemic, as opposed to moral or prudential, we think the project requires careful handling. After all, the notion of virtue is familiar first and foremost from ethics, and the way in which it might be useful for 

Epistemic Virtues and Virtues with Epistemic Content



epistemology is an open question. Moreover, it’s not clear what relation the virtues just mentioned have to more traditional moral virtues, for example. Indeed, it seems plausible that some of them just are paradigm cases of garden-variety moral virtues, as opposed to some other kind of virtue. We think that a fully worked out account of the intellectual virtues should be able to clearly address this issue. It seems central to our understanding of the relevance of an investigation into some virtue or another for epistemology. It also seems central to our understanding of the scope of epistemology itself. Our aim in this chapter is twofold. First, we defend a particular way of individuating virtues, including intellectual ones. Second, we will also argue that this way of individuating virtues gives us reason to be somewhat more cautious in our claims to make progress in epistemology by investigating intellectual virtues: the number of intellectual virtues that count as epistemic is considerably smaller than the current debate on virtue responsibilism (henceforth VR) would have us think. Here is the plan. In Sections . and ., we look at two ways of individuating epistemic normative notions – “content individuation” and “value individuation” – and argue that the latter is preferable. Crucially, value individuation supports a distinction between genuine epistemic virtues and non-epistemic (but, e.g., ethical, prudential, etc.) virtues with epistemic content. In Section . we look at a number of virtues that have been investigated in the literature on VR. We argue that several of them fail to qualify as genuinely epistemic virtues. Rather, they are plausibly moral virtues with epistemic content. In Section . we conclude.

.

Content Individuation

How can we individuate distinctively epistemic virtues? In order to answer this question, we start by looking into the way in which philosophers in the epistemic norms literature individuate those epistemic norms, and take it from there. Let's first consider the following proposal: Content Individuation (CI) If a norm N concerns epistemic features required for permissible ϕ-ing, then N is an epistemic norm.

It is safe to say that, if there is such a thing as a received view in the epistemic norms literature concerning what that literature is theorizing 

We will get back to the issue of how to type virtues in Sections . and ..



, ,   

about in the first place, it is CI. Many philosophers, for instance, when they ask what the epistemic norm for ϕ-ing is, take themselves to be asking, roughly, how much epistemic warrant one needs for proper ϕ-ing. For example, here is Jennifer Lackey (arguing that cases of isolated second-hand knowledge show that knowledge is not the epistemic norm of assertion): It should be emphasized that it is clear that the problem with the agents in the above cases is that it is not epistemically appropriate for them to flat-out assert that p . . . One reason this is clear is that the criticism of the agents concerns the grounds for their assertions. (Lackey : , emphasis in original)

Thus, according to Lackey, insofar as norms concern epistemic grounds, they will be genuine epistemic norms. On a similar note, here is Ishani Maitra (on a view about the nature of assertion that she takes to be widely endorsed): Assertions are governed by an alethic or an epistemic norm – that is, a norm that specifies that it is appropriate to assert something only if what is asserted is true, or justifiably believed, or certain or known. (Maitra : )

There is very good methodological reason to endorse CI: it is both simple and user friendly. CI provides a neat and straightforward way to individuate epistemic norms, ensuring that the debate can be framed on common terminological ground. Most importantly for present purposes, CI might help us to individuate distinctively epistemic virtues. After all, if epistemic norms are individuated by content, as CI has it, it is independently plausible that the same goes for epistemic virtues. Virtue Content Individuation (VCI) If a virtue V concerns epistemic features required for virtuous ϕ-ing, then V is an epistemic virtue.

But of course, the sorts of virtues that the VR literature has focused on, including open-mindedness, intellectual humility, curiosity, epistemic courageousness, and temperance, clearly concern epistemic features. By VCI, these virtues are epistemic virtues. In fact, there is yet another way of getting the same result. To see how, note that virtues are widely taken to be normative. The following is an attractive way of capturing this thought: 

For explicit endorsements see, e.g., Benton , Lackey , Maitra , Brown . For implicit assumptions to this effect, see Gerken , Hawthorne , Littlejohn and Turri .

Epistemic Virtues and Virtues with Epistemic Content



Normative Charge of Virtues (NCV) One’s actions and states T-ought (morally, epistemically, etc. ought) to manifest T-virtuous (morally, epistemically, etc. virtuous) character traits.

By NCV, virtues are associated with oughts. Note also that there is a type correspondence between virtues and associated norms. For instance, if the virtue to be manifested is a moral virtue, the ought at issue in NCV is a moral ought, if the virtue to be manifested is an aesthetic virtue, the ought at issue in NCV is an aesthetic ought, and so on. NCV gives us the result that virtues of a certain type are associated with oughts of the same type. Now we may expect to use our recipe for individuating epistemic norms to home in on distinctively epistemic virtues. After all, given that types of virtues are by NCV associated with the corresponding types of norms, if we have a recipe for individuating types of norms then we might be able to use it to individuate the associated virtues as well. In particular, we get the result that if the norm at issue in NCV is epistemic, then so is the associated virtue. And since the sorts of virtues that the VR literature has focused on are associated with norms that concern epistemic features and since, by CI, norms that concern epistemic features are epistemic norms, once more we arrive at the result that the associated virtues are indeed epistemic virtues. There are thus two ways of arguing that the virtues discussed in the VR literature are genuinely epistemic virtues. And while it might be thought that this is all entirely as it should be, on reflection there is reason to think otherwise. To see why, note that both arguments rely on CI for their motivation. The first uses CI to motivate VCI, which is key to that argument, and the second uses CI expressly in the relevant derivation. The trouble is that there is excellent reason to think that CI is false. More specifically, as we will argue momentarily, first, CI does not generalize in the right way and, second, it is intuitively inadequate. To see why CI doesn’t generalize as it should, let’s get the generalization on the table. Here is what it looks like:



NCV requires some noteworthy qualifications. First, it’s quite plausible that it needs to be restricted to relevant actions, that is, those that can be performed in a way that manifests the relevant virtue. For instance, while it is plausible that I ought to be generous in my dealings with friends, it does not seem to be the case that I ought to be generous in sitting down. Second, the oughts at issue here are defeasible. For instance, if I have overriding reason not to be generous in my dealings with friends (perhaps because someone threatens to kill me if I am), it is not the case that I ought to be generous, at least not all things considered. Since these qualifications are of little consequence for the purposes of this chapter, in what follows we will take them as read.



, ,    Generalized Content Individuation (GCI) If a norm N concerns features of type T required for permissible ϕ-ing, then N is a norm of type T.

There is excellent reason to think that GCI is false. Consider, for instance, traffic norms: driving one’s car within city bounds will surely be subject to whatever the local traffic regulations have to say. Imagine that the relevant traffic norm forbids one from driving faster than  mph (call it ‘DRIVE-’). However, imagine also that a terrorist group has placed a bomb in the centre of town and you are the only one able to diffuse it. In order to get there in time, you must break the traffic norm and drive  mph. Somewhat superficially, we might imagine that the following instance of a more general moral norm applies: when millions of people’s lives are at risk, and you have the ability to save them only if you drive at a certain speed, you morally ought to drive at that speed (call it ‘DRIVETO-SAVE’). Clearly, in the case at hand, DRIVE-TO-SAVE overrides DRIVE- and renders driving  mph the all things considered proper thing to do. Most importantly for present purposes, note that what we have before us is a case in which a distinctively moral requirement, DRIVE-TO-SAVE, has traffic-related content: it regulates the morally (and all-things-considered) appropriate speed. If that is the case, however, it is evident that just because a norm has traffic-related content – just because it regulates the appropriate speed – it does not follow that it is a traffic norm. GCI fails for traffic norms. What’s more, the traffic case is hardly isolated. Similar examples can be found for many types of normativity. For example, imagine a case of a school that operates a norm according to which students must wear pink shirts and red trousers. Just because the school’s norm regulates a fashionrelated feature – that is, color of clothing – it need not follow that it is a fashion norm. Or consider a case of a chess player who is offered a million dollars if they move the rook diagonally during a leisurely game of chess. In this case, our chess player prudentially ought to move the bishop diagonally. But, again, from the fact that we are dealing with a norm that has chess-related content, it does not follow that the norm is a chess norm. In 

There are many problems with CI that we will not discuss here. One such problem is: how is the typing supposed to work for biconditional norms: “Hit the emergency brake if and only if you see someone stuck on the escalator.” Is this a norm for permissible brake-hitting? If so, it has features of accident-witnessing as permissibility requirement. Or is it a norm for permissible accidentwitnessing? If so, it has features of brake-hitting as permissibility requirements. Is it both and does it have both as permissibility requirements and so would be classed as a norm of two types?

Epistemic Virtues and Virtues with Epistemic Content



fact, it is clear that it isn’t. There are many examples of norms with X-related content that are not X norms. Hence, GCI must be false. What about CI? Given that GCI does not hold, there is reason to believe that the same goes for epistemic norms: just because a norm regulates the appropriate degree of an epistemic feature, it need not follow that it is an epistemic norm. It may indeed happen to be an epistemic norm; but it may also be the case that it is a norm of a different nature – say, a prudential or moral norm – with epistemic content. It may be a norm of some other kind that simply happens to regulate the (morally, prudentially, etc.) proper degree of an epistemic feature. To see this, consider the following examples: SING. One must sing only songs one knows. JUMP. One must not jump in lakes unless one knows how to swim. ASSERT. One must assert only what one knows. BELIEVE. One must believe only what one has sufficient evidence for. All four norms have epistemic content. Accordingly, CI will predict that they are all distinctively epistemic norms. However, that is intuitively implausible. In fact, it is widely agreed that while ASSERT and BELIEVE are clearly epistemic norms, SING and JUMP are not. Instead, SING is plausibly an aesthetic and/or prudential norm and JUMP is plausibly a prudential norm, although both have epistemic content. What transpires, then, is that just because a norm has epistemic content, it does not follow that it is an epistemic norm. Just because it concerns epistemic features – in other words, what epistemic position one needs to be in in order to permissibly ϕ – it does not follow that it concerns distinctively epistemic permissibility. No, it can also regulate an epistemic feature required for prudentially, morally, aesthetically – and so on – permissible ϕ-ing. What CI allows us to home in on are norms with epistemic content rather than genuinely epistemic norms. If all of this is right, however – that is, if CI fails as a criterion for individuating genuinely epistemic norms – we should also be suspicious of  

For an argument that ASSERT is a genuine epistemic norm and thus unlike SING and JUMP, see, e.g., Kelp . One legitimate question at this point is: where does this discussion leave the epistemic norm for action and/or practical reasoning? A full answer to this question falls outside the scope of this chapter. One of us has argued elsewhere (Simion ) in quite some detail that, in virtue of the distinction between epistemic norms and mere norms with epistemic content, it is a mistake to lump together the epistemic propriety of practical reasoning with that of action. Practical reasoning is governed by genuine epistemic norms, action is not. For a similar view, see Fantl and McGrath .



, ,   

the ways of individuating epistemic virtues that it serves to motivate. For instance, since CI fails to distinguish between epistemic norms and norms (moral, prudential, aesthetic, etc.) merely with epistemic content, we may legitimately wonder whether VCI correspondingly fails to distinguish between genuinely epistemic virtues and virtues of different stripes (e.g. moral, prudential) with epistemic content. And of course, this worry is only more pressing for the individuating recipe that relies on CI directly. Indeed, on closer inspection, one encounters such cases in the literature. Consider, for instance Roberts and Wood’s example of the intellectual virtue of epistemic temperance: In Toni Morrison’s novel Beloved, the slave-narrator comments on a slaveowner who was a cut above the average, that he looked away when the slave women were nursing their infants. This intentional foregoing of acquaintance expresses respect for the women’s privacy and a sense of the limits that human proprieties set to appropriate knowledge . . . To be an indiscriminate ogler is a trait of bad intellectual character, a failure of discipline of the will to know. (Roberts and Wood : )

Although this virtue is associated with an epistemic feature, it is not at all clear that it qualifies as a genuinely epistemic virtue rather than a virtue of some other denomination that has epistemic content. After all, by the lights of the authors themselves, this particular virtue expresses “respect for the women’s privacy and a sense of the limits that human proprieties set to appropriate knowledge” (Roberts and Wood : , emphasis added). As such, it is more natural to read it as a kind of moral virtue, albeit one that has something to do with the epistemic domain – a moral virtue with epistemic content. Indeed, epistemic temperance through the foregoing of acquaintance looks like an example of a moral virtue, the exercise of which limits the gathering of information and knowledge. And from a purely epistemic perspective, at any rate, this seems like precisely the opposite of what the exercise of an epistemic virtue would be apt to limit (see also Section . below). Let’s take stock. The challenge this chapter aims to meet on behalf of VR is to find a criterion for normative typing that will allow VR to remain within the boundaries of epistemology proper and to determine whether and to what extent the intellectual virtues being mapped by VR theorists are genuinely epistemic virtues. We have seen that there are two ways of doing this, both motivated by a popular proposal in the epistemic norms literature. According to the defenders of this view, epistemic norms and – by extension – virtues are concerned with distinctively epistemic features; that is to say, epistemic norms and virtues, as such, are norms and virtues with

Epistemic Virtues and Virtues with Epistemic Content



epistemic content. We have seen, however, that content individuation for epistemic norms and virtues runs into trouble on at least two counts. First, it doesn’t generalize to other normative domains in the way it ought to. And second, it makes counterintuitive predictions in the epistemic domain. In what follows, we will look at a different proposal for individuating genuinely epistemic normative notions. We argue that it is not only theoretically superior to CI but also intuitively adequate. However, with this improved method to hand, many of the virtues discussed in the VR literature still fail to qualify as genuinely epistemic virtues. Rather, they are moral virtues with epistemic content.

. Value Individuation This section’s intention is to propose a theory-neutral individuation recipe for epistemic virtues, based on a widely accepted claim concerning the relation between the axiological and the deontic. In previous work (Simion , ), one of us has argued that the theory of normativity has a widely accepted answer to our question ready to hand – norms can be typed by the type of good they are associated with: Value Individuation (VI) A norm N is of type T if and only if N is associated with goods of type T.

According to VI, prudential norms are associated with prudential goods, moral norms are associated with moral goods, and so on. All normative domains have goods (values) that are central to them, in virtue of the kind of normative domains they are: survival is a prudential good; promisekeeping is a moral good; politeness is a social good; beauty is an aesthetic good; money is a financial good. Similarly, if etymology is any guide to what normative domain ‘the epistemic’ is supposed to refer to, knowledge is an epistemic good. VI is of course still rather vague, and in particular the association relation at issue requires spelling out. After all, one way in which a norm can be associated with a particular good is by requiring more or less of that good. This, however, will put us back in the same trouble we faced with content  

We will say more about what ‘associated with’ means below. For an argument from VI to there being no epistemic norm for action, but only an epistemic norm for practical reasoning – in virtue of the latter, but not the former, being plausibly conducive to epistemic goods – see Simion . For a defense of the sufficiency direction of the knowledge norm of assertion via VI, see Simion . For an argument to the effect that VI renders the knowledge norm of assertion perfectly compatible with classical invariantism, see Simion .



, ,   

individuation: just because a norm requires me to know how to swim before jumping into lakes, and is thereby associated with an epistemic good – that is, knowledge – it does not follow that it is an epistemic norm. We propose that the association relation stands for one or another direction of explanation: either the goods explain the norm, or the other way around. To see how this goes, it may be worth noting that VI is valuetheoretically neutral in the sense that it doesn’t come with any substantive commitments about the relation between the axiological and the deontic. This is because the association claim between norms and goals of the same type does not imply any particular direction of explanation. As a result, it is compatible with both of the two leading views about the relationship between the axiological and the deontic. Teleologists (e.g. Moore [], Sidgwick [], and Slote []) explain the ‘ought’ in terms of the ‘good’; they claim that the norm of type X is there to guide us in reaching the good of type X. In contrast, deontologists (e.g. Scanlon [], Ewing [], Rabinowicz and Rönnow-Rasmussen []) reverse the order of explanation: according to “fitting attitude” accounts of value, for example, the goods of type X are only valuable to begin with because the norm of type X gives us reasons to favour them. Crucially, in either case, the mere association claim at issue in VI holds. Since we are interested specifically in the epistemic domain, we want to take a quick look at how the association claim may be unpacked for distinctively epistemic norms within both a teleological and a deontological framework. Here is Peter Graham for an explicit statement of the teleological direction of explanation: Epistemic norms in this sense govern what we ought to say, do or think from an epistemic point of view, from the point of view of promoting true belief and avoiding error. (Graham : )

Here is Kurt Sylvan for a statement of the deontological direction of VI: [C]entral epistemic properties like justification, coherence, and substantive rationality derive non-instrumental epistemic value in virtue of the fact that they manifest different epistemically fitting ways of valuing accuracy. (Sylvan , section ., emphasis in original)

Note also that VI makes plausible predictions about epistemic norms and norms with epistemic content. For instance, BELIEVE is correctly characterized as an epistemic norm. After all, it is a norm associated with the 

For a good general overview of the relevant literature in value theory, see, for instance, Schroeder .

Epistemic Virtues and Virtues with Epistemic Content



epistemic good of true belief. The easiest way to see this is by looking at the teleological direction of the association claim. Believing what one has good evidence for believing is a good means to true belief; it promotes true belief. Consider, by way of contrast, SING. Rather than being associated with distinctively epistemic goods (e.g. promoting true belief ), it is associated with an aesthetic good. Again, in a teleological framework, singing only songs one knows is a way of promoting beauty. As a result, VI classifies it, again correctly, as an aesthetic norm, albeit one with epistemic content. Now, just as we took CI to motivate the parallel content-based recipe for individuating epistemic virtues, we may take VI to motivate the obvious value-based recipe for so doing. Here goes: Virtues Value Individuation (VVI) A virtue V is a distinctively epistemic virtue if and only if V is associated with epistemic goods.

One might initially worry whether VVI preserves VI’s value-theoretic neutrality. After all, given that virtues are associated with epistemic goods, one may wonder whether we will be committed to a distinctively teleological value theory. Fortunately, the answer to this question is no. After all, it may be that, in accordance with VI, the axiological is analyzed in terms of the deontological. If so, it may still be that norms explain both goods and virtues. By the same token, VVI does preserve the value theoretic neutrality after all. Interestingly enough, there is reason to believe that virtue responsibilists are sympathetic to VVI. In several places, leading VR figures seem to hint at something along VVI lines. Let’s start with what Roberts and Wood () say about this issue: The difference between our study and a study in virtue ethics is simply that we are interested in the relations between the virtues and the intellectual goods. (Roberts and Wood : )

Here is Montmarquet on epistemic virtues: What I want to suggest, then, as a first approximation, is that the epistemic virtues are those personal qualities (or qualities of character) that are conducive to the discovery of truth and the avoidance of error. (Montmarquet : )  

We’ll leave it to the reader to figure out how the association claim works within the deontological framework and for other norms (e.g. JUMP and ASSERT). Again, we can also individuate epistemic virtues by VI and NCV. Since it’s easy enough to see that the results will be the same, we will not go into details about this here.



, ,   

Moreover, Jason Baehr’s view about what distinguishes epistemic virtues from other types of virtues seems very similar to VVI: While structurally similar to moral virtues, they [epistemic virtues] are also distinct from what we ordinarily think of as moral virtues on account of aiming at distinctively epistemic goods like truth, knowledge, and understanding. (Baehr : )

Last but not least, here is Linda Zagzebski: I will argue that truth conduciveness is an essential component of intellectual virtues and I will attempt to ground these virtues in the motivation for knowledge. (Zagzebski : )

In the next section, with our preferred individuation method in hand, we examine some consequences for VR. We argue that, if VVI is correct, many of the purportedly epistemic virtues discussed in the literature turn out (once again) to be moral virtues with epistemic content rather than genuinely epistemic virtues.

. Epistemic Virtues and Virtues with Epistemic Content We submit that many of the VR virtues fail to count as epistemic virtues by the lights of VVI. Let’s start with an easy case. Consider, again, the virtue of epistemic temperance discussed by Roberts and Wood. Intuitively, Roberts and Wood are right: the slave-owner who looked away when the slave women were nursing their infants was, to this extent, manifesting virtue. As Roberts and Wood well put it: “To be an indiscriminate ogler is a trait of bad intellectual character, a failure of discipline of the will to know” (: ). However, since we have seen that typing norms and virtues by content will not do, the question is: when Roberts and Wood talk about “bad intellectual character,” what type of badness is that? Is it genuinely epistemic (i.e. associated in the relevant way with epistemic goods) or rather badness of a different sort (i.e. associated with different types of goods), albeit badness that has epistemic content? It is hard to deny that it must be the latter. After all, what seems to be going on here is that a properly epistemic virtue, “the love of knowledge” (which is, in the relevant way – either teleologically or deontologically – associated with an epistemic good, in this case knowledge), is being overridden (“disciplined”) by another virtue, namely the virtue of epistemic temperance. Plausibly, though, the 

To be sure, the claim here is that the slave-owner manifests virtue in this specific regard, and not, of course, that the slave-owner in any way manifests virtue by being a slave-owner.

Epistemic Virtues and Virtues with Epistemic Content



latter is not really associated with epistemic goods in the relevant sense. After all, on neither direction of explanation does it plausibly stand in an association relation with epistemic goods. It is neither conducive to their acquisition, nor does it give one reasons to favor them. Rather, what seems to be the case is that epistemic temperance stands in the relevant relation with moral goods, such as respect for privacy or discretion. If all this is so, then Roberts and Wood are guilty of the following mistake: it is not the case, as they would have us believe, that the slaveowner is an epistemically virtuous person in virtue of tempering his will to know in the relevant case; rather, the slave-owner manifests a moral virtue with epistemic content by tempering his will to know. On a similar note, consider Heather Battaly’s discussion of epistemic temperance as well as the corresponding vice which she calls “epistemic self-indulgence”: [T]he passions and actions associated with these traits are epistemic rather than physical, and include wanting, consuming, and enjoying beliefs, knowledge, and belief-forming practices. I argue that the epistemically temperate person desires, consumes, and enjoys only appropriate epistemic objects, only at appropriate times, and only in appropriate amounts. The epistemically self-indulgent person, however . . . desires, consumes, and enjoys epistemic objects at inappropriate times (e.g., while having sex with his partner); or desires, consumes, and enjoys epistemic objects too much (thus preventing him from pursuing other things of value). (Battaly : )

Note that the goods secured by “epistemic temperance” are again social or prudential, precisely at the expense of epistemic goods. According to Battaly herself, the epistemically temperate person sacrifices the consumption of epistemic goods for the sake of “pursuing other things of value.” Conversely, it is not clear why “epistemic self-indulgence” should be considered an epistemic vice. After all, on both directions of explanation it is strongly associated with epistemic goods rather than bads. Thus, by the lights of VVI, it seems that Battaly’s discussion also manifests the familiar mistake. Importantly, this is not to say that a case cannot be made for the claim that epistemic temperance is an epistemic virtue proper or that epistemic self-indulgence is an epistemic vice. Here is, very roughly, how an argument to this effect might go. Consider someone who spends all of their epistemic resources on acquisitioning disparate items of knowledge, about completely unrelated topics. Perhaps this person should temper their will for a high quantity of epistemic goods, in favor of the quality of the epistemic goods. That is, perhaps a project of diving more deeply into a



, ,   

particular domain, in order to achieve understanding thereof, would be more epistemically worthwhile than one of simply acquisitioning disparate items of knowledge. At the very least, we think that this question is well worth investigating. Nenad Miscevic’s () discussion of curiosity (which he takes to be the basic epistemic virtue) raises similar concerns. Miscevic carefully draws our attention to the fact that what he means by the “virtue of curiosity” is something that excludes nosiness. Indeed, he stipulates the term “curiosity+” to pick out a kind of curiosity that is not too strong. Curiosity+ is just strong enough to secure some epistemic goods, without thereby giving rise to bad moral side effects. Here is Miscevic: Curiosity, when a virtue, call it curiosity+, includes knowledge of appropriateness, and motivation for appropriate exercise. Curiosity-, the vicious inquisitiveness, is not really curiosity . . . We would then in general have two sub-species of cognitive intrinsic desire to know, intrinsic curiosity+, and curiosity-, the bad intrinsic curiosity. The first is truly a virtue, the second is not . . . Typical . . . cases of low-level object curiosity [are] aiming at private and intimate matters of others (nosiness), or [are] connected to morally problematic goals or consequences. (Miscevic : –)

As the previous discussion of epistemic temperance makes clear, by the lights of VVI Miscevic also makes the familiar mistake here: conduciveness to moral goods does not bear on whether or to what extent a particular character trait is an epistemic virtue. Indeed, it is unclear why “curiosity-,” or nosiness, is not legitimately understood as an epistemic virtue, precisely insofar as it is associated with epistemic goods, and despite the fact that it plausibly fails to qualify as a moral virtue. Consider a teleological direction of explanation: nosiness, as morally indecent as it might be, will definitely be one good character trait to have for securing epistemic goods, such as knowledge or true beliefs. It is, then, unclear why we should endorse Miscevic’s distinction between good curiosity (curiosity+) and bad curiosity (curiosity-), rather than allowing that curiosity is simpliciter an epistemic virtue – albeit one that may or may not also count as a moral virtue, depending on whether it brings about bad moral consequences. One question that champions of VR might rightly ask at this point is the following: Is it not the case that, on the Aristotelian model, only proper exercise of a ‘virtue’ counts as an instance of genuine virtue, whereas improper exercise does not really count as an instance of virtue? Here is Miscevic on this issue, referencing a discussion by Philippa Foot:

Epistemic Virtues and Virtues with Epistemic Content



The courage of a whistleblower is courage, the bravery of an SS-officer is not . . . Similarly, nosiness is not really curiosity, at best it is pseudocuriosity . . . Curiosity, when a virtue . . . includes knowledge of appropriateness, and motivation for appropriate exercise. Curiosity-, the vicious inquisitiveness, is not really curiosity. (Miscevic : )

Regardless of whether this is correct, we can simply grant the point. Miscevic’s case remains problematic anyway, for the reasons outlined. Here is why, very briefly: ‘proper exercise’ is a normative notion. As such, it requires typing. There is such a thing as the epistemically proper exercise of a virtue, the morally proper exercise of a virtue, and so on. If VVI is right, epistemically proper exercise of curiosity will represent an instance of the epistemic virtue of curiosity. Once again, it is not clear why episodes of nosiness will not count as epistemically proper exercises of curiosity – after all, they are apt to secure epistemic goods. This Aristotelian rejoinder, then, does not help remove the suspicion of the familiar mistake in these cases. It comes to light that instances of the familiar mistake resulting from a failure to employ a clear individuation recipe for virtues are easy to encounter in the VR literature. By the lights of VVI, even virtues that have been at the very heart of the VR literature – such as open-mindedness, or intellectual autonomy – deserve closer scrutiny. We’d like to close with a brief examination of open-mindedness. Of course, at least at first glance, open-mindedness comes across as a paradigmatic epistemic virtue: being open-minded opens one up to properly appreciating the views of others. Note, however, that whether a given case really qualifies as proper appreciation may depend on, () the direction of explanation for unpacking VVI; and () contextual features. To see this, consider open-mindedness in a teleological framework according to which virtues count as genuinely epistemic only insofar as they are conducive to epistemic goods. Now, plausibly, for most people of average epistemic endowment, being openminded is indeed conducive to epistemic goods. After all, when undertaking intellectual projects most of us are likely to encounter better (as in epistemically better) ideas/views, and so forth, than our own. Being receptive to these other ideas/views will thus be conducive to epistemic improvement. In contrast, though, it is not clear that the same is the case when we move further up the scale of epistemic endowment. For a being that is well above average in cognitive ability, open-mindedness may even conducive to epistemic bads. After all, in such a case, it may be conducive to abandoning perfectly fine beliefs in the light of misleading evidence. An open-minded mathematical genius might wrongly abandon worthwhile beliefs in the light of less qualified testimony.



, ,   

Perhaps, then, even open-mindedness – one of the central intellectual virtues in the VR literature – is more accurately characterized as a moral virtue with epistemic content. After all, being an open-minded person quite plausibly means taking the view of other human beings seriously; this, arguably, is but an instance of acting in accordance with a more general moral law requiring us to respect humanity. Perhaps what does count as an epistemic virtue proper is what we might call intellectual justice: for instance, the intellectually just person gives the deserved weight to other people’s testimony (e.g. Fricker ). Thus, intellectual justice is plausibly an instance of acting in accordance with a more general epistemic norm requiring one to proportion one’s beliefs to the available evidence.

. Conclusion Many of the VR virtues either fail to count as epistemic virtues by the lights of VVI, or only do so very contingently, given particular contextual parameters. Of course, much more could be said about these issues. Perhaps there is an argument to be made that open-mindedness is a genuine epistemic virtue after all, even by the lights of VVI. Perhaps Miscevic is right and nosiness is not a genuine epistemic virtue. This chapter does not exclude any of this. Rather, our ambition here has been to draw attention to the danger of making a familiar mistake in the absence of a proper individuation recipe for epistemic virtues. We need to proceed carefully in our project of identifying and analyzing epistemic virtues. R E F E R EN C E S Alston, W. . Beyond Justification: Dimensions of Epistemic Evaluation. Ithaca, NY: Cornell University Press. Baehr, J. . The Inquiring Mind: On Intellectual Virtues and Virtue Epistemology. Oxford: Oxford University Press. . “Four Dimensions of an Intellectual Virtue,” in Mi, C., Slote, M. and Sosa, E. (eds), Moral and Intellectual Virtues in Western and Chinese Philosophy. London: Routledge. Battaly, H. . “Varieties of Epistemic Vice,” in Matheson, J. and Vitz, R. (eds), The Ethics of Belief. Oxford: Oxford University Press. Benton, M. . “Knowledge Norms,” in Fieser, J. and Dowden, B. (eds), Internet Encyclopedia of Philosophy. www.iep.utm.edu/kn-norms/, Accessed: June , . Ewing, A. C. . The Definition of Good. London: Macmillan. Fantl, J. and McGrath, M. . “Evidence, Pragmatics, and Justification,” Philosophical Review : –.

Epistemic Virtues and Virtues with Epistemic Content



Fricker, M. . Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press. Gerken, M. . “Warrant and Action,” Synthese : –. Graham, P. . “Epistemic Normativity and Social Norms,” in Henderson, D. and Greco, J. (eds), Epistemic Evaluation: Purposeful Epistemology. Oxford: Oxford University Press. Hawthorne, J. . Knowledge and Lotteries. Oxford: Oxford University Press. Kelp, C. . “Assertion: A Function First Account,” Nous : –. Lackey, J. . “Assertion and Isolated Second Hand Knowledge,” in Brown, J. and Cappelen, H. (eds), Assertion: New Philosophical Essays. Oxford: Oxford University Press. . “Deficient Testimonial Knowledge,” in Henning, T. and Schweikard, D. (eds), Knowledge, Virtue and Action. London: Routledge. Littlejohn, C. and Turri, J. (eds). . Epistemic Norms: New Essays on Action, Belief, and Assertion. Oxford: Oxford University Press. Maitra, I. . “Assertion, Norms, and Games,” in Brown, J. and Cappelen, H. (eds), Assertion: New Philosophical Essays. Oxford: Oxford University Press. Miscevic, N. . “Epistemic Value. Curiosity, Knowledge and ResponseDependence,” Croatian Journal of Philosophy : –. . “Curiosity – the Basic Epistemic Virtue,” in Mi, C., Slote, M., and Sosa, E. (eds), Moral and Intellectual Virtues in Western and Chinese Philosophy. London: Routledge. Montmarquet, J. . Epistemic Virtue and Doxastic Responsibility. Lanham, MD: Rowman and Littlefield. Moore, G. E. . Principia Ethica, revised edition. Cambridge, UK: Cambridge University Press. Roberts, R. and Wood, W. . Intellectual Virtues: An Essay in Regulative Epistemology. Oxford: Clarendon Press. Scanlon, T. . What We Owe to Each Other. Cambridge, MA: Harvard University Press. Schroeder, M. . “Value Theory,” in Zalta, E. (ed.), The Stanford Encyclopedia of Philosophy, http://plato.stanford.edu/archives/sum/entries/value-the ory/, Accessed: June , . Sidgwick, H. . The Methods of Ethics, th edition. Indianapolis, IN: Hackett. Simion, M. . “Assertion: Knowledge is Enough,” Synthese : –. . “No Epistemic Norm for Action,” American Philosophical Quarterly : –. . “Assertion: The Context Sensitivity Dilemma,” Mind & Language. Slote, M. . Beyond Optimizing. Cambridge, MA: Harvard University Press. Sylvan, K. . www.kurtlsylvan.com/research.html, Accessed: June , . Zagzebski, L. . Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge, UK: Cambridge University Press.

 

Difficulty and Knowledge Fernando Broncano-Berrocal

. Difficulty in the Theory of Knowledge Some things are more difficult to know than others. It is more difficult to come to know how many grains of sand there are in a whole dune than in a heap. Proving the Poincaré conjecture is certainly more difficult than coming to know what  +  equals. The larger the distance away, the more difficult it is to recognize a facial expression. The question that this chapter will address is: How can the relationship between difficulty and knowledge be accounted for in epistemology? As obvious as it seems, explaining that knowledge can be difficult in familiar epistemological terms is less straightforward than one might initially think. Suppose with evidentialism that knowledge is a function of one’s evidence. What is the difference in difficulty between coming to know how many sand grains are there in a dune and in a heap? One idea (in line with evidentialism) is that the more evidence needed to know something, the more difficult it is to know it. In order to know how many sand grains there are in a dune, one certainly needs to gather more evidence, which explains the difference in difficulty. But if the amount of evidence is the key factor that accounts for the relationship between difficulty and knowledge in general, what explains the fact that solving a complex Sudoku puzzle is more difficult than solving a simpler one? Here the difference does not seem to lie in the quantity or quality of one’s evidence – the evidence is roughly the same in both cases – but in the skill required to find a solution. Suppose now with reliabilism that knowledge is a function of the reliability of one’s belief-forming processes. One idea (in line with reliabilism) is that the less reliable an agent is in some domain of inquiry, the more difficult it is for them to know propositions in that domain. Accordingly, the difference in difficulty between solving a complex Sudoku puzzle and a simpler one can be explained in terms of the fact that one needs to be 

Difficulty and Knowledge



more reliable in solving Sudoku puzzles to solve the former. However, while appealing to the notion of reliability may help account for cases in which the difference in difficulty is down to a difference in skill, it can hardly explain cases in which it is down to a difference in effort. For example, imagine a very rudimentary calculator with only three buttons ‘,’ ‘+’ and ‘=,’ which is designed to help kids learn natural numbers. There is no other way to add numbers but to press ‘’ and ‘+’ as many times as needed and then press ‘=’ to get a result. In addition, suppose that the calculator, being designed for kids, automatically corrects syntactic errors that may be produced in the course of typing, e.g., long strings of symbols. In such a device, calculating a large natural number is arguably more difficult than calculating a small one. Yet, this difference in difficulty neither lies in the reliability of the belief-forming method, nor in the skill needed. First, the calculator is as reliable when the result is a big number (e.g. ,,) as when the output is a small one (e.g. ). Second, since the calculator automatically corrects eventual syntactic errors, the only skill needed is that one is able to type ‘’ and ‘+’ as many times as one wishes and that, at some point, one presses ‘=’ to find the result. In this way, calculating ,, is more difficult than  simply because it takes more time and effort to introduce the relevant values. The difference in difficulty is not a difference in reliability or skill. These considerations do not show that the relationship between difficulty and knowledge cannot be accounted for in evidentialist or simple reliabilist terms but serve to illustrate that explaining this relationship in familiar epistemological terms (such as evidence or reliability) is less straightforward than one might initially think. Enter virtue epistemology. Virtue epistemology provides a promising theoretical framework for explaining and elaborating on the idea that knowledge can be difficult. The main reason is that it gives a central role to the notion of skill in the explanation of the nature of knowledge and, as we have seen, this notion seems to play a key part in the understanding of the concept of difficulty. The aim of this chapter is to explore these conceptual connections and account for the relationship between difficulty and knowledge in virtue epistemological terms. In so doing, I will make two relevant points about virtue epistemology. The first point will be negative: current views make certain assumptions concerning the role that situations play in the acquisition of knowledge that not only prevent them from explaining the relationship between difficulty and knowledge but also lead them to false claims about the very nature of knowledge. The second point will be positive: the way to amend this problem is to adopt the idea



 -

that knowing can be difficult in the way I will suggest. In this manner, the chapter aims to give insights both into the relationship between difficulty and knowledge and into the (virtue epistemological) nature of knowledge. The chapter is structured as follows. In Section ., I will introduce the two main branches of virtue epistemology and the main assumptions that each makes concerning the role that situations play in the acquisition of knowledge. The main point will be that character-based virtue epistemology, also known as virtue responsibilism, does not provide a suitable theoretical framework for explaining the idea that knowledge can be difficult; by contrast, competence-based virtue epistemology, also known as virtue reliabilism, is prima facie compatible with it. However, as I will argue in Section ., a key assumption of virtue reliabilism – namely, the assumption that cognitive abilities are reliable dispositions to form true beliefs in appropriate conditions – not only precludes the theory from accounting for the relationship between difficulty and knowledge, but also renders its main tenet – the thesis that knowledge requires the manifestation of cognitive ability in appropriate conditions – false. In Section ., I will connect the virtue reliabilist framework with recent work on notions of achievement and difficulty. The aim will be to shed some light on the idea that knowledge can be difficult as well as on the virtue reliabilist thesis that knowledge is a cognitive achievement. Section . will follow up on the ideas introduced in Section . and will develop the idea that acquiring knowledge is a special kind of challenge, and one with varying degrees of difficulty.

. The Situational Assumptions of Virtue Epistemology Virtue epistemology offers a simple answer to the question of what the nature of knowledge is: knowing that p is a matter of coming to believe p truly because of an exercise of intellectual virtue. Yet, as simple as it looks, there is some controversy about how to unpack this answer and especially about how to understand the relevant intellectual virtues. Some virtue epistemologists think that the relevant intellectual virtues are stable reasons-responsive character traits that motivate or orient individuals toward the attainment of epistemic goods, including knowledge. Open-mindedness, curiosity, creativity, intellectual courage and intellectual perseverance are some examples. Crucially, just as character traits in general are supposed to produce cross-situationally consistent behavior independent of normatively irrelevant situational factors, the cognitive performances of agents 

See Baehr  for extensive discussion on this approach to intellectual virtues.

Difficulty and Knowledge



possessing character-based intellectual virtues are supposed to be consistent in a great variety of situations and independently of such factors. For example, the courageous person is supposed to exhibit courage in a wide range of dangerous or challenging situations, and more importantly, independently of, e.g., the presence of bystanders. In the same vein, to exhibit intellectual courage, one must consistently challenge widespread opinions regardless of bystanders. Likewise, to count as intellectually perseverant, one must carry on one’s inquiry or pursue one’s intellectual project whenever one’s beliefs are contested, independently of whether one is tired or sad or one’s favorite show is live on TV. To be clear, the claim is not that character-based intellectual virtues are situation-independent: one is not expected to be intellectually courageous or perseverant while sleeping or dead. Instead, the claim is that they are relativized to a wide variety of environments, in that intellectually virtuous agents should manifest them in all of them. In addition, if they are supposed to be independent of anything, it is of normatively irrelevant situational influences (e.g. the presence of bystanders, energy or mood elevators and so on). These two features (cross-situational consistency across a wide range of environments and independence of normatively irrelevant factors) constitute what we may call the situational assumption of character-based virtue epistemology (also known as virtue responsibilism). Other virtue epistemologists theorize that the relevant intellectual virtues are based on cognitive faculties such as perception, reasoning or memory. More specifically, they conceive them as stable cognitive dispositions – as opposed to motivational or conative character traits – to form true beliefs reliably. Understood thus, intellectual virtues are particular kinds of competences, skills or abilities (e.g. visual, inferential, memorial) whose goal is to form true rather than false beliefs – competences, skills or abilities in general are defined as stable dispositions to achieve goals reliably. Importantly, an assumption that epistemologists upholding this conception of intellectual virtues make is that abilities, in general, and cognitive abilities, in particular, are reliable dispositions to achieve goals in appropriate conditions. What are the appropriate conditions of an ability, on the reliabilist model? One idea is that the appropriate conditions of an agent’s ability   

See Alfano  for relevant discussion on the cross-situational component of character traits and corresponding virtues. See Greco  and Sosa  for helpful discussion on this way of understanding intellectual virtues. Throughout the chapter, I will use the terms ‘competence,’ ‘skill’ and ‘ability’ interchangeably. Thanks to John Greco for urging me to clarify this and for helpful suggestions.



 -

to achieve a certain kind of goal are the type of conditions under which the agent, by exercising their ability, can achieve goals of that kind with a degree of reliability (e.g. a ratio of true-to-false beliefs or a ratio of successful-to-failed shots) that is enough to meet the relevant evaluative standard (e.g. knowledge-level reliability; gold-medal-level reliability). That cognitive abilities are relativized to appropriate conditions thus understood is the situational assumption of competence-based virtue epistemology (also known as virtue reliabilism). The previous is a minimal characterization of what appropriate conditions are, but at least one that most authors in the virtue reliabilism literature should accept. There are, of course, further options. One may opt for understanding appropriate conditions as normal conditions and then give an account of what kind of conditions count as ‘normal,’ e.g., in terms of statistical normality (conditions that are typical), etiological normality (conditions where the ability was learned or selected), teleofunctional normality (conditions that enable an ability to comply with its causal-functional role, if any) or even in terms of practical relevance (conditions suitable for the practical task at hand). One can also bypass the identification of appropriate conditions with normal conditions and yet use some of the previous considerations to explain what appropriate conditions are. One can also perhaps sidestep giving a general account of appropriate conditions for all abilities and use different senses of appropriateness for different types of abilities. Thus, for innate abilities, appropriate conditions might be those to which we are adapted as a species. For example, appropriate conditions for human visual abilities plausibly include a set of light conditions ranging from full sunlight to dim light but excluding pitch dark – a species with infrared night vision would have different appropriate conditions. For learned abilities, appropriate conditions might be those where the ability was learned in the first place. For example, appropriate conditions for the ability to hit baseballs successfully include a set of wind conditions ranging from no wind to relatively windy conditions but excluding hurricanes – no one learns how to hit baseballs in hurricane conditions. For extended abilities, such as the ability to navigate the environment using a smartphone app like Google Maps, appropriate conditions might include the kind of conditions under which the relevant tool can comply with the function it is designed for (e.g. being charged).  

See Greco () for relevant discussion of these options. For useful discussion of how active externalism (including the extended cognition hypothesis) relates to epistemology, see Carter et al. .

Difficulty and Knowledge



All these theoretical alternatives state different ways to specify conditions under which one would exercise a given ability with enough reliability for the relevant goal and evaluative standard and, in this way, they are compatible with (in fact, specifications of ) the minimal characterization of appropriate conditions offered above. What conditions are inappropriate for an ability, then? Given this minimal characterization, it seems that inappropriate conditions are whatever conditions in which an agent’s otherwise reliable ability would not be reliable enough. In this way, on the reliabilist model, factors such as the presence of bystanders, one’s mood or one’s energy levels may render an environment inappropriate for exercising a cognitive ability, so long as they may reduce the reliability of one’s abilities in that environment. However, note that whatever account one gives of the kind of conditions that prevent an agent from exercising their abilities reliably enough (inappropriate conditions), one needs to supplement it with an explanation of why some conditions count as inappropriate despite the fact that one would exercise them with enough reliability. For example, a skillful archer shooting metal arrows at a giant magnetic target seems to be in inappropriate conditions for manifesting their archery skills despite getting close to infallibility. The reason is that the conditions are unduly helpful, in that a terrible archer would hit the target almost infallibly. In a sense, such conditions actually mask any archery skills. ..

Virtue Responsibilism: A Problematic Framework

There is a tension between the situational assumption of virtue responsibilism and the idea that knowledge may involve different degrees of difficulty. On the one hand, it is intuitively more difficult to manifest epistemically appropriate behavior in some circumstances than in others. Consider the virtue of intellectual courage. It was certainly more difficult for Galileo Galilei to challenge geocentrism in his time and thus exhibit intellectually courageous behavior than for Albert Einstein to challenge Newtonian mechanics – Einstein’s life was not at stake. Yet, the situational assumption of virtue responsibilism demands that intellectually virtuous   

See Alfano  and Turri  for an elucidation of why this version of virtue epistemology is not troubled by such situational influences. See also Sosa’s guardian angel case later in this chapter. As we will see, this special set of inappropriate conditions bears on the relationship between difficulty and knowledge.



 -

agents consistently manifest appropriate epistemic behavior across a wide range of circumstances. One way character-based virtue epistemologists may bring the two ideas together (the idea that situations feature varying degrees of difficulty and the idea that intellectual traits must exhibit cross-situationally consistent epistemic behavior) is to argue, first, that character-based intellectual virtues come in degrees; second, that the degree to which one is intellectually virtuous depends on the range of conditions across which one would manifest the relevant responsibilist trait. Thus, circumstances outside that range are such that it is more difficult to behave virtuously in them. To see this, suppose that Galileo and Einstein were equally intellectually courageous, in that they would have manifested intellectually courageous behavior across the same range of conditions, e.g., Einstein would have publicly defended the theory of relativity in Galileo’s times, and the other way around. Now consider a contemporary scientist, Cowardeo, who is much less courageous than they were. In particular, in Einstein’s circumstances, Cowardeo would manifest intellectual courage, but in Galileo’s, he would not. What explains why Galileo's circumstances are more demanding than Einstein’s for Cowardeo is that the former are not part of the range of action of his intellectual virtue. Some virtue responsibilists might not agree with this externalist characterization of character-based intellectual virtues, but it at least helps explain how learning intellectually virtuous behavior is possible: one learns how to be intellectually virtuous by including more and more types of circumstances in the set of circumstances over which one’s intellectual traits are defined (i.e. by expanding its range of action) until one gets close to what an ideal intellectually virtuous agent would do. For instance, in order to learn how to become (or be more) intellectually courageous one must expose oneself to a variety of situations in which holding one’s ideas publicly puts one at risk of incurring different types of personal costs (this is what intellectual courage consists in). Accordingly, exhibiting intellectually courageous behavior may be more or less difficult when one is at risk of (in decreasing degree of difficulty) being murdered, incarcerated, stripped of citizenship or deprived of one’s dinner. However (and here comes the objection to virtue responsibilism), not all conditions serve for expanding the range of action of one’s intellectual traits. The former are certainly the sort of circumstances in which one is supposed to manifest intellectually courageous behavior if one is to be 

Thanks to Chris Kelp for the suggestion.

Difficulty and Knowledge



correctly attributed this virtue, i.e., the type of circumstances that serve to define what intellectual courage is. By contrast, the presence of bystanders, being in a good mood or rested are normatively irrelevant situational influences, in that intellectually courageous people are supposed to consistently manifest virtuous behavior independently of whether or not such factors are in place. The problem is that such apparently irrelevant situational influences do have a bearing on whether we end up manifesting intellectually virtuous behavior, as the situationist critique to characterbased virtue epistemology has plausibly shown. Consider the intellectual virtue of creativity. Some empirical studies show that cognitive tasks that require creativity are better accomplished when subjects are in a good mood, but being creative-while-in-a-good-mood is for sure not an intellectual virtue. Now, if our epistemic behavior is subject to normatively irrelevant situational influences such as mood elevators, mood depressors, the presence of bystanders or even sounds and smells, it is easy to see that these factors also have a bearing on how difficult it is for us to manifest virtuous epistemic behavior. For example, being intellectually courageous while happy is arguably easier than when one is depressed. However, it follows from the notion of intellectual courage that, if one is in its possession, one ought to manifest it independently of such factors. In other words: such factors should pose no difficulty when it comes to manifesting intellectually courageous behavior, but the problem is that they do. Consequently, it is unclear how the relationship between difficulty and knowledge can be accounted for in virtue responsibilist terms. To put it differently, one could argue that to learn how to become (or be more) intellectually courageous one must not only expose oneself to different situations in which holding one’s ideas publicly puts one at risk of incurring different types of personal costs (as pointed out earlier), but also to situations in which the aforementioned normatively irrelevant factors are in place. However, accepting this idea entails casting doubt on the very nature of intellectual courage. While behaving intellectually-courageously-while-atrisk-of-incarceration is the kind of thing intellectually courageous people are supposed to do, behaving intellectually-courageously-while-in-a-good-mood is not. On a character-based model, then, being in a good or a bad mood cannot be part of the explanation of why it can be more or less difficult to manifest (or to learn how to manifest) intellectually virtuous behavior and hence to acquire knowledge. 

For a very helpful monograph on moral and epistemic situationism, see Alfano .



 - .. Virtue Reliabilism: A Promising Framework

Virtue reliabilists are happy to accept that intellectual virtues and, in particular, cognitive abilities are situation dependent, even on normatively irrelevant factors (irrelevant for the manifestation of responsibilist traits). After all, unlike intellectually virtuous character traits, which are supposed to consistently manifest appropriate epistemic behavior across the board and, more importantly, independently of normatively irrelevant situational influences, cognitive abilities are expected to operate reliably under a much more restricted set of circumstances. These include factors that are normatively irrelevant for manifesting responsibilist traits but that are normatively relevant for manifesting cognitive abilities, insofar as they may have a bearing on how reliably they are exercised. This is best illustrated by empirical evidence, which shows that cognition is very sensitive to situational influences (e.g. if a statement is printed in an easier-to-read color, we are more likely to believe it; if we are tired, we are more likely to overestimate the distance of an object; if we hear spoken arguments that are uttered quickly, we are more likely to believe their conclusions). Although some might worry that this speaks against the existence of cognitive abilities, the upshot, as several virtue reliabilists have argued, is that the range of action of our cognitive abilities is not broad but narrow. Accordingly, virtue reliabilists can in principle argue that knowing a proposition is difficult to the extent that one attempts to form a belief in it in inappropriate circumstances, i.e., when such epistemically disrupting situational influences are in play, and easy (or easier) otherwise. If we add to this possible explanation of the relationship between difficulty and knowledge the fact that virtue reliabilism defines knowledge in terms of the same concept that serves to define difficulty (namely, the concept of skill), there does not seem to be any prima facie tension between the competence-based framework and the idea that knowing can be difficult. However, as we will see in Section ., virtue reliabilism can neither account for the relationship between difficult and knowledge, nor 

  

Epistemologists defending virtue reliabilist accounts or virtue reliabilist conditions on knowledge include Broncano-Berrocal , ; Carter ; Greco , ; Kelp ; Jarvis ; Pritchard ; Turri  and most prominently Sosa , . Recall: the kind of circumstances that sustain enough reliability (or truth conduciveness) are what virtue reliabilists call appropriate conditions. See Olin & Doris  for an overview of the relevant empirical evidence.  Olin & Doris  share this skepticism. See Sosa  and Turri  for this reply.

Difficulty and Knowledge



for the nature of knowledge, unless it drops its key situational assumption first.

. Virtue Reliabilism and the Appropriateness Requirement Virtue reliabilism gives the following simple answer to the question of what propositional knowledge is: knowing that p is a matter of coming to believe p truly because of an exercise of cognitive ability. This simple definition of knowledge is backed up by a more complex and explanatorily powerful normative framework for evaluating beliefs, which is in turn based on the assumption that to hold a belief – or alternatively, to form a belief – is a sort of performance, namely a cognitive performance. The guiding idea is that, since beliefs are kinds of performances, they can be evaluated along the same normative dimensions of assessment that serve to evaluate performances in general. In this sense, virtue reliabilists think that (at least) four dimensions of assessment are relevant to accounting for knowledge: ) success, ) competence (or ability or skill), ) success because of competence, and ) achievement. Success. Performances can be successful or unsuccessful. A performance (e.g. making a recipe) is successful if and only if its aim is accomplished (e.g. when the resulting dish is tasty). Analogously, beliefs (i.e. cognitive performances) are (epistemically) successful if and only if their aim is accomplished (i.e. when they are correct). Competence. Performances can be also deemed competent or incompetent. A performance is competent if and only if a competent or skillful agent (i.e. skillful in the performance’s domain) produces it (e.g. a dish can be considered competently made when it has been made by a competent cook). Likewise, an agent's belief is competent if and only if the agent uses their cognitive abilities to form it. What is the relationship between the competence and success dimensions of assessment? They are related in that competent performances are often successful (and the other way around), but they are ultimately independent. First, if a performance is competent, it is not necessarily successful. A competent chef’s finest dish may be ruined by the hair of a careless kitchen helper. Likewise, beliefs that are the product of cognitive  

See Chrisman, forthcoming, for relevant discussion of this assumption. Performances can have more than one aim, e.g. a cook can aim at making a delicious and a beautiful dish. Likewise, at least according to Sosa : , beliefs may be aimed at truth but also at some practical goal (e.g. self-relief ).



 -

ability may turn out to be false. One must recall that abilities and cognitive abilities are reliable, not infallible, which means that they can produce unsuccessful outputs from time to time. Second, even if a performance is successful, it is not necessarily competent. A chef’s apprentice might end up making a delicious cake by mixing random ingredients in random proportions and baking it all in the oven. Analogously, epistemically incompetent agents may form true beliefs by luck. For example, from time to time fortune-tellers get things right about the future by reading tarot cards. Success because of competence. Thus, there are performances (making a cake, forming a belief ) that are successful (delicious, true) by sheer luck. In other words, there are successful performances whose success is not because of (or due to, or in virtue of ) skill or ability. One popular way to understand this is as follows: to succeed in some endeavor because of ability (cognitive or otherwise) is for that success to manifest the exercise of such an ability. By way of comparison, a delicious cake made by a well-prepared pastry chef manifests the pastry-making skills of the chef. However, if an apprentice succeeds in making the same cake by randomly mixing ingredients, the fact that it is delicious does not manifest any pastry-making skill. Analogously, finding out the solution to a complex calculation task by properly following the right steps manifests mathematical competence. In contrast, correctly guessing the solution manifests none. Recall that virtue reliabilists think that knowledge is a matter of believing a true proposition p because of an exercise of cognitive ability. This now translates as the thesis that one knows p if and only if one’s cognitive success (i.e. the fact that one gets things right about p) manifests the exercise of one’s cognitive abilities. Crucially (and here is where the situational assumption of virtue reliabilism becomes relevant to the theory of knowledge), precisely because cognitive abilities are conceived as dispositions that lead to the formation of true beliefs in appropriate conditions, virtue reliabilists assume (without much argument) that an agent’s cognitive success manifests their cognitive abilities only if exercised in such conditions (understood in terms of the minimal characterization of appropriate conditions given in Section .). This means that knowing a proposition requires that one’s cognitive success manifests the exercise of one’s cognitive abilities under appropriate conditions thus understood. Let’s call this the appropriateness requirement. 

Virtue reliabilists (e.g. Turri ) tend to use the analogy of physical dispositions (e.g. solubility) to make this point: in the same way salt manifests its solubility only in appropriate conditions

Difficulty and Knowledge



The appropriateness requirement is not an innocent assumption. Virtue reliabilists use it to handle counterexamples. For example, according to Kelp () if one consults a thermometer that is fluctuating randomly and one (by luck) forms a true belief about the temperature on this basis, one does not possess knowledge because the situation is not appropriate for placing trust in reliable devices such as thermometers, where placing trust only in reliable devices is regarded as an ability. According to Greco (), the driver in the famous fake barn case does not manifest visual competence when forming the belief that there is a barn in front of him when looking at a real barn, because the presence of nearby fake barns makes the environment inhospitable for his typically reliable visual capacities. Finally, according to Sosa (), if a competent archer shoots an arrow and an unexpected gust causes its flight to deviate but a guardian angel puts it back on track, the archer does not manifest their shooting abilities because the circumstances are inappropriate – note that the archer’s circumstances are inappropriate in a different sense than the circumstances in the previous two cases: this time the likelihood of success is high; what happens is that the circumstances are too friendly in that shots would be successful no matter how badly one shot. Achievement. The fourth dimension of assessment is that of achievement. Achievements are a subset of successful performances. In particular, as understood in the virtue epistemological literature, they are successes because of ability – viz., successes that manifest ability. For example, the goal of a striker is an achievement if they score because of their abilities, i.e., if the goal manifests them. A delicious cake is an achievement if its deliciousness manifests the pastry-making abilities of the chef. An excellent dissertation in is an achievement if its quality manifests the Ph.D. student’s academic skills. Why do virtue epistemologists appeal to the notion of achievement if the success-because-of-competence dimension of assessment already serves to account for knowledge? Because it allows them to give the following

 

(e.g. when stirred into water but not into gasoline), the success of an agent’s performance is taken to manifest the agent’s abilities only if the conditions are appropriate. But there is an important respect in which physical and agential dispositions differ. Whereas it is physically impossible for salt to dissolve into gasoline – the probability of this happening is  – it is not impossible but difficult for agential abilities to be manifested in inappropriate conditions – the probability of this happening is greater than . See Section .. for more on this point. See the Temp case in Section .. Accordingly, we can distinguish two senses of inappropriateness: inappropriate conditions for achieving success and inappropriate conditions for manifesting ability. Most inappropriate conditions are inappropriate in both senses, but sometimes, as Sosa’s case illustrates, they come apart.



 -

simple solution to the value problem, i.e., the problem of explaining why knowledge is more valuable than mere true belief. The argument, as reconstructed by Pritchard (Pritchard et al. : ) goes like this: ) An achievement is a success because of ability, i.e., a success that manifests ability; ) knowledge is a cognitive success that manifests cognitive ability; therefore, ) knowledge is a cognitive or intellectual achievement (the cognitive achievement thesis); ) since achievements are finally valuable, ) so is knowledge. ..

Against the Appropriateness Requirement

There is no doubt that the virtue reliabilist explanation of the final value of knowledge is simple and elegant, but the cognitive achievement thesis, when combined with the aforementioned appropriateness requirement, leads to what we can call the problematic cognitive achievement thesis: the idea that knowledge is a cognitive achievement, understood as a cognitive success that manifests the agent’s cognitive abilities only if the conditions are appropriate (as characterized above). It does not take much argument to see what is problematic with the problematic cognitive achievement thesis: it is precisely when the circumstances are inappropriate that human abilities may be most manifested (if successfully exercised) and therefore when the greatest achievements may be accomplished. Indeed, many successes count as great achievements precisely because the circumstances in which they are skillfully attained are inappropriate. Thus, what is problematic with the thesis is the appropriateness requirement. By way of illustration, consider a feat such as the first successful ascent of Mount Everest by Tenzing Norgay and Edmund Hillary in . Three decades of failed attempts are a good indication that the conditions on Everest were not the most appropriate for the climbing skills of mountaineers at that time. After all, appropriate conditions are the sort of conditions in which one’s abilities are reliable enough and hence conditions in which it is likely that one will succeed by exercising such abilities. But Norgay and Hillary succeeded against all odds (i.e. in spite of being in inappropriate conditions) thanks to their climbing skills. The same can be said about Annette Fredskov who, after being diagnosed with multiple sclerosis, ran  marathons in  days in  and , despite conditions of extreme tiredness and muscle soreness. Another illustrative feat is that of Fernando Parrado and Roberto Canessa, two of the sixteen survivors of the  airplane crash in the remote Andes, who after two

Difficulty and Knowledge



months of extreme survival conditions decided to embark on an expedition to look for help and save the rest. It took them ten days of persevering through the merciless peaks and valleys of the Andes, without mountain clothing or proper equipment, in conditions of low oxygen, extreme exhaustion, sickness and bitter cold, to reach the end of the snowline. Soon after, they found a horseman who contacted the authorities, who immediately set up a rescue team. In these success stories great abilities are manifested in inappropriate conditions – indeed, the magnitude of these achievements is greater precisely because of it – which not only invalidates the idea that abilities and skills can only be manifested in appropriate conditions, but also the corresponding claim that achievements are successes that can only be accomplished in such conditions. The same considerations apply to the intellectual domain. Consider intellectual feats, such as decisive breakthroughs in science. Take the Poincaré conjecture. This Millennium Problem withstood the efforts of mathematicians for almost a century, until Grigori Perelman solved it in . The inappropriateness of the conditions when trying to find proofs for problems of such a caliber does not obviously lie in the harshness of the environment (like Mount Everest or the Andes), but in the fact that one may need to connect fields of mathematics that apparently have little to do with each other, or resort to complex mathematical knowledge that is beyond the understanding of even very skillful mathematicians, or develop new techniques or concepts. Such enormously difficult problems are not unfamiliar to philosophers either. Consider a hard problem: How do physical processes in the brain give rise to the conscious experience of the mind and the world? To answer 



Two caveats are in order. First, succeeding in inappropriate conditions does not necessarily mean that the relevant success is an achievement, let alone a feat, as one may succeed completely by fluke in such circumstances. Second, inappropriate conditions are typically such that they make success unlikely, but we have also seen that some circumstances may be inappropriate precisely by making success unduly likely or by masking competent success. Sosa’s guardian angel case and the case of the archer shooting metal arrows at a giant magnetic target are two examples. Unlike the usual kind of inappropriate conditions (conditions in which success is unlikely), this unusual set of inappropriate conditions can make things easier and thereby lead to lesser achievements. A different source of inappropriateness when pursuing an intellectual endeavor is not its complexity but adverse social or cultural circumstances. For instance, the outstanding discoveries of Marie Curie (who holds two Nobel prizes for her research on radioactivity and the discovery of new chemical elements such as radium and polonium) are even more remarkable in view of the fact that her scientific career was often hindered by male colleagues simply because of the fact that she was a female researcher. These kinds of inappropriate conditions do not make the relevant success unlikely given the exercise of reliable intellectual abilities. Instead, what they make unlikely is the very exercise of such reliable abilities.



 -

this question one may need to connect different disciplines that apparently have little to do with each other, such as neuroscience and philosophy, or use knowledge about the world or the brain that is beyond current scientific understanding (e.g. how can we manipulate physical matter so as to give rise to conscious states?), or perhaps develop rather radical ideas (e.g. that consciousness is a fundamental property of the world, just as are space, time or mass). Given these inappropriate intellectual conditions (i.e. the current state of the art), it is very unlikely that philosophers or scientists, by exercising their reliable intellectual capacities, will come up with a satisfactory solution to the hard problem in the short term (at least one that would be accepted by all parties in dispute). However, suppose that someone did: that person’s solution would be considered an extraordinary intellectual feat that we would be happy to call ‘knowledge.’ The bottom line is that some decisive breakthroughs count as cognitive or intellectual achievements despite being made in inappropriate conditions, i.e., in conditions where intellectual success is unlikely given the exercise of otherwise reliable cognitive abilities. Therefore, the cognitive achievement thesis that we have previously labelled ‘problematic’ – that is, the thesis that knowledge is a cognitive achievement, understood as a cognitive success that manifests cognitive ability only if the conditions are appropriate – is false because the appropriateness requirement it assumes (namely, the assumption that knowledge requires the manifestation of cognitive ability in appropriate conditions) is false. In order to avoid possible misunderstandings, let me make more explicit my concerns about this flawed assumption of virtue reliabilism. The appropriateness requirement says that an agent’s cognitive success can only manifest their reliable cognitive abilities in appropriate conditions, i.e., in conditions that sustain or guarantee the threshold of reliability that is presumably enough for knowledge and hence conditions that make cognitive success likely. As one might have noticed, this requirement introduces a qualitative (i.e. an all-or-nothing) criterion that splits circumstances into two types: ) appropriate conditions, i.e., those in which cognitive abilities are reliable enough, cognitive success is likely and hence knowledge can be acquired; and ) inappropriate conditions, i.e., those in which cognitive abilities are not reliable enough, cognitive success is unlikely and knowledge cannot be acquired – because no eventual cognitive success can manifest the relevant cognitive abilities when exercised in such circumstances. However, it is actually the other way around: it is precisely when circumstances are inappropriate that our intellectual abilities may be most manifested (if successfully exercised) and the greatest breakthroughs

Difficulty and Knowledge



in knowledge may be accomplished. Note that the use of examples of intellectual feats such as decisive breakthroughs in science is no coincidence: an intellectual feat is the best example of a cognitive achievement, and if anything deserves to be called ‘knowledge,’ this is a breakthrough in science. Yet, these are things that are sometimes accomplished in impoverished epistemic conditions. The final reason to drop the appropriateness requirement is that splitting circumstances into those deemed ‘appropriate’ and ‘inappropriate’ offers an incorrect representation of the relationship between difficulty and knowledge. The requirement says that no cognitive success can manifest the exercise of cognitive ability (and hence amount to knowledge) when the relevant conditions are inappropriate. Therefore, knowledge in inappropriate circumstances is not just difficult to achieve (as one would naturally expect) but impossible. But then it is unclear how human learning can be explained. In order to expand the range of action of our abilities (the conditions in which are reliable), we often need to expose ourselves to conditions outside that range, i.e., to conditions in which our performances are less reliable (even very unreliable) and hence conditions where success is unlikely. This is how we typically come to learn new things, i.e., how we come to acquire new practical and propositional knowledge. In this way, the learning process is such that we incorporate new circumstances to the range of action of our abilities. Yet this is surely something that does not happen automatically, but gradually – hence why it seems misguided to split circumstances categorically, i.e., into circumstances where one can know and circumstances where knowledge (learning) is impossible. The reality is that at some point during the learning process we may succeed cognitively despite it being at something unlikely. In some such cases, we consider the relevant cognitive successes to be achievements, even feats, and indeed, some of them (such as extraordinary breakthroughs in science) may be properly called ‘knowledge.’

. Knowledge as a (More or Less) Difficult Cognitive Achievement The key tenet of virtue reliabilism that knowledge is a cognitive achievement (a cognitive success that manifests cognitive ability) turns out to be false on the assumption of the appropriateness requirement. 

Note that if φ-ing is difficult for S, the probability that S will φ may be low, but greater than zero, whereas if φ-ing is impossible for S, the probability that S will φ is zero.



 -

This requirement also leads to an incorrect representation of the idea that knowledge can be difficult. In what follows, and to amend the latter problem, I will combine the virtue reliabilist framework with recent work on the notions of achievement and difficulty. This will help understand how knowledge relates to difficulty without appealing to considerations about appropriate or inappropriate conditions, and will hopefully shed some light on the nature of knowledge as well (especially on the thesis that knowledge is a kind of cognitive achievement). If the cognitive achievement thesis is correct, then the extent to which knowledge is difficult corresponds to the extent to which it is difficult for a cognitive success to qualify as a cognitive achievement in the way virtue reliabilists understand this notion. In what sense can a belief be considered a more or less difficult achievement? Whereas difficulty is a property of activities, performances or processes and then derivatively of the states that result from them, achievements are states, namely the results of successful performances. Consider Usain Bolt breaking the m world record in the  Berlin World Championships in Athletics. The activity/performance/process of running faster than anyone else in an official competition is very difficult. Having crossed the finish line in . seconds, the state of being beyond the finish line in . seconds is an achievement. It is an achievement because (in keeping with the virtue reliabilist understanding of the notion of achievement) it is a success that is due to an exercise of ability. Likewise, beliefs are states that (according to virtue reliabilism) count as intellectual or cognitive achievements if their correctness (i.e. their ‘success’) is due to the exercise of cognitive ability. This is the sense in which beliefs can be considered achievements. ..

Achievements: Their (Class-Relative) Nature vs. Their (Value-Relative) Magnitude

In a narrow sense, by ‘achievement’ we mean an exceptional accomplishment, i.e., a feat, or as Bradford () puts it a “capital-A achievement.” 



Three authors who have recently cast light onto the notions of achievement and difficulty are Gwen Bradford ; Hasko von Kriegstein , ; and Alexander A. Guerrero . This section is largely indebted to their work. Because I am interested in investigating how knowledge, qua cognitive achievement, can be more or less difficult, here I am assuming that achievements can be more or less difficult, i.e., I don’t consider being difficult (i.e. difficult to a high degree) a necessary component of achievements. There is some controversy concerning this point, though. See Bradford  and von Kriegstein , for extensive discussion.

Difficulty and Knowledge



The first ascent to Mount Everest, the proof of the Poincaré conjecture or Marie Curie’s discoveries are practical and intellectual achievements in this sense: they are successes because of an exceptional display of ability and effort. In a broader sense, however, we call an achievement any success because of ability, even if mundane, such as tying a shoelace, carrying a heavy shopping bag, threading a needle or finding out whether it’s raining outside. What is the difference between capital-A and mundane achievements? In a nutshell: their magnitude. The magnitude of an achievement typically has to do with the effort invested and the skill needed to attain it. But very effortful or skillful performances can still result in mundane achievements. For example, counting the grains of sand in a heap one by one is no doubt a very effortful success. But it is not a capital-A achievement. Burping at  decibels is a very skillful success, but again, not a capital-A achievement. The reason is that the resulting knowledge (in the former case) and the resulting loudness (in the latter) have little significance or value. This suggests that the magnitude of achievements features a significance or value condition: the more significant or valuable a practical or intellectual success, the more of an achievement it is. Knowledge can be a capital-A intellectual achievement, such as the proof of the Poincaré conjecture, the (eventual) discovery of the ultimate cure for cancer or an (eventual) satisfactory solution to the hard problem of consciousness. But for the most part, knowledge is a mundane achievement, such as deducing the answer of a problem in a math exam or knowing the departure time of the next train from a station. As with the magnitude of practical achievements, the magnitude of intellectual achievements depends to a large extent on how significant or valuable they are. However, the value or significance of a cognitive success (of the true proposition believed) is not relevant to whether it counts as an intellectual achievement. The crucial element is whether it manifests the cognitive 



Suppose that on three different occasions I sign three cheques for different fundraising events for cancer research: one for $, one for $ and the third for $,,. On the present proposal, there is an increase in the magnitude of the achievement, even if its nature is the same in the three cases (the effort and the skill manifested are the same). Donating $,, is certainly more of an achievement than donating $ or $ (in the sense of having more personal costs, greater benefits for the benefactor and more impact on the target group or society). I use the terms ‘cognitive’ and ‘intellectual’ interchangeably. That said, the term ‘intellectual’ denotes successes or achievements that result from higher forms of inquiry (e.g. scientific research), whereas ‘cognitive’ denotes successes or achievements that result from more basic cognitive faculties, such as perception.



 -

abilities of the agent as well as the effort invested. To see why value considerations are not important, consider two scenarios. First scenario: to pass the time, someone correctly counts the grains of sand in a heap one by one. Second scenario: finding out how many grains of sand are there in a heap (with equal number of grains) is the only way to quench the thirst of blood of a furious god and save humanity from extinction; a hero succeeds in counting the grains, one by one. In both scenarios, the relevant performances lead to the same cognitive success, i.e., a true belief about the number of grains, which manifests ability and effort to the same degree. The difference between the two cases lies in the significance or value of the resulting knowledge. Does this mean that the relevant cognitive success in the low-stakes scenario doesn't count as a cognitive achievement? No, it is certainly less of an achievement (not a capital-A achievement for sure), but a cognitive achievement after all. In other words, whereas the value of an achievement decreases or increases its magnitude, it does not affect its nature or status qua achievement. Since the aim is to investigate how difficult it is for a cognitive success to qualify as a cognitive achievement in the relevant sense and thus as knowledge, in what follows I will omit considerations about the value of cognitive achievements. Suppose with virtue reliabilism that knowledge is a cognitive achievement (which may have different magnitudes). There is a problem in the offing. On the one hand, knowledge is considered a universal phenomenon – at least universal in the sense that it is attributed to and may be instantiated by many creatures. On the other hand, the notion of achievement is always relativized to a reference class, which may include only one individual (i.e. successes that count as achievements only for those who produce them). For example, a baby’s first steps are an achievement for the baby as an individual and perhaps also for the entire baby collective, but such a remarkable success might not be considered an achievement for healthy adults (at least not in the capital-A sense of achievement). How can knowledge, qua achievement, be considered universal if the notion of achievement is relative? The first thing to note is that some achievements are considered universal at least in the sense that the relevant reference class is humanity as a whole. Consider for instance the eradication of infectious diseases or a  

See von Kriegstein , , for helpful discussion on the distinction between the magnitude and the nature of achievements. Again, see Bradford  and von Kriegstein , , for extensive discussion of this point.

Difficulty and Knowledge



stable world peace: they are considered universal accomplishments of humanity. We also tend to take this more comprehensive stance toward relevant successes that are accomplished for the first time in history, such as the first ascent of Mount Everest or the first time a human landed on the moon. But the sense in which knowledge is a universal achievement is not the same as the sense in which these achievements are considered universal: the latter are universal because they are feats, i.e., because of their magnitude, but as we have seen knowledge can be mundane. The reason knowledge constitutes a universal achievement is the fact that it is a distinctive state that can in principle be possessed by any being capable of holding doxastic states with propositional content or at least capable of having contentful mental representations. This includes humans, other higher-order mammals, many animals with less sophisticated nervous systems and complex forms of artificial intelligence. Of course, this does not mean that beings in this list actually have knowledge, because their methods of belief formation may be unreliable. The idea is rather that knowledge is a kind of state that we would universally attribute to doxastic agents only, i.e., to agents with belief capabilities. An unreliable doxastic agent (e.g. a brain in a vat) may not be credited with actual knowledge but could in principle come to acquire knowledge. By contrast, we would not say the same thing about a stone, a chair or a mountain: they could not in principle come to acquire knowledge – in certain key senses of ‘could.’ To see this more clearly, suppose that the actual world contains two types of things that by their constitution cannot have knowledge: stones and brains in vats. It takes much more of a departure from the actual world to change the constitution of a stone so it can have knowledge than to change the constitution of a brain in a vat. After all, the latter may be restricted by constitution to fairly unreliable ways of forming beliefs (barely anything more than chance), but unlike the stone, brains in vats satisfy the precondition that makes knowledge universal: having the capacity to form contentful mental representations (beliefs). This is why the brain in a vat, unlike the stone, can in principle have knowledge. What reference class is relevant to evaluate knowledge as a distinctive cognitive achievement then? Unsurprisingly, the relevant reference class 

What exactly makes knowledge distinctive is an open question: virtue reliabilist might argue that only true beliefs that have the ‘knowledge’ trademark manifest cognitive ability; others might argue that knowledge is distinctive because it serves to flag good informants or stop inquiry. The way we fill in the details is not as relevant as the general point that, for all beliefs possibly held by doxastic agents, there is a subset that we distinctively call ‘knowledge.’



 -

that is assumed by default in non-contextualist epistemology (at least in virtue epistemology) is the class of agents capable of forming contentful doxastic representations. This reference class may of course be subdivided into a narrower reference class when it comes to evaluate whether there is actual knowledge (e.g. the class of agents who form contentful doxastic representation in reliable ways, who in doing so manifest cognitive ability, agents who follow their evidence and so on). In epistemology, there are of course many competing views about how to define this reference subclass of knowers in a context-independent way. Why should this reference subclass not be fixed by context? Because if the reference class relative to which an agent’s cognitive success qualifies as a knowledge-level cognitive achievement included only individuals that are salient in the context of assessment, then too many cognitive successes would count as knowledge. For example, suppose that your preferred way of forming beliefs is wishful thinking. Suppose that you get things right this time. Is your cognitive success then a cognitive achievement? Given that most of the time you form false beliefs, and given the class relativity of the general notion of achievement, this does count as a cognitive achievement for you in your present context (after all, albeit luckily, you got things right). But relative to a relevant non-contextually fixed reference subclass of doxastic agents (e.g. reliable doxastic agents, doxastic agents whose cognitive successes manifest their cognitive abilities, doxastic agents who follow their evidence and so on) your cognitive success does not represent a (knowledge-level) cognitive achievement. To be clear, some cognitive successes count as cognitive achievements but do not amount to knowledge because there are always contextually fixed reference classes of individuals following bad epistemic practices (e.g. wishful thinking) for which such cognitive successes count as cognitive achievements. In particular, their less epistemically defective practices (e.g. lucky wishful thinking) can be considered achievements for them. However, once again, relative to the relevant non-contextually fixed reference subclass of doxastic agents that serves to evaluate whether there is actual knowledge, bad epistemic practices are what they are: bad epistemic practices. In this way, the class relativity of the notion of achievement is no impediment to conceptualizing knowledge as a (special kind of ) cognitive achievement.



When the kind of knowledge is reflective knowledge, the typically assumed reference class excludes human infants and non-human animals.

Difficulty and Knowledge



.. Difficulty: Likelihood, Skill, Effort and Resources Similar considerations apply to the notion of difficulty. Difficulty is always relative: when we deem something difficult, we always think of it as difficult for the agents of a relevant reference class. This is obviously relevant to the question of how knowledge can be difficult. How is the relevant class of agents fixed in each case? Here is Bradford: The relevant class of agents for any particular instance of difficulty is determined, I think, fairly straightforwardly. Like many things of this nature, it is determined by context. Whatever activity we are interested in evaluating, it seems that we can quite easily pick out the relevant class of agents for whom it may or may not be difficult. The default may be the class of adult human beings, but of course many activities are difficult for four-year-olds, but not for the typical adult. Whatever class it is that is relevant for the evaluation of whether or not some activity is difficult is generally apparent from the context of the activity, or can be discerned or stipulated without much controversy. (: )

How does this bear on the explanation of the idea that knowledge can be difficult, particularly once knowledge is theorized as a special kind of cognitive achievement? As we have just seen, when we evaluate whether a given cognitive success amounts to a knowledge-level cognitive achievement (i.e. a cognitive achievement in the relevant sense and not only for the agent in question), the reference class that epistemologists assume by default (i.e. not contextually) is a subclass of the class of agents capable of holding doxastic states. The idea is that if relative to this reference subclass a cognitive success counts as a cognitive achievement, then it can be universally considered knowledge (universally for all doxastic agents). However, when it comes to the assessment of whether a certain proposition is difficult to know, things are different. If Bradford is right, here the relevant reference class is fixed by the context of assessment, not by default. This means that knowing can be difficult relative to a reference class that may include any number of individuals from all agents capable of holding beliefs to one single agent. To see this, take a simple cognitive success, such as answering correctly what  +  equals. Is it difficult to know what  +  equals? It depends. If the relevant reference class only includes educated adults, then it is not difficult. If it only includes twoyear-olds, then it is difficult. It is a matter of what sort of individuals the salient reference class includes. We can generalize these ideas for any practical or intellectual endeavor E in the following way: If the most skillful agents in a contextually fixed



 -

reference class are unlikely to succeed in E, then E is difficult for all members of the reference class. Conversely, if the less skillful agents in a reference class are likely to succeed in E, then E is easy for all members of the reference class. Comparatively, E is more difficult for members of reference class R than for members of reference class R if the less skillful agents in R are more likely to succeed in E than the most skillful agents in R. In this way, finding out what  +  equals is more difficult for twoyear-olds than for educated adults because the most skillful two-year-olds are less likely to arrive at the correct answer than the less skillful adults. So far I have been assuming a probabilistic conception of difficulty according to which how difficult (or easy) is it for an agent to reach a goal corresponds to how likely the agent is to reach it. In the previous paragraph, I also assumed that the factor that bears most weight in terms of an agent’s chances of success with respect to some goal – and therefore for how difficult it is for the agent to reach that goal – is how skillful that agent is in the relevant domain. But as Bradford () and von Kriegstein (, ) have suggested (and as pointed out before), another relevant factor that determines the difficulty of reaching a goal is the amount of effort invested. What is the relationship between invested effort, exercised skill and likelihood of success? Von Kriegstein nicely summarizes how they are related: 



The present project explores how the concept of difficulty relates to intellectual endeavors in general and especially in connection with the notion of achievement. A different question, which cannot be answered from the armchair, is how to measure difficulty in each domain. In some domains, there are objective metrics, such as videogames or the high jump. In the latter case, for example, the higher the bar is, the more difficult it is to jump over without dislodging. In some other domains, there are also metrics for difficulty, but they are subjective. For example, in bouldering there are detailed scales that measure the physical difficulty of a ‘boulder problem’ (i.e. the route that a climber takes to complete a climb). The first person who sets the route makes a subjective assessment of the relevant grade, which may be then slightly modified by consensus. In some other domains, there are no scales or metrics, but there is a shared implicit agreement on what is easy and what difficult in the domain. For example, in the study plan of a math degree linear algebra is typically taught before complex analysis because it is easier, just as propositional logic is taught before modal logic in philosophy degrees. In other areas, it is not obvious what is more difficult, e.g., some might find Frege an easy read, but some might find it difficult and opt for Husserl instead. The same sorts of considerations apply to the measurement of skill levels. Sometimes, it is easy to pin down the features that make someone an expert in the domain, but sometimes this is less obvious. Empirical research often helps shed light on the issue. For example, in the case of bouldering one study concluded that “more advanced ability climbers make greater use of foot holds, with associated lowering in physiological response (oxygen uptake and heart rate) across all slope inclinations” (Baláš et al. : ). In the domain of physics, another study concluded that a key factor that distinguishes experts from novices in the representation of physics problems is that “the experts initially abstract physics principles to approach and solve a problem representation, whereas novices base their representation and approaches on the problem's literal features” (Chi et al. ). See von Kriegstein , for this account.

Difficulty and Knowledge



Actions generally involve a combination of effort and skill, and this combination (together with facts about the goal or intention) determines the likelihood of success. As a general rule skill and effort can be substituted for each other to some degree: a less skilled agent can make up for her lack of talent with great effort, and a lazy agent may compensate her lack of effort through great skill. Keeping skill level constant, then, we should generally expect a correlation between effort and contribution to likelihood of success. That means that effort, via this contribution, has an achievementenhancing effect. (Von Kriegstein : )

How is effort to be understood in this context? A plausible view, advanced by von Kriegstein, understands effort as a matter of dedicating resources. Von Kriegstein conceptualizes the relevant resources as internal, but does not say much about their nature (only that they can be physical or mental). Let me elaborate on the notion of resource then. The term ‘resource’ may be used to refer to whatever method one has when it comes to accomplishing a certain task (φ-ing). However, you may have a method that is completely useless for φ-ing. When that happens, and if you have no other method at your disposal, we would correctly say that you have no resources for φ-ing. It seems, then, that the notion of resource features some sort of reliability condition. One plausible idea is that one’s resources for φ-ing are one’s abilities, where these need not be internal, as there may be abilities based on extended cognitive processes. However, this might not yet suffice for capturing the notion of resource: it also seems to feature an availability condition. Suppose that you want to hunt a rabbit for dinner and that you are very skillful at hunting rabbits with your bow but someone has stolen it. In a sense, we might say that you have the resources for hunting a rabbit: you are a skillful hunter after all. In another (I think more plausible) sense, you don’t have the resources: while you have the relevant abilities, you cannot exercise them. It seems then that in order for an agent to have resources for φ-ing, they not only need to have the ability to φ but also be in a position to φ, i.e., be in conditions that enable them to exercise their ability to φ. In other words, exercising an ability to φ is something that must be available in the circumstances of φ-ing. The fact that certain enabling conditions for exercising an ability to φ are in place does not entail that the conditions are appropriate for φ-ing. Conceptually, these are two different things. In the previous example, you 

Guerrero  makes a similar point by distinguishing two kinds of difficulty: effort and skill related.



 -

cannot hunt a rabbit because you don’t have your bow. Having a bow is a condition that enables you to exercise your ability to hunt rabbits and, consequently, when you have a bow at your disposal, you have the resources for hunting a rabbit – the presence of rabbits is also an enabling condition. But the fact that you are in conditions that enable you to exercise your hunting skills is compatible with the fact that you are in ‘inappropriate’ conditions for hunting, i.e., conditions in which hunting a rabbit successfully is difficult or unlikely despite your abilities (e.g. strong winds, poor light conditions and so on). These latter kinds of conditions determine the success (or failure) of your performances rather than enable the exercise of your abilities. Having resources for φ-ing only requires that your enabling conditions are in place. It does not require that you are in favorable determining conditions (i.e. conditions that make it likely that you will φ). A clarificatory point is in order, as some might have noticed an ambiguity between two competing conceptions of the notion of resource in the previous paragraphs. On the present understanding of the notion of resource, you may have resources to φ even in conditions in which φ-ing is unlikely (unfavorable determining conditions). Call this the permissive conception. But it could be argued that if φ-ing is unlikely in certain conditions, then you don’t have the resources to φ in such conditions. Call this the restrictive conception. I think the ambiguity originates in a prior ambiguity regarding the notion of ability. On a restrictive conception, you don’t have the ability to φ if your conditions are such that φ-ing is unlikely. Having an ability to φ, in this restrictive sense, consists in having a complete ability to φ, i.e., in not only possessing whatever constitutional basis makes you disposed to φ in the relevant way but also in being adequately situated for φ-ing (both internally and externally). On a permissive conception, by contrast, you can have the ability to φ by having the relevant constitutional basis even if you are not adequately situated for φ-ing, i.e., even if you are unlikely to φ. What conception of resources should we adopt? There are reasons to opt for the permissive understanding, as the restrictive reading leads to counterintuitive results. Suppose that two agents have the same constitutional bases for the disposition to φ but are in different situations. Do we want to say that they have different resources for φ-ing? A more intuitive claim seems that they have the same resources, but they are in different situations. The restrictive reading does not allow to say that, however. If this is not convincing enough, consider another reason. Imagine a quite skillful agent in unfavorable conditions and a not so skillful agent in favorable

Difficulty and Knowledge



conditions, so that the likelihood that they will φ is the same. The restrictive conception must say that they have the same resources for φ-ing, but that seems wrong. A trained mathematician has objectively more resources for solving a mathematical problem than a primary school student, even if in a given situation the likelihood that they will solve it is the same. With this account of the notion of resource in place, what is effort then? According to von Kriegstein, it is the degree to which one’s resources are put to use. As a general rule, the more resources one has at one’s disposal, the more likely it is that one will succeed in accomplishing a task and, therefore, the easier the task will be. But this needs qualification. First, as we have seen, the likelihood of success also depends on the circumstances, which may be more or less demanding. Second, as von Kriegstein (: ) points out, “potential resources simply have no bearing on the likelihood of success as long as they remain untapped.” Given the previous characterization of the notion of resource, to ‘tap’ a resource for φ-ing is just to exercise the corresponding abilities in conditions that enable such an exercise. With these two caveats in place, we may say that what determines the likelihood of success and therefore the degree of difficulty in reaching a goal is how many resources one actually uses of those at one’s disposal as well as the kind of determining conditions one is under. What is the time span in which one can use one’s available resources for reaching a goal? Von Kriegstein (: ) helpfully sheds light on the issue with another proposal: the relevant time span is “the interval between the adoption of the goal and the last moment when it is possible to reach it.” According to von Kriegstein, two types of factors can set a limit beyond which it is not possible to reach a goal with one’s resources. The first factor is the very content of the goal. Some goals, such as filing a legal document or answering the questions of an exam, have deadlines, i.e., they do not count as ‘reached’ unless they are accomplished within a certain time. One can of course complete these tasks after their deadlines, but one would be thereby reaching different goals. The second factor that can set a limit beyond which it is no longer possible to reach a goal are the circumstances in which one tries to reach it. For example, suppose that the goal is to prepare a delicious dish with the fresh ingredients in the fridge. This goal 

Time should intuitively count as a resource too, one might think. That is probably right. For the sake of explanation, however, I follow von Kriegstein in treating time as something separately from one’s resources, so we can say that there is a time span for using one’s available resources which is not itself part of one’s resources.



 -

can be only accomplished while the ingredients are fresh, otherwise the dish will not be delicious no matter how well one cooks. Finally, as von Kriegstein points out, the limiting case of a deadline being set by the circumstances is the very death of the agent. Complex mathematical problems are often like this: mathematicians die before solving them. The temporal dimension is important for understanding the relationship between difficulty and knowledge. According to the main tenet of virtue reliabilism, an agent has knowledge if and only if their cognitive success manifests their cognitive abilities. But this can be now qualified as follows: when an agent has knowledge at t, their cognitive success manifests the exercise of their cognitive abilities in a given time span up to t, which means that what is manifested is not only how competent they are, but also the degree to which they puts their cognitive abilities to use in that interval of time, i.e., their effort. The reason the notion of effort should be included in the virtue reliabilist analysis of knowledge is that, as we have seen, ability and effort can be substituted for each other. The implication of this is that the same piece of knowledge can manifest varying degrees of cognitive ability as long as a higher degree of effort offsets a lower degree of ability. By way of illustration, suppose that an expert solves a complex Sudoku puzzle easily and quickly. Their cognitive success (i.e. their knowledge of the solution) manifests great cognitive ability and little effort. In comparison, a novice is less likely to solve the puzzle, i.e., it is more difficult for them. But with dedication and effort they can solve it, and if they solve it their cognitive success can be considered knowledge with the only difference being that, unlike the expert’s cognitive success, it will manifest more effort than cognitive ability. This interplay between cognitive ability and effort in increasing an agent’s chances of cognitive success is at the core of the relationship between difficulty and knowledge.

.

Knowledge As a (Greater or Lesser) Challenge

This section will condense the foregoing discussion in a simple account of the relationship between difficulty and knowledge. The account is based on two main points. The first point is that, although they often go 

In both cases, the relevant cognitive successes count as cognitive achievements. In the expert case, the relevant cognitive achievement presumably has a greater magnitude because the expert solves the puzzle more efficiently (she needs to invest less effort). This gives rise to an interesting question concerning the value of knowledge: is it more valuable to know something using a greater degree of skill or a greater degree of effort?

Difficulty and Knowledge



together, the difficulty of succeeding in a cognitive endeavor – which depends on how competent the agent is and on how much effort they put in – is different from the difficulty of manifesting competence and effort. There are cases in which they come apart. The second point is that the difficulty of accomplishing a cognitive task (such as acquiring knowledge) does not require the intentional adoption of a goal, because knowledge can be conceived as a challenge of the situation that can be met unintentionally. Let’s start with the second point. Cognitive tasks are diverse: they range from effortful forms of inquiry (e.g. solving complex mathematical problems) to automatic information processing (e.g. perceiving the color of an object). While we may intentionally aim at accomplishing a cognitive task (e.g. solving a differential equation), most cognitive tasks do not need the intentional adoption of cognitive goals to the extent that most of the times we simply have no propositional attitude toward the proposition believed prior to believing it. This marks an important difference with success in the practical domain, which is typically the product of intentional action. That being so, we cannot understand the difficulty of knowing in terms of intentionally adopted cognitive goals being reached. Instead, as I will propose next, we can better understand it in terms of situational challenges being intentionally or unintentionally met. The general idea, in a nutshell, is the following: in each situation, circumstances or context, there are potentially many challenges that an individual in the situation can meet. Each challenge poses different degrees of difficulty for the individual in the situation, where how difficult it is for an individual to meet one of these situational challenges depends on the type of challenge in question, their level of skill and the amount of effort invested. By way of illustration, consider a soccer game. There are potentially many challenges that a striker can meet: scoring a goal, scoring a hat-trick, making two assists, dribbling past three defenders in a single play, following the orders of the coach, avoiding being shown a yellow card, not getting injured and so forth. All these challenges have different degrees of difficulty for the striker, which may be different from the degrees of difficulty for other players exhibiting different levels of ability and effort. For example, it is certainly easier for a striker to score a goal or to dribble past another player than for a defender, because strikers are in general more competent at those tasks than defenders. Likewise, it is presumably more 

I use the terms ‘situation,’ ‘context’ and ‘circumstances’ loosely and interchangeably.



 -

difficult for a striker to avoid getting injured than for a defender (strikers tend to be more exposed). In a given situation, there are challenges for the agents involved that are irrelevant to them or the domain. In soccer, for instance, it is irrelevant whether a player does a perfect déboulé or lands a jump in fifth position, but this is very relevant in ballet. In addition, impossible challenges, because they are impossible, are also typically deemed irrelevant (e.g. the challenge of running faster than light in a -meter race). The bottom line is that what challenges are relevant or salient in a situation is determined by context or by the domain itself. Crucially (and here comes one of the main points of the section), an agent in a certain context or situation does not need to intentionally adopt a goal to meet the challenges that are relevant or salient. For example, a striker’s main aim may be to score a goal, but they may end up scoring a hat-trick, in which case they overcome two challenges by reaching one goal. The same applies to cognitive tasks. For example, consider perceptual tasks. Suppose that an agent enters a room with only two objects. There are several perceptual challenges available to this agent in the situation that they may accomplish using their perceptual abilities. For example, they can find out how many objects are there in the room, what color these are, their shape, their location and so on – richer perceptual environments obviously feature many more and more complex perceptual challenges. The point is that, no matter how many perceptual challenges are available in the situation, one does not need to intentionally adopt the goal of meeting them in order to actually meet them. It is possible to overcome them spontaneously or unintentionally by carrying out other perceptual tasks as long as these challenges are available in the situation. Another way to put it is this: as in the case of soccer, one can meet cognitive challenges that are different from one’s initial cognitive goals. For instance, when an ophthalmologist asks you to tell the letters on the eye chart, you may intentionally adopt this perceptual goal. But suppose that while you are trying to identify the letters, you notice that the eye chart is white and has a rectangular shape. In this way, you overcome two perceptual challenges (what is the shape of the eye chart? what is its color?) in addition to the perceptual challenge you were trying to meet in the first place (what are the letters in front of you?). What is the time span in which an agent can put her resources to use in order to meet a challenge in a given situation? As we have seen in Section .., this can depend on the content of the relevant challenge (e.g. if it involves a deadline) or on the circumstances themselves. For example, the

Difficulty and Knowledge



challenge of scoring three goals in a game has a fixed time span: the duration of the game. By contrast, the time span of complex intellectual challenges often corresponds to the time span of the lives of those trying to overcome them: it is not rare that a scientist devotes their entire academic life to challenges such as finding the cure for HIV or a very hard mathematical problem. For more mundane cognitive challenges, such as simple perceptual tasks, the time span typically depends on the circumstances. For example, the time span for perceiving the location of an object is the time the object is at that location. My proposal, then, is that knowing is a matter of meeting a more or less difficult cognitive challenge in a given situation. However, knowing is a special kind of challenge in that it does suffice that one succeeds cognitively, i.e., it is not the challenge of succeeding cognitively simpliciter, but the challenge of succeeding cognitively in such a way that one’s cognitive success manifests one’s cognitive abilities and effort. Let’s generalize these ideas for any practical and intellectual endeavor. Let φ-ing be any practical or intellectual task (e.g. performing an action, believing a proposition). In a given situation, we can distinguish three general types of challenge: (C) The challenge of successfully φ-ing at t: an agent overcomes this challenge if and only if they are successful in φ-ing at t (e.g. if and only if they form a true belief ). (C) The challenge of competently φ-ing at t: an agent overcomes this challenge if and only if they use their abilities to φ at t (e.g. if and only if their belief is produced by their cognitive abilities). (C) The challenge of successfully φ-ing at t because of one’s abilities and effort: an agent overcomes this challenge if and only if their successful φ-ing at t manifests the exercise of their abilities and effort, i.e., if and only if their success is an achievement (e.g. if and only if the fact that they come to believe a true proposition is due to a dedicated exercise of cognitive ability). Acquiring knowledge is a challenge of the latter type: it is not the mere challenge of succeeding cognitively in a competent way but in a way that makes one's cognitive success qualify as a cognitive achievement in the relevant sense. Meeting C entails meeting C and C, but the converse does not hold. Someone can get things right by wishful thinking (i.e. they meet C) but this cognitive success does not manifest cognitive ability. Someone can also use their cognitive abilities to form a true belief (i.e. they meet C) but the



 -

fact that they get things right may still be by luck (as happens in Gettier cases), which means that their cognitive success is not because of their cognitive abilities (i.e. they don’t meet C). In this way, precisely because C entails C and C but not the other way around, meeting C is (at least in principle) more difficult than meeting C or C. Sosa’s guardian angel case is an illustrative example of this mismatch in difficulty. When the competent archer shoots an arrow and an unexpected gust causes it to deviate but a guardian angel puts it back on track, (i) the shot is successful (C is met), (ii) the archer is using their abilities to shoot (C is met), but (iii) the shot does not manifest the archer’s exercise of ability: it manifests the intervention of the guardian angel instead (C is not met). In these circumstances, meeting C is very easy because success is likely independent of how competent the archer is: no matter how badly they shoot, the guardian angel will correct the trajectory of the arrow and make it hit the mark. What is difficult is precisely to hit the mark by manifesting competence in archery. The difficulty of acquiring knowledge is like this. In most situations, if we exercise our cognitive abilities to form beliefs (and thus meet C), the resulting beliefs are not only true (C is met), but their correctness also manifests our cognitive abilities and effort (C is met). Sometimes, however, these challenges come apart and so does the degree of difficulty in overcoming them. To see this more clearly, consider a case that is analogous to Sosa’s guardian angel situation but in the intellectual domain: [Temp] forms his beliefs about the temperature in his room by consulting a thermometer on the wall. Unbeknownst to Temp, however, the thermometer is broken and is fluctuating randomly within a given range. Nonetheless, Temp never forms a false belief about the temperature by consulting this thermometer since there is a person hidden in the room, next to the thermostat, whose job it is to ensure that whenever Temp consults the thermometer the temperature in the room corresponds to the reading on the thermometer. (Pritchard et al. : –)

In this case, Temp is presumably competent at reading and trusting reliable devices (C is met), he forms true beliefs (C is met), but his cognitive success does not manifest this competence; instead, it manifests the intervention of the hidden agent (C is not met). As in Sosa's case, meeting C is very easy in Temp’s circumstances, because he is very likely to form true beliefs about the temperature in the room (no matter how reliable or unreliable he is). What is difficult is overcoming C: Temp is unlikely to meet this challenge because the intervention of the hidden agent systematically thwarts the manifestation of his cognitive abilities.

Difficulty and Knowledge



Why does the intervention of the agent thwart the manifestation of Temp’s abilities? Traditional virtue reliabilism answers this question in terms of the appropriateness requirement (i.e. the idea that an agent’s cognitive success can manifest their cognitive abilities only in appropriate conditions). The traditional, unsatisfactory answer is simply that Temp does not manifest his cognitive abilities because the intervention of the agent makes the conditions inappropriate. As we have seen, however, the appropriateness requirement is incorrect (and should therefore be dropped) for three related reasons. First, it implies that manifestation of cognitive ability and hence knowledge is something that cannot happen in inappropriate conditions, but this is not true: it is when the conditions are inappropriate that our cognitive abilities may be most manifested (this is the case with intellectual feats, such as decisive breakthroughs in science). Second, and relatedly, it offers an incorrect representation of the relationship between difficulty and knowledge: it makes knowledge impossible (rather than unlikely) in inappropriate conditions. Third, it is unclear how learning can be explained if we assume the appropriateness requirement. Therefore, we need an alternative explanation of what thwarts manifestation of ability. A possible one goes as follows. Two factors prevent an agent’s cognitive success from manifesting their cognitive abilities. The first is luck. The second is masking circumstances. Let’s start with the first. Recall that there are two kinds of inappropriate conditions for the exercise of cognitive ability. The first type makes cognitive success unlikely (e.g. poor light conditions, psychedelic drugs and so on). In such conditions, you may use your cognitive abilities (e.g. your visual faculties) and you may even succeed cognitively by forming true beliefs against the odds – i.e., challenges C and C can be met in them. However, those conditions are such that one typically gets thing right because of luck, not because of the exercise of one’s abilities – i.e., challenge C cannot typically be met under them. Let’s consider the second factor: masking circumstances. Temp’s circumstances – just like the archer’s circumstances in Sosa’s guardian angel case or when shooting metal arrows at a giant magnetic target – are different: cognitive success is likely, to the point that it may be even guaranteed. This feature makes the circumstances masking, in the sense that no matter how unreliable one’s belief-forming methods are, one will end up forming true beliefs. In such circumstances, a cognitive success that is produced by the exercise of cognitive ability does not manifest such an exercise. In a sense, that exercise is masked by the circumstances. Take Temp’s case. Temp has the ability to trust reliable devices. Suppose that another agent, Tempo,



 -

lacks this ability and often trusts unreliable devices. In Temp’s circumstances, it does not matter whether one uses Temp’s or Tempo’s method: the intervention of the hidden agent guarantees cognitive success no matter what. In this way, the circumstances prevent Temp’s cognitive success from manifesting his normally reliable cognitive abilities by masking its exercise. In Sosa’s guardian angel case, or in the case of the competent archer shooting metallic arrows at a giant magnetic target, the circumstances are also masking: it does not matter whether the archer makes competent or deliberately incompetent shots. Any shot is successful no matter what. In conclusion, two relevant situational factors influence the difficulty of knowing: factors that make cognitive success unlikely and factors that, by making cognitive success likely, mask the manifestation of cognitive ability.

.

Concluding Remarks

In view of the foregoing discussion, we are now in a position to gloss the relationship between difficulty and knowledge as follows: how difficult it is for S to know that p at t in circumstances C corresponds to how likely S is to meet the challenge of forming a true belief in p at t in C in such a way that the fact that S comes to believe p truly at t manifests S’s cognitive abilities and effort up to t. In other words, the difficulty of knowing is simply the difficulty of attaining a cognitive success that counts as a cognitive achievement in the relevant sense (a true belief that manifests cognitive ability and effort). This completes the explanation of the idea that knowledge can be difficult. 

There is a possible objection lurking around that might have crossed the mind of the interested reader: what about fake barns? In the perennial fake barn case, a driver crosses an area full of indistinguishable fake barns and is lucky enough to look at the only real barn in the area. The objection is that, while the driver ignores that there is a barn in front of them (at least this is what many epistemologists think), it is seemingly very easy for the driver to manifest cognitive ability: all they need to do is to look at the barn and form the corresponding true belief that there is a barn in front of them. But why is that in the normal case in which there are no fakes around it is as easy to manifest visual competence and yet the driver would acquire knowledge as a result? In reply, I have elsewhere contended (Broncano-Berrocal , , forthcoming) that this diagnosis of the fake barn case (and similar cases) rests on the false assumption, often made by virtue reliabilists, that manifesting cognitive ability is just a matter of manifesting belief-forming ability. As I have extensively argued, we also engage a range of abilities that do not play a belief-forming but a precautionary role. Whereas belief-forming abilities are abilities to form true beliefs reliably, what I call precautionary cognitive abilities are those that prevent the activation of one’s belief-forming abilities in circumstances where it is unlikely that the latter will produce correct outputs (i.e. true beliefs). An example of a precautionary visual ability is the disposition not to form visual beliefs about the color of objects when the illumination is poor.

Difficulty and Knowledge



REF ERE NCE S Alfano, M. . Character as Moral Fiction. Cambridge, UK: Cambridge University Press. Baehr, J. S. . The Inquiring Mind: On Intellectual Virtues and Virtue Epistemology. Oxford: Oxford University Press. Baláš, J., Panáčková, M., Jandová, S., Martin, A. J., Strejcová, B., Vomáčko, L., and Draper, N. . “The Effect of Climbing Ability and Slope Inclination on Vertical Foot Loading Using a Novel Force Sensor Instrumentation System,” Journal of Human Kinetics : –. Bradford, G. . Achievement. Oxford: Oxford University Press. Broncano-Berrocal, F. . “A Robust Enough Virtue Epistemology,” Synthese (): –. . “Purifying Impure Virtue Epistemology,” Philosophical Studies (): –. forthcoming. “Epistemic Dependence and Cognitive Ability,” Synthese. Carter, J. A. . “Robust Virtue Epistemology as Anti-Luck Epistemology: A New Solution,” Pacific Philosophical Quarterly. doi:./papq.. The crux of the matter is that in the fake barn scenario the driver does manifest belief-forming ability (after all, they counts with a reliable visual system that automatically produces the true belief that there is a barn when they see it), but fails to manifest precautionary visual ability, in that they should have stopped belief formation. Why? Because the circumstances (viz., the fake barn area) are such that they would too easily have formed a false belief. In other words, that is something that the driver should have done because it is a real possibility in their circumstances (unlike, e.g., the possibility of being a brain in a vat). Of course, they do form a true belief against all odds, but this is precisely why they fail to manifest reliable precautionary ability. Once knowledge is understood in terms of the manifestation of all epistemic resources (i.e. not only of the agent’s belief-forming abilities and effort, but also their precautionary cognitive abilities), we can easily explain why the driver lacks knowledge: they only manifest part of the epistemic resources they ought to manifest. In this way, knowing in circumstances where there are visually indistinguishable fakes is extremely difficult because our visual system is wired in such a way that we automatically take them to be real. Consequently, manifesting precautionary cognitive ability (i.e. manifesting the disposition to stop visual belief formation when cognitive success is unlikely) is something extremely difficult to accomplish in such circumstances. In such cases, in order to meet challenge C – i.e., the challenge that we can now better understand as the challenge of succeeding cognitively because of all epistemic resources – we often need to resort to external aid to acquire the missing precautionary disposition. For example, in the fake barn situation the driver could use a fake barn detector: the information provided would reliably dispose them not to form beliefs about the presence of barns when in fact there are none. More mundanely, someone could tell the driver that they are in a fake barn area, thus endowing them with the disposition to suspend judgment altogether. However, this sort of crucial information is not typically available in such unusual circumstances and this is the reason why it is so difficult to acquire knowledge in them: one needs to be epistemically cautious and one does not even know this to be the case. The same considerations apply in general. In order to succeed and meet difficult challenges, sometimes the best we can do is to stop the current course of action, reassess and decide whether or not the current strategy is going to work. But this is not an easy task, especially if changing strategy is what we ought to do and we ignore it. In this sense, reliably applying caution constitutes an ability in itself, namely a precautionary ability – cf. von Kriegstein’s interesting related discussion on wasted effort (von Kriegstein ).



 -

Carter, J. A., Kallestrup, J., Palermos, S. O., and Pritchard, D. . “Varieties of Externalism,” Philosophical Issues : –. Chi, M. T. H., Feltovich, P. J., and Glaser, R. . “Categorization and Representation of Physics Problems by Experts and Novices,” Cognitive Science : – Chrisman, M. forthcoming. “Epistemic Normativity and Cognitive Agency,” Noûs. Greco, J. . “Knowledge and Success from Ability,” Philosophical Studies : –. . Achieving Knowledge. Cambridge, UK: Cambridge University Press. . “A (Different) Virtue Epistemology,” Philosophy and Phenomenological Research : –. Guerrero, A. . “Intellectual Difficulty and Moral Responsibility,” in J. W. Wieland and P. Robichaud (eds), Responsibility: The Epistemic Condition. Oxford: Oxford University Press. Jarvis, B. . “Knowledge, Cognitive Achievement, and Environmental Luck,” Pacific Philosophical Quarterly : –. Kelp, C. . “Knowledge: The Safe-Apt View,” Australasian Journal of Philosophy : –. Olin, L. and Doris, J. M. . “Vicious Minds,” Philosophical Studies : –. Pritchard, D. . “Anti-Luck Virtue Epistemology,” Journal of Philosophy : –. Pritchard, D. Haddock, A., and Millar, A. . The Nature and Value of Knowledge: Three Investigations. Oxford: Oxford University Press. Sosa, E. . A Virtue Epistemology. Oxford: Oxford University Press. . “How Competence Matters in Epistemology,” Philosophical Perspectives : –. . Judgment and Agency. Oxford: Oxford University Press. . “Virtue Theory Against Situationism,” in M. Alfano and A. Fairweather (eds), Epistemic Situationism. Oxford: Oxford University Press. Turri, J. . “Manifest Failure: The Gettier Problem Solved,” Philosophers’ Imprint : –. . “Epistemic Situationism and Cognitive Ability,” in M. Alfano and A. Fairweather (eds), Epistemic Situationism. Oxford: Oxford University Press. von Kriegstein, H. . “Effort and Achievement,” Utilitas : –. . “On Being Difficult: Towards an Account of the Nature of Difficulty,” Philosophical Studies (): –.

 

What Is Epistemic Entitlement? Reliable Competence, Reasons, Inference, Access Peter J. Graham

In contemporary epistemology, the term ‘entitlement’ has caught on, having been used by Robert Brandom, Tyler Burge, Fred Dretske, Christopher Peacocke, Michael Williams, and Crispin Wright, among many others, though often for very different purposes, as you might imagine. This chapter is about Tyler Burge’s distinction between entitlement and justification. His distinction falls within his overall ‘entitlement’ framework. His framework cuts across all of his writings in epistemology and informs, and is informed by, his philosophical and empirically informed investigations into the natures of our various cognitive capacities, including perception, interlocution (testimony), memory, reasoning, critical reason, self-knowledge, and reflection. Burge’s writings on these topics sprawl across a number of papers, many of them collected in his Cognition through Understanding (Oxford, ), with a number still to be collected, especially his papers on perception. Because I’ve written and talked about Burge’s views in epistemology, people often ask me about his distinction between entitlement and justification. I could refer them to a number of his articles, but they report finding Burge a challenging read and would like something a little easier going, which pulls everything together in one place. That’s what I’ll try to provide here. Our interest is not merely scholarly. Burge’s approach is a live contender. Expounding his approach will raise issues currently in contention, including the role of reasons, the nature of inference, and the internalism/ 

 

Boghossian ; Brandom , ; Burge , , a, a, ; Dretske ; Peacocke , ; Williams ; Wright a, b, . Very few of these authors have the same thing in mind when using the term ‘entitlement.’ Unnamed references throughout are to Burge’s works. For papers collected in Burge a, page numbers are those in the  printing. Other papers taking on this task are Casullo ; Majors ; and Gerken . See also Graham, Pedersen, Bachman, and Rosa .





 . 

externalism divide. Understanding Burge’s path through the territory should deepen our own grasp of these issues and more. To understand Burge’s distinction between entitlement and justification, you need to know that he systematically uses ‘entitlement’ in two different senses. In the first, we have an entitlement to rely on a belief-forming competence. This is an entitlement that attaches to our reliance on the competence in general. In the second, we have entitlements for beliefs – entitlements to hold or form particular beliefs through exercises of a competence. Here the entitlement attaches to the belief formed or sustained by the competence exercised on a given occasion. When Burge distinguishes between entitlement and justification, he primarily has this second sense in mind. You also need to know that sometime between  and  Burge changed his mind on how to draw the distinction. In ‘Content Preservation’ () and ‘Perceptual Entitlement’ (a) he relied on access to draw the distinction between entitlement and justification. He has since dropped accessibility as a distinguishing feature. If you think you know Burge’s current formulation from reading those two papers, think again. The chapter goes as follows. In Section . I explain what Burge means when he says we have a general entitlement to rely on a belief-forming competence. In Section . I explain the reliabilist element in his account of warrant. In Section . I explain his distinction. He draws the distinction in terms of reasons. And so, Section . includes a long discussion of his account of reasons. Readers interested in the role of reasons in epistemology should be especially interested in this material. In Section . I provide a diagnostic decision-tree and apply it to four representative cases to cement our mastery of the distinction. In Section . I explain why Burge’s distinction is not the inferential vs. non-inferential warrant distinction. I there offer a brief taxonomy of various kinds of psychological processing, including inference. Readers following recent debates about inference should find this section worthwhile. In Section . I explain why Burge changed his mind.

. Our Entitlement to Rely on a Competence This section addresses what Burge means when he says we have a general right or entitlement to rely on a belief-forming competence. 

Burge does not highlight the difference between his two uses, though he marks them on occasion. Sometimes he calls one a general entitlement and the other a particular entitlement. See Burge a: , , ; c: , n. ; e: –. b: ; . See also Majors , sec. .

What Is Epistemic Entitlement?



Burge argues that we enjoy entitlements to rely on interlocution, perception, memory, deduction, self-knowledge, reflection, and so on. When he does, he has both a negative and a positive point in mind. Burge’s negative point expresses agreement with the externalist movement in epistemology. According to the externalist, higher non-human animals, very young children, and even ordinary epistemologically unsophisticated adults have plenty of knowledge and warranted beliefs, especially perceptual knowledge and warrant. The externalist adopts a non-skeptical stance, assuming the correctness of these ordinary judgments. Given our ordinary concept of knowledge, we can know that higher non-human adults, young children and ordinary adults possess a great deal of knowledge. The externalist also observed that animals, small children, and some ordinary unreflective adults do not know – or even cannot know – that their belief-forming processes are reliable routes to external reality. Higher non-human animals, small children, and some unreflective adults lack meta-justifications in favor of the reliability of their belief-forming competencies. The externalist then concluded that knowledge and warrant – especially perceptual knowledge and warrant – cannot require the ability to represent one’s belief-forming competencies and abilities, or the ability to argue for their effectiveness. Knowledge and warrant in general cannot require metajustifications. This is then a negative point about what knowledge does not require. For many routes to knowledge, knowledge does not require metajustifications. Epistemology studies knowers. Knowers in general don’t have to be nascent epistemologists in order to be knowers. Burge claims we know this negative point a priori. We reflect on our ordinary concept of knowledge, on cases involving animals and young children, our practices of knowledge and knowledge attribution, and come to see that meta-justifications are not required for ordinary warrant and knowledge. Burge’s first use of ‘entitlement’ then signals agreement with the externalist movement. To say we have an a priori entitlement to rely on our belief-forming competencies is to say we don’t need a meta-justification establishing the effectiveness of our competencies for warranted reliance and knowledge. But to make this negative point is only to go so far. Many so-called moderate internalists or mentalists in epistemology can accept this negative point, but without embracing the externalist (especially the reliabilist) program tout court. Burge goes further.



 . 

Burge’s positive point is then that our general entitlement to rely on a competence “resides in” (b: ; b: ; b: ) or “rests on” (b: ) the reliability of the competence in normal conditions. To say that we have an entitlement to rely on a competence is then to say that the competence on which we rely is a reliable belief-forming competence. I shall say more about this momentarily. These points so far are now largely familiar in contemporary epistemology, as many philosophers have taken the externalist turn and see our warrant and knowledge as arising from exercises of reliable competencies instead of sophisticated meta-justifications. Burge’s papers in epistemology then take up our “entitlement to rely” on various belief-forming competencies. ‘Perceptual Entitlement’ (a) is about perception. ‘Content Preservation’ () concerns memory and testimony-based beliefs. ‘Comprehension and Interpretation’ (b) discusses linguistic understanding. ‘Our Entitlement to Self-Knowledge’ () is about self-knowledge. Other papers take up reasoning, reflection, intellection, and further explore the epistemology of memory. In each case, Burge takes up two tasks: (i)

(ii)

First, to argue a priori, by reflecting on cases and our practice, that we don’t need a meta-justification for warranted beliefs that arise from exercises of the competence. Which is just to say that we have an a priori prima facie entitlement to rely on the competence. We can formulate these entitlements to rely on the competence in the form of a principle. For example, we are a priori prima facie entitled to take the deliverances of perception at face value. We might call this the Perception Principle. Second, to give an account or rationale for why it should be that the competence is a good route to truth and knowledge. That is, explain why it is that the competence is reliable in normal conditions when functioning normally. Burge sometimes calls these accounts or rationales “justifications” for the principles just mentioned, for they explain the major part of the underlying entitlement to rely on the competence, viz. the reliability of the competence. They explain why reliance on the practice is “rationally justified” (: ; c: , n. ). This is Burge’s positive point as applied to the competence in question. The rationale can be a priori or empirical; the reliability of the competence can have an a priori or empirical basis. Which answer is correct depends on the details of the case. To say that our general entitlement to rely on a competence is a

What Is Epistemic Entitlement?



priori – his negative point that a meta-justification is not required for warrant – is then not ipso facto to say that the reliability of the competence wherein our entitlement resides is a priori. Again, it depends on the details of the case. Let’s look at three cases to make this more concrete: Perception. For (i), in ‘Perceptual Entitlement’ and elsewhere Burge repeats externalist considerations against requiring meta-justifications for perceptual warrant. The Perception Principle is true. For (ii), though Burge once argued that perception is necessarily reliable in normal conditions (, n. ; : ; b: ), he now holds that it is only empirically true that perception is reliable in normal conditions (a: –; c:  n. ). The underlying rationale for the Perception Principle is empirical: our entitlement to rely on perception has an empirical basis. Deduction. Here Burge means our competence with simple deductive inferential rules. For (i), in various places Burge argues we do not need to propositionally represent the inference rule or the reliability of the rule to have an inferential competence to reason in accord with the rule. Our entitlement “resides in [our] actual competence to make the relevant deductive transitions, not in an ability to understand and represent the rule governing the competence” (b: ). A Deduction Principle is thus true: we have a prima facie entitlement to transition from premises to conclusions through deductive reasoning. For (ii), Burge argues a priori that our competence with deductive reasoning is necessarily truth-preserving when functioning normally. The underlying rationale for a Deduction Principle is a priori: unlike perception, our entitlement to rely on deduction has an a priori basis; deductive reasoning is truth-preserving by its nature. Interlocution. For (i), Burge argues in ‘Content Preservation’ that we do not need to represent the communication channel and have independent grounds for believing it is reliable; we do not need reasons to believe that our interlocutors are sincere and competent to form prima facie warranted beliefs through interlocution (: –, c: –). To make his case, he focuses on children, language-learning, and our ordinary practice. Burge’s so-called Acceptance Principle is true: a person is entitled to accept as true something that is presented as true and that is intelligible to 

I have addressed issues surrounding the Acceptance Principle in my , , a, and b and elsewhere. For an excellent book-length treatment, see Shieber .



 . 

them, unless there are stronger reasons not to do so (: ). This makes Burge an anti-reductionist about interlocution. For (ii), Burge surprisingly argues that interlocution is a priori necessarily reliable in normal conditions when functioning normally. The underlying rationale for the Acceptance Principle is a priori: unlike perception but like deduction, our entitlement to rely on interlocution has an a priori basis; unlike perception but like deduction, interlocution is necessarily reliable in normal conditions. We can now see what it would be to deny that we have an entitlement to rely on a belief-forming process: either deny that the process is a reliable competence or deny that the competence produces warranted beliefs without a meta-justification of its reliability. Without an entitlement to rely on a belief-forming competence, a belief formed on the process would have to gets its warrant from some other source, if warranted at all. Sellarsians deny the Perception Principle: for perceptual warrant we must possess a meta-justification supporting perceptual belief. The ‘inferential internalist’ denies the Deduction Principle: for warranted belief through inference, we need to be aware of our inference and its truthpreserving power. The ‘reductionist’ about testimony denies the Acceptance Principle: for warranted testimony-based belief, we need independent reasons to believe the speaker is sincere and competent. Burge, of course, disagrees with all three. This should provide enough context to understand Burge’s first, general sense of ‘entitlement.’ With this understanding under our belts, we can now turn to Burge’s second, particular sense of ‘entitlement.’ On this second use, entitlement is a species of warrant. So, to understand his second use of the term, we should first address what he means by ‘warrant.’

. Warrant Is a Good Route to Truth Given what I have said so far that our general entitlement to rely on a competence resides in the reliability of the competence, it should come as no surprise to learn that for Burge warrant for beliefs formed through the exercise of a belief-forming competence should arise from the reliability of the competence. This makes Burge a reliabilist about warrant, but not just any kind of reliabilist. Burge holds a teleological, competence-based, proper functionalist account of warrant. Burge anchors his account in a teleological, normative 

I critically examine Burge’s case for this in my b.

What Is Epistemic Entitlement?



framework. The representational function of belief is truth. The representational function of a belief-forming competence is to form true beliefs. A belief-forming competence that is reliable in normal conditions when functioning normally meets natural norms – standards for doing well in meeting or furthering functions – a priori associated with the competence’s representational function. A competence that is functioning normally and reliably serves true beliefs in normal conditions meets a standard for doing well in furthering its representational function. Warrant then consists in meeting such a norm, for warrant is a good route to truth and knowledge. Warrant arises through the exercise of a reliable belief-forming competence that is reliable in normal conditions when functioning normally; a competence that is reliable in normal conditions when functioning normally is a good route to truth and knowledge. Even so, warrant does not suffice for knowledge. There are warranted true beliefs that fall short of knowledge. Some warranted beliefs are even false. Burge is clearly not a simple reliabilist. Sheer matter of fact reliability is not sufficient for warrant. Warrant requires a reliable competence; no competence, no warrant. An accidentally reliable ‘clairvoyant’ power does not confer warrant. Nor is sheer matter of fact reliability necessary for warrant. Warrant does not automatically lapse outside of normal conditions. A massively deceived brain-in-a-vat might still enjoy warranted beliefs. Provided the competence is reliable in normal conditions, an exercise of a competence confers prima facie warrant when functioning normally in any conditions. Burge’s view of warrant then lies near ‘virtue reliabilist,’ ‘competence,’ and ‘proper functionalist’ views that now populate much of the literature. Burge (a) himself noted similarities with Sosa’s “virtue reliablism” and Plantinga’s “proper functionalism.” Here’s how Burge’s idea works for perception. Perception is a competence to accurately represent objects and their attributes in the physical environment. The competence was formed and individuated through past patterns of interaction between the perceptual system and its subject  



a, b, , . See Burge forthcoming for his fullest treatment of ‘clairvoyance’-type cases. See Burge a for his fullest treatment of ‘demon-world’ cases. For discussion of whether warrant should persist in demonworld cases, see Gerken  and Graham . For example, see Bergmann , Greco , Kelp , Plantinga , Pritchard , and Sosa , among many others. I have developed the view that epistemic warrant consists in the normal functioning of the belief-forming competence when the competence has forming true beliefs reliably as an etiological function. See Graham , a, b, .



 . 

matter. The patterns begin with objects and their attributes in the environment; continue through a medium such as light; impact sensory receptors that in turn trigger processing within the perceptual system; eventuate in perceptual representations that in turn guide behavior in the organism’s environment. These past patterns then partly explain the formation and nature of the perceptual system. These patterns specify what counts as normal functioning and normal conditions. Normal functioning is working or operating the way the system worked or operated when it was formed, when it came to be what it is. Formative functioning determines normal functioning. Normal conditions are then those kinds of conditions where the system, in working or operating normally, came to be what it is, where it acquired the capacity to represent environmental objects and attributes. Formative environments determine normal environments. Higher animals, especially humans, then form (mostly true) perceptual beliefs though conceptualizations of (mostly accurate) perceptual representations, a non-inferential transition that reformats the perceptual representation into a perceptual belief. Warrant for perceptual beliefs then arises from the normal exercise of our competence to form perceptual beliefs. Reliable in normal conditions when functioning normally, perception is a good route to truth. The formative patterns ground a non-accidental explanatory path to success. When perceptual belief formation manifests competence, it runs along this path. Imagine a flow chart that describes the steps in the normal functioning of the perceptual system and the transition to perceptual belief. Consider a scene of objects and properties distributed in the individual’s environment. Light bounces off the objects and enters the eye. The perceptual system transforms the information through various channels. The modular processing of the perceptual system produces a perceptual representation, a 



This point applies both to competencies acquired in virtue of species membership (e.g. the human visual system) and competencies acquired by the individual through individual learning. For more on this way of specifying normal conditions and normal functioning, see Millikan , Majors and Sawer , Graham a, Gerken . ‘Conceptualization’ is Burge’s term of art for the process whereby perceptual representations are reformatted into propositional form (turned into perceptual beliefs). Conceptualization is neither inferential nor meta-representational. It does not involve a higher-order act of applying concepts to percepts. Conceptualization reformats perceptual attributives into concepts (propositional attributives), reformatting the perceptual representation into propositional form. The subject matter thereby remains the same throughout: physical objects and their properties. A perceptual belief is then, for the most part, the perceptual representation reformatted. See Burge a: –, b, b, and .

What Is Epistemic Entitlement?



state of the whole individual. The system processes; the individual sees. If the flow chart describes a good, reliable route to forming true beliefs in normal conditions, then beliefs formed along this flow chart enjoy prima facie warrant. When a perceptual belief is occasioned along this path, the belief is formed on a good route to truth. All forms of warrant – not just perceptual warrant – involve reliability in normal conditions. For all cognitive competencies (interlocution, deductive and inductive reasoning, memory, introspection and self-knowledge, reflection, etc., not just perception), warrant runs on such paths, on good, reliable routes to truth. Different competence, different formative patterns; different formative patterns, different pathways; different pathways, different route to truth. That’s Burge’s account of warrant.

. The Reasons Criterion We can now proceed to Burge’s distinction between two kinds of warrant: entitlement and justification. Burge sees the distinction in both psychological and epistemological terms: there is, Burge thinks, an important difference between two kinds of warrant because there is an importance psychological difference in the ways we form and sustain beliefs. What then is the basis for the difference? Burge originally drew the distinction in terms of accessible reasons. Justifications were warrants involving accessible reasons for belief; entitlements were warrants that did not involve accessible reasons. But more recently he has changed his mind. We will see why in Section .. He now draws the distinction solely in terms of the presence or absence of operative reasons in the particular exercise of the belief-forming competence. Justifications arise from exercises of good, reliable routes to truth that involve reasons, accessible or not. Entitlements arise from exercises of good, reliable routes to truth that do not involve reasons.



‘Perceptual Entitlement’ (a) makes it clear perceptual warrant requires reliability. Here are passages from Burge’s  Dewey Lectures that make it clear that all forms of warrant require reliability: “A propositional state or occurrence is warranted, on a given occasion, if and only if it is the result, on that occasion, of the exercise of a representationally well-functioning propositional competence that provides a representationally reliable, epistemically good route to veridicality, allowing for the natural limitations of the competence with respect to its subject matter, and limitations of the information available to it” (a: –). “To explain any epistemic warrant, one must explain what makes a warranted state an objectively good route to truth, given an individual’s background competencies. Being a good route implies reliable veridicality in normal conditions” (a: ).

 • •

 . 

Justifications = warrants involving reasons Entitlements = warrants without reasons

Imagine a flow chart (as it were) describing the exercise(s) of the good route to truth – the particular exercise(s) of the competence(s) forming or sustaining the belief. Are there reasons involved? If yes, the warrant for the belief is a justification. If no, the warrant for the belief is an entitlement. Following Mikkel Gerken, let’s call this the Reasons Criterion. But what are reasons? We won’t grasp Burge’s distinction until we understand his conception of reasons. Unsurprisingly, Burge has a complicated conception. For the sake of completeness, and so as to have most of the details of Burge’s view of reasons in one place, I’ll list all the points in his conception that I know. Most important for our purposes, however, are the last two. They tell us exactly what counts as a reason when it comes to grounding his distinction. We’ll see right away that he takes the notion of reasons seriously. For Burge, unlike many other epistemologists, ‘reason’ is not merely a commonly used synonym for ‘ground,’ ‘basis,’ ‘evidence,’ or ‘justifier.’ ‘Reason’ has a narrow use in Burge’s hands. Here are the details: •

•







Reasons are defined by their role. Reasons are actual and potential answers to ‘why’ questions. Why did the professor go home? Because he needed to do his laundry. That was his reason. Reasons are actual or potential steps in explanations. That is their job, role, or function. Answers to ‘why’ questions are necessarily propositional. Only something propositional can complete a ‘because’ clause or answer a ‘why’ question. This is a matter of the logical grammar of reasons. (b: ) This is a familiar point in the reasons literature.

Here is Burge from the introduction to his  collection of papers, Cognition through Understanding. “A justification is a warrant that consists partly in the operation or possession of a reason. An individual is justified if and only if the reason is operative or relied upon in the individual’s psychology. An entitlement is a warrant whose force does not consist, even partly, in the individual’s using or having a reason” (b: –). Gerken . Gerken argues against Burge for a rival criterion he calls the Reason Criterion. On this conception, justifications are warrants that arise from exercises of the faculty of reason; entitlements are warrants that do not. I intend to discuss Gerken’s view elsewhere. Witness John Hyman: “[T]he basic function of a reason is to explain something – in other words, to make something intelligible or understood – and more particularly to explain why something is the case. This is why there are reasons why but no reasons how, what, which, when, where, or who . . . [T]he primary function of a reason is, as it always is, to explain why something is the case” (Hyman : –).

What Is Epistemic Entitlement?



• Hence, non-propositional states and entities cannot be reasons. This excludes perceptions (though representational, they are non-propositional) and sensations (for they are frequently not even representational, let alone propositional) (a: ; b, a, a). • This also rules out worldly states, objects and events. “A thunderbolt or a sea otter cannot be a reason . . . Nature grounds reasons. It provides entities and situations representations of which can form a reason . . . Nature is not a text and is not made up of propositions or reasons” (a: –). • What then is the ontology of reasons? Reasons are abstract propositional representational contents of psychological states, marked with a certain mode, such as belief, judgment, or intention (a: ; Audi : –). • What then is it to have a reason? Having a reason involves thinking the reason, having a psychological state with that propositional content and mode (b: –). Burge thereby respects the distinction between there being reasons, on the one hand, and having reasons, on the other. • Normative reasons are then a subset of explanatory reasons. Normative reasons explain why someone should believe or intend something. • Reasons to believe are theoretical reasons. Theoretical reasons explain why one should believe something. Theoretical reasons are normative reasons to form or sustain a belief (a: ). 







John Pollock () and Christopher Peacocke (), among others, recognize the nonpropositional character of perceptual representations, but even so call them reasons for perceptual belief. Pollock calls perceptual representations “half-doxastic” on the grounds that they are representations and committal like beliefs, but different in being non-propositionally structured. Pollock proposes to call perceptual representations “reasons” while acknowledging “this that is stretching our ordinary use of the term ‘reason’” (Pollock : ). Burge sees the stretch as a “grammatical” mistake. Non-propositional states are ipso facto not reasons. Note well that for Burge perceptual experiences (conscious perceptual representational states) are genuinely representational though not propositional (see especially Burge a, ). Propositional representations are a species of representational content; propositional contents do not exhaust the category of representational (intentional) contents. This makes Burge a ‘propositionalist’ (of sorts) about reasons, as opposed to a ‘statist’ or a ‘factualist.’ Other propositionalists include Schroeder ; Fantl and McGrath ; and Comesana and McGrath . For a recent review of the ontology of reasons and their role in epistemology, see Littlejohn . It is common to classify justifications (normative reasons) as a species of explanations (explanatory reasons). “[R]easons in general are explanations, and . . . a justification is a particular kind of explaining, namely, an explanation of why something is right or just . . . a justification is an explanation” (Hyman : –). Theoretical reasons are “potential steps in . . . explanations that, in effect, show why one should believe . . . something” (a: ). They “provide a kind of explanation of the credibility of what they are reasons for” (b: ).

 •

• •

•

•







 . 

Being a believer entails having some reasons. This is because the capacity to have beliefs requires engaging in some deductive reasoning, and deductive reasoning combines with premises to provide reasons for conclusions. So, a believer will have some beliefs that play this role, if they have any beliefs at all. Reasons are the basic units in reasoning (a: ); they “enjoin thinking . . . in accordance with them” (: ). Reasons are elements in reasoning, in belief formation, maintenance, and revision. Some animals can reason deductively and so have reasons. As some studies of animal cognition suggest, some animals can reason from the belief that the food is either in the red cup or in the blue cup, that the food is not in the red cup, and then conclude that the food is in the blue cup. This requires competence with logical constants, and so the capacity for deductive reasoning (a: , n ). The rationalizing explanation need not be second-order or metarepresentational. A “reason is commonly not a meta-representational consideration that makes reference to shoulds, beliefs, intentions, or representational contents. Reasons can be about object-level matters. They are reasons by virtue of their potential use in supporting and explaining the ‘why’ of conclusions” (a: , n. ). Hence the steps comprising the explanation can be in the individual’s psychology, even if the individual cannot give the explanation. Having potential steps in explanations in one’s psychology is one thing. Being able to give the explanation is another. Some non-human animals have reasons, but they cannot give explanations for the belief-worthiness of their beliefs (a: ; : , ).

“[H]aving beliefs requires being able to carry out [deductive] inferences, and the relevant inferences must sometimes support beliefs. Inferentially supported beliefs are justified, or are at least backed by beliefs [reasons] that are capable of justifying them” (a: –, n. ). See Burge a, a, and for qualifications, see Burge c. “I think that some non-human animals engage in deductive inference. They have perceptual beliefs, memories, and some simple logical constants. They think according to simple rules of inference. They believe no logical truths, lack a concept of logical validity, and lack a concept of truth. They make valid inferences without understanding what they are doing. Whether this is empirically so about some non-human animals, this level of understanding seems conceptually possible” (Burge b: , emphasis added). Here Burge disagrees with mainstream Sellarsians who seem to think that though ‘intelligent’ in various ways, animals lack deductive inferential abilities and so lack propositional attitudes. For empirical evidence suggesting that some animals have deductive abilities, see Call , . For discussion, see Rescorla  and Graham b. “The explanation need not be in meta-representational terms. It is not essentially about belief or truth. It is fundamentally at the same level as the belief: p because r, where p is the content of the belief and r is the reason” (b: , n. ).

What Is Epistemic Entitlement?



• Reasons differ from critical reasons. Critical reasons evaluate reasons. Critical reasons are about reasons; they represent reasons as such, as reasons. Critical reasoning requires meta-representational abilities and epistemic concepts. Animals and small children have reasons and can reason. They lack critical reasons and cannot critically reason. • There are two broad categories of theoretical reasons. First, they are (or involve) the contents of self-evident propositional attitudes (e.g. cogito thoughts or basic logical or mathematical thoughts), for the reason for the belief-worthiness of the attitude is contained within the attitude itself. • Second, theoretical reasons are warranted premises in deductive and inductive arguments that combine with a competence with the inference rule to explain the belief-worthiness of the conclusion for the individual. The warranted premise is not a reason in isolation but only in conjunction with a competence with a rule. That’s Burge’s account of reasons. The earlier points in the bullet list help lead up to and make sense of the last two points – those we need under our belt to understand Burge’s distinction, for they delimit the class of theoretical reasons: theoretical reasons are either understood contents that self-evidently warrant belief, or other warranted beliefs that combine with a competence with a deductive or inductive inference rule to provide a rationalizing explanation of the belief-worthiness of the conclusion for the individual (e: ). They are steps in the individual’s mind – steps at the object level – that enter into explanations of the belief-worthiness of what they are reasons for. Returning to the justification versus entitlement distinction, justifications are then warrants involving reasons (when a reason is operative in the good route to truth). Entitlements are warrants without reasons (when there is no operative reason in the good route to truth). If self-evident, or derived from deductive or inductive reasoning, the warrant is a justification. Otherwise it is an entitlement. That’s the Reasons Criterion. To classify the warrant for a belief, again imagine the flow chart describing the good route to truth. Ask if there are any operative reasons in the flow chart. If there are, the overall warrant for the belief is a justification. If there are none, then the overall warrant for the belief is an entitlement. Besides calling an overall warrant for a belief a justification when an operative reason is involved, Burge also calls operative reasons themselves justifications. It is perfectly fine to call operative reasons justifications, for



 . 

when we ask someone for the justification for their belief, we are often asking for the argument that supports their belief (their reasons). We are not asking whether the overall warrant is a justification. (That’s a question for the professional epistemologist.) But even so I find this dual use confusing. My terminological preference would be to reserve ‘justification’ and ‘entitlement’ for kinds of warrants that attach to beliefs, and to reserve ‘reasons’ and ‘non-reasons’ for elements in the exercise(s) of the competence(s) that contribute to the warrant for the belief. Reasons serve justifications. Everything else serves entitlements. We would then only use ‘justification’ for warrant that results when operative reasons are involved, and simply call operative reasons ‘reasons’ instead of also calling them ‘justifications,’ as Burge often does. That way we’re only using ‘justification’ for one thing: a kind of warrant. And when ordinary people ask one another for their ‘justifications’ for their beliefs, we’ll take that as ordinarily meaning to ask for their reasons. What about entitlements? Does Burge ever call the other elements in the exercise(s) of the competence(s) that are not operative reasons entitlements? No. Instead he calls them contributions to an entitlement. A perceptual representation is not an entitlement, but a contribution to an entitlement. Perceptual representations do not entitle perceptual beliefs; they contribute to entitlement. They are a part of a good route to truth; they are not a good route all on their own. They are not the epistemic grounds for perceptual belief; at best they are just one factor. Many overall warrants involve a mix of contributions to entitlements and justifications as elements. In such cases, the overall warrant is a justification, despite the complexity of the good route to truth (b: ; a: ; b: ; c: , n. ). When it comes to justifications as total warrants, there are always elements that are not operative reasons in the good route to truth. Contributions to entitlement are ubiquitous. I’d like to introduce another terminological innovation. Instead of using ‘entitlement’ in both a general sense (as in our entitlement to rely on a belief-forming competence in general) and in a particular sense (for the kind of warrant a particular belief enjoys, depending on the presence of reasons in the good route to truth), I propose using ‘epistemic right to rely’ or ‘epistemic license’ for the general sense, and reserving ‘entitlement’ for the particular sense. That way we are only using ‘entitlement’ for one thing: a kind of warrant. To sum up, in this and the previous section we have expounded two important ideas. The first is that warrant arises from the exercise of a reliable belief-forming competence (a good route to truth and knowledge),

What Is Epistemic Entitlement?



a competence that is reliable in normal conditions when functioning normally. The second is that there are two kinds of warrant: entitlements and justifications, where the distinction turns on reasons. These two ideas are orthogonal. You can conceive of warrant as Burge does, but without seeing much point to the role of reasons in epistemology. Hilary Kornblith (), I imagine, sees things this way. Or you can agree with Burge about the importance of reasons in epistemology – agree that it grounds an important distinction between two kinds of warrant (you can even use ‘justification’ and ‘entitlement’ to mark the difference), but reject a competence, virtue reliabilist account of warrant. Crispin Wright (a, ) and Fred Dretske (), for example, might fall in here. Of the two ideas, which one is more fundamental, as it were? In my estimation, the fundamental notion for Burge is reliable competence. Fundamentally, Burge is a teleological reliabilist, competence theorist about warrant. The difference between the two kinds of warrant turns on reasons; their commonality turns on competence. Burge’s view isn’t fundamentally an ‘entitlement’ theory (whatever that might mean, taken in the abstract) but fundamentally a teleological reliabilist, proper functionalist, competence theory of epistemic warrant. In a recent paper Kurt Sylvan and Ernest Sosa have also argued for the incorporation of (possessed normative) reasons into a competence account of knowledge. Like Burge, they argue that some beliefs are properly based on reasons: “[We] agree that one species of justified belief may require possessing sufficient epistemic reasons” (Sylvan and Sosa : ). Here they concur with Burge that a species of warrant (“justified belief”) turns on reasons. They also agree that some beliefs are not warranted by reasons. Hence, they seem to agree with Burge’s distinction, though they don’t use his nomenclature. They furthermore agree that reliable epistemic competence, at root, is doing the work, even when reasons are involved. “[E]pistemic reasons . . . ground one species of justified belief, but competences are the real foundation” (Sylvan and Sosa : ). Although they don’t mention Burge, they’ve got their fingers on the same set of issues.

. A Decision-Tree and the Scope of Entitlement In this section I will run through a decision-tree and four examples to crystalize our understanding. The fourth example will raise an interesting issue about the nature of inference that we will subsequently pursue.



 . 

Here’s the decision-tree. Start with a warranted belief. Then go through these steps: STEP ONE Is the warranted belief warranted through self-evidence? • If yes, the overall warrant is a justification. • If no, go to step two. STEP TWO Is there some other warranted propositional attitude that is a part of the warrant, a part of the good route to truth? • If no, the overall warrant is an entitlement. • If yes, go to step three. STEP THREE Does the other warranted attitude combine with an inferential competence with an inference rule to help provide a rationalizing explanation of the credibility of the conclusion for the individual? Does the other warranted attitude function as a reason for the individual? • If yes, the overall warrant is a justification. • If no, the overall warrant is an entitlement. Figure . captures the three steps in a flow chart:

Figure .

A decision-tree and the scope of entitlement

These three steps should cover every case.

What Is Epistemic Entitlement?



Let’s now run through four cases. The first three appear throughout the Burge corpus on warrant. The fourth appears in one paper in particular, ‘Epistemic Warrant: Humans and Computers’ (b). CASE ONE: Cogito Thoughts and Simple Self-Evident Thoughts An individual thinks the thought I exist or I am thinking, or the thought +=, and thereby believes the thought through understanding its content. These beliefs are self-evidently warranted. “[The cogito proposition], as thought on the occasion, is self-evident in the sense that it constitutes a reason for itself” (b: ). By STEP ONE, the warrant is a justification. Since no inference is involved, the warrant is a non-inferential justification.

CASE TWO: Perceptual Beliefs An individual forms a perceptual belief that is a sphere through the exercise of a reliable perceptual competence that involves conceptualizing the perceptual representation that sphere. Apply STEP ONE. Is the belief self-evident? No. Move to STEP TWO. Apply STEP TWO. Is there a warranted propositional attitude involved in the flow chart describing the exercise of the competence? No. The perceptual representation is not a warranted propositional attitude: perceptual representations are non-propositional. No other propositional attitude is involved in the formation of the perceptual belief. Since conceptualization (in Burge’s sense) is not inference, a fortiori the transition to perceptual belief does not involve competence with a deductive or inductive inference rule. By STEP TWO, the warrant is an entitlement. (Hence the title of Burge’s  paper, ‘Perceptual Entitlement.’)

CASE THREE: Deductive Inference Suppose an individual forms the warranted belief that the streets outside are dangerous from the warranted premise beliefs that it is raining freezing rain outside and that if there is freezing rain on the streets, then the streets are dangerous through an exercise of their deductive reasoning competence with modus ponens.

 

For critical discussion of Burge on this point, see Gerken , section .. “Even though it is immediate [non-inferential], the self-knowledge that is illustrated in the pure, self-verifying instances of cogito is warranted by justification . . . [S]elf-knowledge of this sort is warranted through a justification . . .[these thoughts] are . . . reasons for themselves” (b: , emphasis added).



 . 

Apply STEP ONE. Is the conclusion self-evident? No. Move to STEP TWO. Apply STEP TWO. Are there warranted beliefs in the warrant? Yes. The premises are other warranted beliefs. Move to STEP THREE. Apply STEP THREE. Do the warranted premise beliefs combine with a deductive or inductive inference rule to constitute a reason? Yes. The two warranted premises clearly combine with a deductive reasoning competence with modus ponens to provide a rationalizing explanation of the credibility of the conclusion for the individual. The other warranted attitudes function as reasons for the individual. Hence, by STEP THREE the warrant is a justification.

CASE FOUR: Associative Inference Burge presents the following example in ‘Epistemic Warrant: Humans and Computers’ (b: –). An intelligent, non-human animal forms the warranted perceptual belief there is danger over there from the warranted belief there is that kind of motion over there. Somehow the animal ‘knows’ that motion of that kind signals danger. But the animal does not make this transition via a connecting premise belief if there is motion of that kind at a location, then there is danger at that location. The individual does not represent the connection. Instead, the psychological transition is the result of evolution by natural selection “hammering in” the reliable connection into the animal’s psychology, producing a reliable truth-preserving psychological competence, so that the animal automatically infers from the premise belief there is motion of that kind at that location to the conclusion belief there is danger at that location. The competent transition from the perceptual belief to the conclusion is a truth-preserving associative inference. The transition does not involve a competence with a deductive or inductive inference rule. Nevertheless, the conclusion is warranted, for the belief is formed on a good, reliable route to truth. Apply STEP ONE. Is the conclusion self-evident? No. Move to STEP TWO.





Their competence with modus ponens is not a premise in their reasoning. In general, inference rules, on pain of regress, are not themselves premises in reasoning. The rules, though most adults can learn and reflect on them and consciously apply them, need not even be represented or representable within the individual’s psychology for the individual to have a competence to reason deductively. For some discussion of reasoning in accord with a rule in this kind of sense, see Gerken , and Quilty-Dunn and Mandelbaum . Since the warrant for the premises involve perceptual warrants (entitlements), and since the individual’s competence with modus ponens is not a reason (it is not a warranted propositional attitude), the warrant for the conclusion belief involves a mix of operative reasons and elements that are not operative reasons. As noted, this is not at all surprising.

What Is Epistemic Entitlement?



Apply STEP TWO. Is there a warranted belief involved in the warrant? Yes, the perceptual belief. Move to STEP THREE. Apply STEP THREE. Does the premise belief combine through deductive or inductive reasoning – with a competence with a deductive or inductive inference rule – to provide a rationalizing explanation for the individual of the belief-worthiness of the conclusion? No. The animal moves from ‘p’ to ‘q’ without an inference rule that would help explain the belief-worthiness of the conclusion for the individual. Hence the perceptual belief does not function as a reason for the animal. “It is not a rationalizing premise within the individual’s psychology” (b: ).

Why does Burge say this? He says the transition is “dumb and nonrationalizing” even though it is part of a good route to truth. “Only an explanation of why” the associative competence “is in place explains and rationalizes the belief-worthiness of the conclusion, given the premise” (b: ). But that explanation is not in the individual’s psychology, and so not a part of an explanation of the reasonability of the conclusion for the individual. The premise belief might be a reason for the suprapsychologist who knows why the associative competence is in place, but not for the animal itself. For the animal, the inference from the premise to the conclusion fails to constitute “some sort of explanation for the acceptability of the conclusion . . . the premise and the transition fail to “combine to go some way toward providing an (object-level) explanation for the inferrer of why the conclusion is credible” (b: ). The warranted perceptual belief is then not a reason for the conclusion for the animal. Hence, there is no reason in the warrant, in the animal’s good route to truth. The warrant for the conclusion is then not a justification but an entitlement, an inferential entitlement. Burge then conjectures that cases like this last one may be more common than we might have thought: It seems to me that more of our elementary induction may take this form than we philosophers are inclined to think. Patterns of inference are hammered into us by evolution. The principles that they instantiate and that explain them may yield little insight into why the conclusion is rationalized, made sense of, or explained by the premises. (b: ) 

“[In “Empirical Warrant: Humans and Computers”] I . . . discuss . . . cases [of inference] in which [the premises are not reasons]. I center on empirically warranted inferences. I assume that the inference tracks likely connections in nature. So it supports the truth of the conclusion. But the inferential transition does not follow a rule that provides any insight at all into the nature or existence of the connection. [So the premise is not a reason.] Still, the . . . premises and inferential transition entitle one to the conclusion” (b: , bold added, material in brackets added).



 . 

Burge says he believes that his account of these cases “provides insight into the status of blind, but competent, truth-tracking, associative inference” (b: ). It seems to me that these cases should not be restricted to patterns “hammered in” by evolution but should include learned associations as well. Learned associations connecting propositional attitudes would then also involve non-rationalizing transitions from warranted premises (‘input’ beliefs) to warranted conclusions (‘output’ beliefs), where the total warrant is not a justification but an entitlement. Given that associative inference may be more common than we’ve been led to believe after the cognitive revolution, the scope of entitlement may be much broader than we might otherwise have been inclined to judge. That’s the three-step decision-tree applied to four cases. If I’ve succeeded, you should now have a thorough understanding of Burge’s current conception of the distinction between entitlements and justifications.

. What Is Inference? A number of authors have assimilated Burge’s distinction to the distinction between inferential (mediated) and non-inferential (immediate) warrant: justifications are inferential warrants, and entitlements are non-inferential warrants. Here is Ram Neta: If a believer is warranted in believing that P on the basis of an inference to her conclusion that P, then, for Burge, this is sufficient for the believer’s warrant for P to be a justification . . . [I]nferential warrants . . . are all justifications. (Neta : )

And here is Crispin Wright: Burgean justification is restricted to . . . reasons . . . Burge’s preference is to restrict the idea of reasons for belief to cases where beliefs are formed/ sustained inferentially. The Burgean notion of entitlement is enlisted in an attempt to explain how non-inferential but empirically based belief can be in rational good standing. (Wright : )

Given what we’ve seen already, this assimilation is mistaken. Our four cases show that the inferential vs. non-inferential distinction is orthogonal to the justification vs. entitlement distinction. There are inferential and non-inferential justifications, and there are inferential and non-inferential entitlements. 

If Mercier and Sperber are right, cases like this should be ubiquitous in human cognition generally, not to mention non-human, animal cognition. See Mercier and Sperber : , , , .

What Is Epistemic Entitlement?



The updated flow chart in Figure . makes this clear:

Figure . A decision-tree and inference

Burge’s distinction isn’t about the presence or absence of inference; it’s about the presence or absence of reasons. But what, you may be wondering, is Burge’s account of inference? Some philosophers (psychologists too) have a very broad conception, where any processing or transitioning in an individual’s psychology involves inference. On this very broad conception, the processing of visual stimuli into various stages by the perceptual system counts as inferential processing. At the other extreme, some other philosophers have a very narrow conception, where inference involves the conscious, deliberate drawing of a conclusion from a set of premises according to an inference rule, where the individual is aware of, or can easily become aware of, both the inference rule and his or her ‘taking’ the premises to support the conclusion according to the rule. 

A paper by Boghossian () sparked renewed interest in understanding inference. Boghossian leans toward the narrow view. (I conjecture that he does so partly because he holds an accessinternalist, responsibilist view of warrant. So, for inference to warrant a conclusion, one needs to ‘have access’ to what one is doing when making an inference, viz. ‘taking’ the premise as a reason for the conclusion.) For persuasive criticisms of Boghossian, suggesting a broader view, but perhaps not as broad as Burge’s, see Quilty-Dunn and Mandelbaum . See also McHugh and Way . For a general critical discussion of arguments from philosophers for the narrow view, see Kornblith .



 . 

Burge rejects the narrow view (b: ). But that does not mean that Burge accepts the broad view either: not every transition ‘in the mind’ is an inferential transition; perceptual processing is not inferential processing. So where, in between these two extremes, does Burge place inference? At a minimum, Burge insists, inferential transitions must be from a propositional mental state (such as a belief ) to another propositional mental state (typically another belief ). “Inference, as I use the term, is processing that connects propositional contents” (b: –, n. , emphasis in the original). Furthermore, an inferential transition must be along a warranting pattern: “any transition from one or more propositional attitudes to another, according to some warranting transition pattern, is propositional inference” (b: ). Inference is not so broad so as to include any mental processing. Nor is it so narrow so as to include only self-conscious, reflective reasoning. It even goes beyond deductive and inductive reasoning so as to include associate inference. It may even include non-conscious, modular inference, as we’ll see further along. This, of course, is not a full-blown theory of inference (far from it). Even so, given Burge’s remarks, as a conjecture we might propose the following taxonomy of psychological processes, where the narrower forms are all instances of the broader forms, but not vice versa: • • • •





processing from any cause to any (kind of ) mental state as output; processing from any mental cause to any mental output; processing from any mental cause to a propositional attitude as output (conceptualization is processing a perceptual representation into a perceptual belief); thinking: processing from propositional attitudes (thoughts) as input to propositional attitudes (further thoughts) as output; “Being non-inferential does not mean not involving processes . . . Similarly, immediacy does not entail lack of processing. A representational state is immediate if it is not the product of an inference” (b: –, n. ). As noted already, Burge rejects the view that processing by the perceptual system involves inference. The inputs and outputs of the perceptual system are not propositional. Perceptual processing does not involve deductive, inductive, or abductive inference, though it can be modeled as if it does. For more discussion of perceptual processing, see Hatfield ; Burge b, ; Hohwy ; Rescorla , . Prior to his  book and his papers circa , Burge often used ‘inference’ closer to his current use of ‘reasoning’: transitioning in thought according to a deductive or inductive rule that functions to transfer warrant from the premises to the conclusion. For instance, see Burge c: –.

What Is Epistemic Entitlement?



• inference: processing from propositional attitudes as input to propositional attitudes as output according to a warrant-conferring or preserving competence or pattern; • reasoning: processing from propositional attitudes as input to propositional attitudes as output according to a competence with deductive or inductive rules. The premise inputs function as reasons for the conclusion output; • critical reasoning: conscious processing from propositional attitudes as input to propositional attitudes as output according to a competence with deductive or inductive rules, with conscious awareness of the premises as reasons for the conclusion; • reflective critical reasoning: conscious processing from propositional attitudes as input to propositional attitudes as output according to a competence with deductive or inductive rules, with conscious awareness of the premises as reasons for the conclusion, and conscious awareness of the inference rule as such. The broadest view of inference would identify it with any cause of any mental state. The narrowest view would identify it with reflective critical reasoning. Burge places inference in between thinking and reasoning. What then is inference? Inference is warranting thinking. Reasoning, on the other hand, is justifying thinking. That’s why not every inferential warrant is a justification.

. Why Did Burge Change His Mind? If that’s Burge’s current conception of the distinction between entitlement and justification, then what was his former conception, and why did he change his mind? Here is how Burge originally posed the distinction in ‘Content Preservation’: Justifications . . . involve reasons that people have and have access to. These may include self-sufficient premises or more discursive justifications. But they must be available in the cognitive repertoire of the subject . . . Entitlements are . . . warrants that need not be . . . accessible to the subject. (Burge : , emphases added)

And here is what he said in ‘Perceptual Entitlement’: Justification is warrant by reason that is conceptually accessible on reflection to the warranted individual . . . Entitlement . . . contrasts with the internalist form of warrant – justification. (Burge a: –, emphasis added; cf. Burge b: , ).

Let’s call this the Accessible Reasons Criterion.



 . 

Three points stand out: ()

()

()

As with the Reasons Criterion, this is a reasons account of justification. Justification arises from operative reasons in the exercise of the competence. However, it adds an access requirement, unlike the Reasons Criterion. The reason must be accessible to serve justifications. Justifications arise from accessible reasons. The Reasons Criterion says nothing about access. The Accessible Reasons Criterion does not explicitly say that entitlements arise from elements in a good route to truth that are not operative reasons. Instead it says that contributions to entitlements need not be accessible to the individual. Inaccessible reasons (if they are possible) could then serve entitlements. The Accessible Reasons Criterion thus conflicts with the Reasons Criterion when it comes to entitlements. The Accessible Reasons Criterion also suggests that one might understand Burge’s distinction by assimilating it to an access version of the internalism vs. externalism distinction. Justification is just access-internalist warrant, and entitlement is just access-externalist warrant. Burge’s ‘innovation’ would then be to give each side in the controversy their own notion. This way of glossing Burge’s distinction is familiar in the literature. Burge frequently glossed his distinction this way himself. But this gloss, we will see, is importantly misleading.

Why did Burge drop the Accessible Reasons Criterion? We will find our answer in his reflections on the possibility of modular reasons and modular reasoning. ‘Modular’ means occurring in a psychological module, a subsystem of the human mind that is generally fast and automatic, whose operations are unconscious and opaque to the central, non-modular systems of the whole individual. Many modular subsystems output to the whole individual. A good example is perception. Perceptual processing of perceptual information is sub-personal, modular processing. It is fast, automatic, unconscious, and opaque to the individual. The output of the system is a perceptual representation, a state of the whole individual. Processing language may be another example. The output is a state of the individual, a representation of



Fodor ; Burge c: , a; Lyons ; Robbins . For a different view of modularity, see Sperber ; Carruthers ; Mercier and Sperber .

What Is Epistemic Entitlement?



the syntax and semantics of the utterance. But the inputs and intermediate stages are sub-personal (Burge c: –; e). Modular reasoning may or may not be possible. But one candidate – that Burge may or may not accept – is pragmatic processing to recover speaker meaning (Sperber and Wilson ). Regardless, Burge certainly considers the possibility of modular reasons. By , if not before, he imagined the possibility of sup-personal propositional attitudes that are unconscious and opaque to the whole individual but are nonetheless potential steps in explanations of the belief-worthiness of conclusions for the individual. He imagined the possibility of modular reasons. Are modular reasons accessible to the individual? If ‘accessible’ means that which ‘can be represented by a higher-order, reflective, introspective ability’ then modular reasons are clearly not accessible. Call this the narrow, reflective sense of access. Here is another sense. A psychological state is accessible if and only if it is a possible input to further non-modular psychological processing – inferential or otherwise – by the whole individual. In this sense, the perceptual beliefs of animals and small children are accessible, for they are inputs to reasoning and inference by the whole individual. Call this the moderate, whole individual sense of access. Are all modular reasons (reasons internal to the module) accessible to the individual in this sense? Obviously not. Here is still another, even broader, sense of access. A psychological state is accessible if and only if it is possible input to further processing of any kind – modular or non-modular – inferential or otherwise – by one or more of the psychological systems or capacities of the individual, subpersonal or personal. Put crudely, a state is accessible iff it is ‘in’ any part of the individual’s psychology that can be the input to any other part. Call this the rather broad, ‘in the mind’ sense of access. Are modular reasons accessible in this sense? Obviously yes. We then have three senses of access: • narrow – accessible to the individual through conscious, whole individual introspective reflection; moderate – accessible to the individual through conscious, whole • individual thinking; • broad – accessible to any part of the mind through any psychological operation. 

Incidentally, this is very close to Pollock’s sense of access. A state for Pollock is internal iff accessible, and accessible iff the potential direct input to belief formation. See Pollock and Cruz : .



 . 

If Burge were to keep the access criterion on reasons serving justifications, he’d then have three choices: . Only reasons that are narrowly accessible serve justifications. All other warranting elements serve entitlements. Young children and animals without introspective reflective access lack justifications. Entitlement, not justification, is then the type of warrant for the warranted premises in deductive arguments in the minds of animals and very young children. Animals and very young children have reasons but not justifications. Animals and small children are inside the space of reasons but outside the space of justifications. . Reasons that are moderately accessible serve justifications. Self-evident beliefs and the non-modular warranted premises in deductive arguments in the minds of animals and small children are then warranted by justifications. Their non-modular reasons serve justifications. Their modular reasons (if they have any) do not. Their modular reasons serve entitlements. . Reasons that are very broadly accessible serve justifications. The nonmodular and modular warranted premises in deductive arguments in subsystems and in the central processing of animals and very young children can then serve justifications. Which sense of ‘access’ does Burge adopt? Burge finds the very broad sense of access possible but not to his liking. The narrow sense is of course possible, as evidenced by its popularity. Even so, Burge favors the moderate view. Accessible for Burge then means within the central system, ‘accessible’ to the whole individual. Accessible then means consciously thinkable. Burge is then forced to make a choice when drawing the entitlement vs. justification distinction. Should he keep the access criterion understood in terms of moderate access? This would have the advantage of including animals and small children in the space of justifications, consistent with his intent. But it would have the disadvantage of excluding operative modular reasons from serving justifications, even though structurally they seem to be no different from reasons: they play the same role as potential steps in explanations. 

Burge reports (, n. ) that when he wrote ‘Perceptual Entitlement’ he believed that reasons were necessarily consciously thinkable, and so accessible; inaccessible reasons were not possible. So extensionally speaking, given that assumption, the Reasons Criterion and the Accessible Reasons Criterion would have been equivalent. But given the possibility of modular reasons, they are not equivalent.

What Is Epistemic Entitlement?



Which way should Burge go? Accept the moderate sense of access and keep the access requirement on reasons serving justifications, or accept the moderate sense of access but drop the access requirement altogether? Burge chooses the latter (b: ). Justifications involve operative reasons, regardless of access. Reasons are potential steps in rationalizing explanations of the belief-worthiness of conclusions for an individual, whether conscious or unconscious, accessible or inaccessible. Why should it matter whether they are accessible to reflection, in the central parts of cognition, or tucked away in modular subsystems? A quote sums this up: I . . . think that a functional-structural conception [of justification] is more basic than a conception in terms of accessibility to consciousness. As I now use the terms, justification consists in warrants by reasons; an individual has a justification only if a relevant reason is present in an individual’s psychology; entitlements are warrants that do not consist, even partly, in reasons . . . If there are modular reasons, they are justifications in the individual’s psychology, although not . . . accessible. (b: –)

That’s why Burge changed his mind. I hope by now you have mastered Burge’s two uses of ‘entitlement,’ the structure of his approach in epistemology, and his distinction between two forms of warrant in terms of the presence or absence of operative reasons. Why make all of these distinctions? Why use these words – ‘warrant’ ‘reason,’ ‘justification,’ and ‘entitlement’ – in these particular ways? The fundamental reason, I believe, is to make sure we track differences in different epistemic kinds, especially the ways those different kinds track fundamental differences in psychological kinds. Epistemology should respect our best philosophy of mind, just as our best philosophy of mind should respect our best epistemology. 



You may be wondering to what extent is Burge an ‘internalist’ given his new formulation. There are often ‘internal’ mental states as elements in good routes to truth, whether the eventual warrant is an entitlement or a justification, after all. Should we then call both species ‘internalist’ forms of warrant? Hardly. For both kinds are kinds of warrant, and warrant requires reliability, something any so-called internalist in epistemology is sure to reject. Though not a simple reliabilist about warrant, as we’ve seen, Burge is a teleological, competence reliabilist about both kinds of warrant for all that. Burge is not seeking a rapprochement between internalism and externalism; he’s declaring victory for one side. See especially Burge , §I, and Graham and Pedersen b. For a discussion of whether some forms of warrant are (roughly) internalist, see Majors : §, and Gerken forthcoming. I have presented earlier versions of some of this material to audiences at Yonsei University in South Korea, Radboud University in the Netherlands, and National Cheng Chi University in Taiwan. I am grateful for helpful comments on those occasions. For comments and conversations that led to substantial improvements, I am grateful to Zachary Bachman, Mikkel Gerken, John Greco, and Eric Wiland.



 .  R E F E R EN C E S

Audi, Robert. . The Architecture of Reason. Oxford: Oxford University Press. Bergmann, Michael. . Justification without Awareness. Oxford: Oxford University Press. Boghossian, Paul. . “Epistemic Analyticity: A Defense,” Grazer Philosophische Studien : –. . “What is Inference?” Philosophical Studies : –. Brandom, Robert. . Making It Explicit. Cambridge, MA: Harvard University Press. . Articulating Reasons. Cambridge, MA: Harvard University Press. Burge, Tyler. . “Content Preservation,” The Philosophical Review : –. . “Our Entitlement to Self-Knowledge,” Proceedings of the Aristotelian Society –. Reprinted in Burge a. . “Interlocution, Perception, and Memory,” Philosophical Studies : –. Reprinted in Burge a. . “Reason and the First Person” in Crispin Wright, Barry C. Smith, and Cynthia Macdonald (eds), Knowing Our Own Minds: Essays on Self-Knowledge. Oxford: Oxford University Press. Reprinted in Burge a. a. “A Century of Deflation and a Moment of Self-Knowledge,” Proceedings and Addresses of the American Philosophical Association. Reprinted in Burge a. b. “Comprehension and Interpretation,” in L. Hahn (ed.), The Philosophy of Donald Davidson. Chicago, IL: Open Court. Reprinted in Burge a. a. “Perceptual Entitlement,” Philosophy and Phenomenological Research : –. b. “Logic and Analyticity,” Grazer Philosophische Studien : –. c. “Perception,” International Journal of Psychoanalysis : –. . “Disjunctivism and Perceptual Psychology,” Philosophical Topics : –. a. Origins of Objectivity. Oxford: Oxford University Press. b. “Origins of Perception,” Disputatio : –. c. “Steps Toward Origins of Propositional Thought,” Disputatio : –. a. “Self and Self-Understanding: The Dewey Lectures ,” Journal of Philosophy : –. Reprinted in Burge a. b. “Epistemic Warrant: Humans and Computers,” first published in Burge a. c. “Postscript: ‘Content Preservation’,” first published in Burge a. d. “Disjunctivism Again,” Philosophical Explorations : –. e. “A Warrant for Belief in Other Minds,” first published in Burge a. a. Cognition through Understanding: Self-Knowledge, Interlocution, Reasoning, Reflection: Philosophical Essays, volume . Oxford: Oxford University Press. b. “Introduction,” in Burge a.

What Is Epistemic Entitlement?



. “Entitlement: The Basis of Empirical Warrant,” in Graham and Pedersen a: –. Call, Josep. . “Inferences about the Location of Food in Great Apes,” Journal of Comparative Psychology : –. . “Descartes’s Two Errors: Reason and Reflection in Great Apes,” in S. Hurley and M. Nudds (eds), Rational Animals? Oxford: Oxford University Press. Carruthers, Peter. . The Architecture of the Mind. Oxford: Oxford University Press. Casullo, Albert. . “What is Entitlement?” Acta Analytica : –. Comesana, Juan and Matthew McGrath. . “Having False Reasons” in Clayton Littlejohn and John Turri (eds), Epistemic Norms. Oxford: Oxford University Press. Dretske, Fred. . “Entitlement: Epistemic Rights without Epistemic Duties?” Philosophy and Phenomenological Research (): –. Fantl, Jeremy and Matthew McGrath. . Knowledge in an Uncertain World. Oxford: Oxford University Press. Fodor, Jerry. . The Modularity of Mind. Cambridge, MA: Cambridge: The MIT Press. Gerken, Mikkel. . Epistemic Reasoning and the Mental. London: Palgrave Macmillan. . “The New Evil Demon and the Devil in the Details,” in Veli Mitova (ed.), The Factive Turn in Epistemology. Cambridge, UK: Cambridge University Press. . “Epistemic Entitlement: Its Scope and Limits,” in Graham and Pedersen . Graham, Peter J. . “Liberal Fundamentalism and Its Rivals,” in Jennifer Lackey and Ernest Sosa (eds), The Epistemology of Testimony. Oxford: Oxford University Press. . “Testimonial Entitlement and the Function of Comprehension,” in Alan Millar, Adrian Haddock, and Duncan Pritchard (eds), Social Epistemology. Oxford: Oxford University Press. a. “Epistemic Entitlement,” Nous : –. b. “Psychological Capacity and Positive Epistemic Status,” in Jill Henderson (ed.), The New Intuitionism. New York: Continuum Press. . “Warrant, Functions, History,” in Abrol Fairweather and Owen Flanagan (eds), Naturalizing Epistemic Virtue. Cambridge, UK: Cambridge University Press. a. “Formulating Reductionism about Testimonial Warrant and the Challenge of Childhood Testimony,” Synthese : –. b. “Sincerity and the Reliability of Testimony: Burge on the A Priori Basis of Testimonial Warrant,” in Eliot Michaelson and Andreas Stokke (eds), Lying: Language, Knowledge and Ethics. Oxford: Oxford University Press. . “Why Should Warrant Persist in Demon Worlds?” in Graham and Pedersen a: –. Graham, Peter J. and Nikolaj J. L. L. Pederson (eds). a. Epistemic Entitlement. Oxford: Oxford University Press.



 . 

Graham, Peter J. and Nikolaj J. L. L. Pederson (eds). b. “Recent Work on Epistemic Entitlement,” American Philosophical Quarterly : –. Graham, Peter J., Nikolaj J. L. L. Pedersen, Zachary Bachman, Luis Rosa. . “Introduction and Overview: Two Entitlement Projects,” in Graham and Pedersen a: –. Greco, John. . Achieving Knowledge. Cambridge, UK: Cambridge University Press. Hatfield, Gary. . “Perception as Unconscious Inference,” in Dieter Heyer and Rainder Mausfeld (eds), Perception and the Physical World: Psychological and Philosophical Issues in Perception. New York: John Wiley: –. Hyman, John. . Action, Knowledge, & Will. Oxford: Oxford University Press. Kelp, Chris. . “In Defense of Virtue Epistemology,” Synthese : –. Kornblith, Hilary. . On Reflection. Oxford: Oxford University Press. . “The Role of Reasons in Epistemology,” Episteme : –. Littlejohn, Clayton . “Reasons and Theoretical Rationality,” in Daniel Star (ed.), The Oxford Handbook of Reasons and Rationality. Oxford: Oxford University Press. Lyons, Jack . Perception and Basic Beliefs: Zombies, Modules, and the Problem of the External World. Oxford: Oxford University Press. McHugh, Conor and Jonathan Way. . “Against the Taking Condition,” Philosophical Issues : –. Majors, Brad. . “What Entitlement Is,” Acta Analytica : –. Majors, Brad and Sarah Sawyer. . “The Epistemological Argument for Content Externalism,” Philosophical Perspectives : –. Mandelbaum, Eric. . “Seeing and Conceptualizing: Modularity and the Shallow Contents of Perception,” Philosophy and Phenomenological Research (): –. Mercier, Hugo and Dan Sperber. . The Enigma of Reason. Cambridge, MA: Harvard University Press. Millikan, Ruth. . Language, Thought, and Other Biological Categories. Cambridge, MA: The MIT Press. Neta, Ram. . “Can A Priori Entitlement be Preserved by Testimony?” in Alan Millar, Adrian Haddock, and Duncan Pritchard (eds.), Social Epistemology. Oxford: Oxford University Press. Peacocke, Christopher. . “Entitlement, Self-Knowledge and Conceptual Redeployment,” Proceedings of the Aristotelian Society : –. . “Explaining Perceptual Entitlement,” in Richard Schantz (ed.), The Externalist Challenge. Berlin: Walter de Gruyter. Plantinga, Alvin. . Warrant and Proper Function. Oxford: Oxford University Press. Pollock, John. . Contemporary Theories of Knowledge. Lanham, MD: Rowman & Littlefield. Pollock, John and Cruz, Joseph. . Contemporary Theories of Knowledge, second edition. Lanham, MD: Rowman & Littlefield.

What Is Epistemic Entitlement?



Pritchard, Duncan. . “Anti-Luck Virtue Epistemology,” Journal of Philosophy : –. Quilty-Dunn, Jake and Eric Mandelbaum. . “Inferential Transitions,” Australasian Journal of Philosophy : –. Rescorla, Michael. . “Chrysippus’ Dog as a Case Study in Non-Linguistic Cognition,” in R. Lurz (ed.), The Philosophy of Animal Minds. Cambridge, UK: Cambridge University Press. . “Bayesian Perceptual Psychology,” in Mohan Matthen (ed.), The Oxford Handbook of Philosophy of Perception. Oxford: Oxford University Press. . “Bayesian Sensormotor Psychology,” Mind & Language : –. Robbins, Philip. . “Modularity of Mind” The Stanford Encyclopedia of Philosophy, Edward N. Zalta (ed.). https://plato.stanford.edu/archives/win/ entries/modularity-mind/. Accessed January , . Schroeder, Mark. . “Having Reasons,” Philosophical Studies : –. Shieber, Joseph. . Testimony: A Philosophical Introduction. New York: Routledge. Sosa, Ernest. . Judgment and Agency. Oxford: Oxford University Press. Sperber, Dan. . “In Defense of Massive Modularity,” in E. Dupoux (ed.), Language, Brain, and Cognitive Development. Cambridge, MA: The MIT Press. Sperber, Dan and Diedre Wilson. . “Pragmatics, Modularity, and Mindreading,” Mind & Language : –. Sylvan, Kurt and Ernest Sosa. . “The Place of Reasons in Epistemology,” in Daniel Star (ed.), The Oxford Handbook of Reasons and Normativity. Oxford: Oxford University Press. Williams, Michael. . “Responsibility and Reliability,” Philosophical Papers : –. Wright, Crispin. a. “On Epistemic Entitlement: Warrant for Nothing, and Foundations for Free?” Aristotelian Society supp. : –. b. “Intuition, Entitlement and the Epistemology of Logical Laws,” Dialectica : –. . “On Epistemic Entitlement (II): Welfare State Epistemology,” in Dylan Dodd and Elia Zardini (eds), Scepticism and Perceptual Justification. Oxford: Oxford University Press.

 

Knowledge-Producing Abilities John Greco

What kind of abilities are apt for producing knowledge, and for giving knowledge its value? Greco (, ) and Sosa (, ) understand the notion of cognitive ability in such a way that true belief attributable to ability is also safe. In effect, the ability condition on knowledge entails a safety condition on knowledge as well. However, Greco argues that abilities must be relativized to environments, whereas Sosa seems to deny this. Pritchard’s (, ) “modest” virtue epistemology denies that ability entails safety, but adds an additional safety condition on knowledge. Finally, Turri considers two alternatives to standard robust virtue epistemology, one that requires more from knowledge producing abilities and one that requires less. “Ampilism,” endorsed by Turri (), requires that knowledge-producing abilities manifest safety as well as truth. “Abilism,” endorsed by Turri (a, b, c), denies that knowledge requires safety. This chapter considers the dispute in light of questions regarding the value of knowledge: In what sense is knowledge valuable? And in what sense of “ability” does the ability condition on knowledge contribute to that value? The discussion proceeds as follows. Section . discusses the view that Pritchard dubs “Robust Virtue Epistemology”: that knowledge is true belief attributable to cognitive ability, and that the ability condition on knowledge entails a safety condition as well. This view, as defended in Greco (, ) and Sosa (, ), has several theoretical advantages, including the elegance with which it accounts for both the nature and value of knowledge. Section . rehearses several proposed counterexamples to the view. In particular, Kallestrup and Pritchard argue that there are cases in which true belief is attributable to ability but not safe. Such examples motivate Pritchard’s own “modest virtue epistemology,” which continues 

Sosa has changed his view in this respect over several years. This chapter considers view in his  and  books. For a discussion of the evolution of Sosa’s view on this point, see Greco .



Knowledge-Producing Abilities



to hold that knowledge is true belief attributable to cognitive ability, but which adds a safety condition to address the examples. Section . looks more closely at the nature of ability in general and of cognitive ability specifically. In particular, it looks at the sense of “ability” in play in robust virtue epistemology’s claim that knowledge is true belief attributable to ability. Notational differences between Greco and Sosa are discussed and resolved, and a detailed account of knowledge-producing ability (i.e. the kind of ability relevant for producing knowledge) is presented. That makes it possible to present a more detailed account of the conditions on knowledge, and to show that the ability condition on knowledge, so understood, entails a specific version of the safety condition. Section . uses the account in Section . to resolve the counterexamples presented in Section .. With the counterexamples neutralized, Section . moves to considerations regarding the value of knowledge and argues that robust virtue epistemology is the superior view. First, that view explains the final value of knowledge as a species of a more general kind – success that is attributable to competent agency, and that is thereby safe success. Second, the view explains the practical value of knowledge partly in terms of the practical value of knowers. Specifically, knowers are persons who can be depended on to get things right, in actual circumstances and in relevantly similar circumstances. Finally, Section . considers Ampilism and Abilism in light of related considerations. Versions of Ampilism are distinguished, and it is argued that the most promising version reduces to the standard robust view. Against Abilism, it is argued that the value of knowledge, both final and practical, partly resides in being reliably successful, as opposed to merely actually successful. Before proceeding, one clarification is worth noting up front. In the discussion below, it is conceded that there are various kinds of ability (or if you like, various senses of the term “ability”), and that various accounts of the ability condition that result do ground the claim that knowledge has a kind of final value and a kind of practical value. Nevertheless, it will be argued, our preferred view better explains the kinds of value that knowledge has. Accordingly, the central thesis of the chapter is this: the precise kind of ability that our preferred view employs to characterize the nature of knowledge best explains the kinds of value, both final and practical, that we think knowledge has.

. Robust Virtue Epistemology and Some of its Advantages Pritchard () characterizes “Robust Virtue Epistemology (RVE)” as the position that knowledge is identical to true belief from cognitive ability.



 

That is, S knows that p just in case S has a true belief that p (rather than a false belief or no belief at all) because S’s belief was produced (or is sustained) by cognitive ability. Versions of this view are defended in Greco (, ), where it is argued that the ability condition on knowledge (properly understood) entails a safety condition (properly understood). Sosa has long defended a virtue-theoretic approach to knowledge but has been inconsistent on the relation between virtue and safety. Nevertheless, in Sosa (, ) he holds that the ability condition on knowledge does entail a kind of safety, and so in those works he should be understood as embracing Robust Virtue Epistemology as Pritchard intends that label. Differences remain among the views defended by Greco and Sosa, however. For example, Greco argues that abilities must be relativized to environments, whereas Sosa seems to deny this. Nevertheless, a shared view emerges that has several theoretical advantages. Most notably, RVE plausibly diagnoses a range of Gettier cases, allows an elegant treatment of the value of knowledge, and yields a detailed account of epistemic normativity. Finally, and more on this below, RVE motivates a specific version of the anti-luck condition on knowledge in terms of ability-relative safety: In cases of knowledge, S would not easily go wrong when S believes from ability (of the right sort). Despite these theoretical virtues, a number of authors have raised objections to the robust view. In Section . we focus on a series of counterexamples.

. Counterexamples to Robust Virtue Epistemology, and Pritchard’s Modest View ..

Barn Facades

The first (alleged) counterexample that we will consider is the famous Ginet-Goldman Gettier case involving barn facades. According to Pritchard, RVE yields the result that there is knowledge in the case, which is counterintuitive. Barn Facade. Henry is driving in the countryside and sees a barn ahead in clear view. On this basis, he believes that the object ahead is a barn. Unknown to Henry, however, the area is dotted with barn facades that are indistinguishable from real barns from the road. However, Henry happens to be looking at the one real barn in the area.   

See Greco  for the evolution of Sosa’s view.  See especially Greco ,  and Sosa . See Greco , . The example is from Goldman . Goldman attributes the example to Carl Ginet.

Knowledge-Producing Abilities



Recently, some philosophers have suggested that Henry does know in the case, and there is some empirical evidence that the folk agree. If one is unconvinced that Henry does not know, alter the case so that Henry drives by a series of barn facades, incorrectly judging in each case that there is a barn. After ten such misidentifications, Henry sees a real barn, and then proceeds to misidentify another series of ten barn facades. The idea that Henry knows in the one case where he forms a true belief seems incredible. ..

Force-Fields

The next case to consider is by Pritchard, offered in defense of his view that RVE must give the wrong verdict in Barn Facade. Archie. Archie . . . selects a target at random, skillfully fires at this target, and successfully hits it . . . Suppose, however, that unbeknownst to Archie there is a force-field around each of the other targets such that, had he aimed at one of these, he would have missed it. It is thus a matter of luck that he is successful, in the sense that he could very easily have not been successful. (: )

Pritchard then comments: “Notice, however, that luck of this sort does not seem to undermine the thesis that Archie’s success is a genuine achievement. Indeed, we would still ascribe an achievement to Archie in this case even despite the luck involved. It is, after all, because of his skill that he is successful, even though he could very easily have not been successful in this case” (: ). Pritchard’s line of reasoning can be reconstructed as follows: . . 



Archie is analogous to Barn Facade in all relevant respects. The case in Archie is a case of success from ability.

For philosophers who argue that there is knowledge in Barn Facade, see Turri a, b, c; Sosa , , . Sosa’s treatment of Barn Facade is somewhat complicated. But the main idea is that Henry’s true belief that there is a barn does manifest his perceptual ability, and so Henry does know in this case. Another aspect of Sosa’s view explains why this result seems counterintuitive. Namely, although S enjoys a kind of basic or animal knowledge in barn facade cases, she lacks a more robust and reflective kind of knowledge. See Sosa : –, –, esp. n. . See also Sosa : – and Sosa : –. Since I think that Sosa’s diagnosis is misguided here, below I will defend the view given in Greco , . Turri a, b presents empirical evidence that non-philosophers regularly attribute knowledge in some barn facade cases. Thanks to John Hawthorne for the modified case, in conversation. Turri  found that subjects tended to attribute knowledge even in iterated barn facade cases. In Turri’s case, “the agent encounters four fakes and verbally misidentifies each as an item of interest, after which she encounters and correctly identifies a genuine item of interest.”



 

Therefore, . The case in Barn Facade is a case of success from ability. (, ) . There is no knowledge in Barn Facade. Therefore, . The case in Barn Facade is a case of success from ability without knowledge. (, ) .. Epistemic Twin Earth Kallestrup and Pritchard () offer the following “epistemic twin earth” argument as an objection to RVE. Once again, the point of the objection is that RVE’s ability condition on knowledge does not entail a safety condition, and therefore RVE allows cases of knowledge without safety. To understand the argument, imagine a twin earth where there is plenty of HO in the global environment, but where there are regions containing only XYZ rather than HO. (As usual, XYZ is an odorless, colorless liquid that is perceptually indistinguishable from water.) Finally, imagine that twin-S (S’s intrinsic physical duplicate) is in a local environment that contains only HO. The environments on Earth and Twin Earth are therefore structured as in Figure .. Since S is a normal perceiver in a normal environment on earth, we can assume that S’s true perceptual beliefs regarding water amount to knowledge. Kallestrup and Pritchard then argue as follows. On epistemic twin earth all watery stuff in twin-S’s global environment is HO. Not only is twin-S therefore able to entertain water-thoughts, a high frequency of twin-S’s water-beliefs as formed in her global environment is Earth

Global: H2O

Epistemic Twin Earth Global: H2O

Regional: H2O

Regional: XYZ

Local: H2O

Local: H2O Twin-S

Figure .

The environments on Earth and Twin Earth

Knowledge-Producing Abilities



true both in actual fact and across relevantly close worlds. Twin-S’s perceptual apparatus is thus equally reliable. Moreover, all watery stuff in twinS’s local environment is HO. When twin-S forms the demonstrative belief that that’s water on the basis of a perceptual experience as of water, her belief is true. Yet, unbeknownst to twin-S, twin-water is abundant in her regional environment. The basis on which twin-S holds that belief is thus such that her belief is only luckily true, in that given the basis for her belief it could very easily have been the case that she would have formed a false belief. (:–)

Kallestrup and Pritchard sum up the problem for RVE this way: The explanation robust virtue epistemology offers of why S has knowledge on earth is that her cognitive success is because of her cognitive ability. The challenge, however, is to explain why twin-S lacks knowledge on epistemic twin earth. The fact that S and twin-S are intrinsic physical duplicates embedded in physically identical global environments means that one cannot possess a cognitive ability that the other lacks. (: )

.. Good and Bad Finally, Kallestrup and Pritchard () offer what might be thought of as a “social” barn facade case – that is, a Gettier case that trades on environmental luck, but in this case social environmental luck. First, consider a competent epistemic agent H, who is embedded in an epistemic community GOOD. In GOOD, most testifiers are reliable. Moreover, most conversations are monitored and policed so as to ensure reliable testimonial exchanges. Kallestrup and Pritchard () assume that H can acquire knowledge in GOOD by accepting a reliable speaker’s testimony. Now consider a second case. Now imagine that H is unwittingly transported to epistemic community BAD which also mostly contains reliable testifiers. The difference is that while the testimonial processes in GOOD are monitored and policed in a knowledge-enabling manner, the corresponding processes in BAD are monitored and policed in a knowledge-precluding manner vis-à-vis H. That is to say, third party epistemic agents reliably ensure that H is mostly exposed to unreliable speakers . . . Now assume that H forms a true belief on the basis of hearing reliable speaker S’s testimony. The proposition in question is again an ordinary one which nearly every reliable speaker will know. (: –)

Kallestrup and Pritchard argue that H lack’s knowledge in BAD and that robust virtue epistemology cannot account for this. In particular, they contend, H exercises the same ability in BAD that she does in GOOD.



 

The difference in knowledge in the two cases, then, cannot be explained in virtue-theoretic terms (Kallestrup and Pritchard : ).

. Ability, Knowledge, and Safety Greco (, ) defines abilities in term of appropriate “conditions,” whereas Sosa (, ) defines abilities in terms of appropriate “shape” and “situation.” These differences are merely notational, as Greco includes the same factors in “conditions” as Sosa divides between “shape” and “situation.” Furthermore, both agree that abilities have an “inner seat” in the subject, and that abilities must be defined in terms or some relevant success. The common position that emerges is this: S has an ability A(Se/Sh/Si/R/D) = S has a disposition seated in inner seat Se to produce some result in range R, when in appropriate shape Sh and appropriate situation Si, with appropriate degree of reliability D.

Greco (, ) explicitly relativizes abilities to environments, Sosa does not. Is this too a merely notational difference? Greco suggests that there is no principled distinction between conditions and environment: “we can think of ‘environments’ as sets of relatively stable circumstances and ‘conditions’ as sets of shifting circumstances within an environment” (: ). In that sense of “environment,” everything included in Greco’s “environment” can be understood to go into Sosa’s “situation.” However, consider a different sense of “environment” that is important to consider in the present context. Let a modal environment be defined as a sphere of logical space centered on a possible world. Modal environments are related to geographical environments in the following way: Whenever a subject S is actually located in a geographical environment, they are located in a modal environment that is partially determined by their geographical environment. For example, suppose that you are now located in New York City. Then your modal environment is partly determined by the fact that you are actually in New York City. All other things being equal, the possible worlds where you are in New York City are closer than the possible worlds where you are in New Jersey, and these latter are closer than the worlds in which you are in California. Now there is certainly a principled difference between a) the general conditions in term of which an ability is defined, and b) the fully determinate conditions that constitute a modal environment. And with this distinction in mind, it should be clear that abilities, defined in terms of general conditions, must be relativized to modal environments, understood in terms of fully determinate conditions. The intersection of the

Knowledge-Producing Abilities



former and the latter creates the set of worlds that must be evaluated to determine whether S has an ability relative to an environment. Again, think of a modal environment E as a space of possible worlds centered on the actual world (or a typical world, or a normal world, etc.). S has an ability, relative to E, just in case S is reliably successful, when in appropriate conditions (appropriate Sh and Si), throughout that modal space. The idea is that a modal environment always presents more detailed actual and potential conditions than any set of conditions characterizing an ability. For that reason, whether S is reliable when in a defined set of appropriate conditions always depends on how conditions are further characterized in the environment. Put differently, the fully determinate conditions constituting a modal environment will outrun the details characterizing appropriate conditions for any defined ability. The more general those latter conditions, the more variability in reliability relative to different modal environments. Hence, it makes perfect sense to say, and in fact we have to say, that the same disposition (even when robustly defined as including appropriate Sh and Si) counts as an ability or no (i.e. as reliable or no) only relative to a modal environment. And this is a different point than saying that abilities must be defined in terms of appropriate Sh and Si. All this is consistent with saying that, often or always, the conversational context in which we ascribe abilities will specify some relevant modal environment. Likewise, many abilities are naturally relativized by default to some kind of normal and/or typical modal environment. For instance, baseball abilities and football abilities are that way. But other abilities will have no natural default environment – the question will always be whether S has such an ability here, or there, or somewhere else. For example: the ability to find food; the ability to stay warm; the ability to stay alive; the ability to discriminate friend from foe; the ability to discriminate barns from non-barns. Moreover, both default environments and environments specified by a context, as with other implied parameters, can always be cancelled and replaced. Suppose that Sofia is going away to a new school and her grandmother is worried about her eating well. I reply that Sofia is perfectly capable of feeding herself. In doing so, I intend that Sofia is capable of finding good food, buying it, preparing it to eat, etc., when in relevant shape (e.g. she is ambulatory and otherwise healthy) and in relevant situation (e.g. the weather is normal – there is no blizzard or extreme flooding). And relative to normal environments, she certainly has that ability – she reliably feeds herself just fine. But she does not have the ability to feed herself in a desert,



 

where there is no food to be had! NB: this is so even when holding shape and situation within normal ranges. That is, even ambulatory and in good weather, etc., Sofia would lack the ability to reliably feed herself in a desert environment. The example illustrates that abilities have to be understood as relative to (modal) environments. Holding specified ranges of situation and shape fixed, there will still be different success rates relative to different (modal) environments, depending on further features of the environments in question. A further theoretical decision point that presents itself is whether abilities are to be defined in terms of what is normal, what is typical, what is salient for practical purposes, or something else. For example, when specifying the relevant seat Se, shape Sh, situation Si, range of truths R, and degree of reliability D for visual perception, should those parameters be defined in terms of what is normal for human beings, what is typical for human beings, what is relevant for present practical purposes, or something else? A similar question arises regarding the specification of a relevant modal environment, for when we ask whether S is reliable throughout the modal space defined by some E, how far out is that space supposed to extend? Is that parameter to be understood in terms of what is normal or typical, what is relevant for present practical purposes, etc.? Greco (, ) and elsewhere argues that knowledge-producing abilities should be specified in terms of practical considerations. The idea is that a conversational context picks out a practical environment, which in turn is defined by a set of actual and potential information-dependent practical tasks. The informational needs associated with the specified practical environment determine all relevant parameters. There is no knockdown argument for choosing this theoretical option, however. Rather, the correct choice regarding how to specify relevant parameters will be a function of a) getting the extension of the concept right, but also b) other constraints on theorizing, such as the point and purpose of knowledge attributions, intuitive relations between knowledge and action, intuitive relations between knowledge and assertion, relations among these and practical reasoning, and considerations about the value of knowledge. For present purposes, however, the choice we make here does not matter. That is because, for a great many cases, what is normal, typical, practically 

There is an important distinction here between what is “normal” in a developmental sense, for example in the evolution of visual perception, and what is “typical” for human beings (or some set of human beings) now. Clearly, these can come apart. In this regard, it is not always clear what a particular theorist has in mind when he or she refers to “appropriate,” “relevant,” “normal,” or “typical” conditions. See Graham  for an exception.

Knowledge-Producing Abilities



relevant, etc. will overlap. This is sufficiently so that, in the arguments that follow, we can ignore theoretical differences on this particular point. With these more general considerations in place, we are now in a position to present an account of knowledge-producing abilities – one that makes all parameters explicit and that relativizes knowledge-producing abilities to a modal environment. We are also in a position to make more explicit the conditions on knowledge that follow. And with that in place, we are able to see how the ability condition on knowledge, so understood, entails a specific version of a safety condition. First, here is the account of knowledge-producing abilities, i.e. the kind of ability that is relevant for producing knowledge: S has a knowledge-producing ability A(Se/Sh/Si/R/D) relative to a modal environment E = S has a disposition seated in inner seat Se to believe truths in an appropriate range R of propositions p . . . pn when in appropriate Sh and appropriate Si, and in E, with appropriate degree of reliability D.

How do we understand “appropriate” here? In other words, how do we understand the relevant parameters Sh, Si, and D? Again, that is a matter of theory choice. Some refer to what is normal, some to what is typical, some to practical context, and plausibly there are other options as well. Likewise, for the characterization of E. But in whatever way we fill in those details, we get the following conditions for knowledge. S knows that p iff for some appropriate A(Se/Sh/Si/R/D) and E: . . . . . .



p is in R S is in E S has Se S is in Sh and Si S’s belief that p is produced by an exercise of A(Se/Sh/Si/R/D) in E (S’s belief is competent) S has a true belief that p because S’s belief is produced by an exercise of A(Se/Sh/Si/R/D) in E. (S’s belief is true because competent, S’s belief is apt).

There is an alternative notation available here, where we define a non-relative knowledge-relevant ability by making the specification of E internal to the characterization of ability. Hence: S has a knowledge-producing ability A(Se/Sh/Si/R/D/E) = S has a disposition seated in inner seat Se to believe truths in an appropriate range R of propositions p . . . pn when in appropriate Sh, Si and E, with appropriate degree of reliability D. But this would seem to be a mere notational difference, with corresponding alternative notations available for our characterizations of knowledge and safety below.



 

Alternatively, S knows that p iff for some appropriate A(Se/Sh/Si/R/D) and E: S’s belief that p is true because it is produced by an exercise of A(Se/Sh/Si/R/D) in E.

We many now see that these conditions for knowledge entail the following: For all p in R, not easily would S believe p (while retaining Se and in Sh, Si) when p is false.

Alternatively, and more exactly: For all p in R, throughout the space of worlds defined by E, almost always when S believes that p (while retaining Se and in Sh, Si), p is true. In fact, that is so to degree D.

And therefore, these conditions for knowledge entail the following safety condition: S’s belief that p is SSS-safe: Not easily would S believe p (while retaining Se and in Sh, Si) when p is false.

Alternatively, and more exactly, Throughout the space of worlds defined by E, almost always when S believes that p (while retaining Se and in Sh, Si), p is true.

. Resolution of Counterexamples This section uses the accounts presented in Section . to address the counterexamples directed at robust virtue epistemology in Section .. ..

Barn Facades

Recall Barn Facade. Henry is driving though countryside populated with barn facades, but sees a rare actual barn in clear view. Pritchard’s diagnosis is that Henry’s true perceptual belief that there is a barn is from ability, and indeed constitutes an achievement, but is nevertheless unsafe. After all, Henry could easily have a false belief that there is a barn. S forms his belief in Sh and Si that are actually appropriate. But that is not enough for SSS-safety. In response, we should agree with Pritchard that Henry’s belief is not safe. More specifically, his belief is not SSS-safe: Henry could easily be in appropriate Se, Sh, and Si, and yet believe falsely that there is a barn.

Knowledge-Producing Abilities



However, Henry lacks the perceptual ability to discriminate barns from non-barns relative to his environment. Therefore, Henry does not believe from ability, and so his true belief is not attributable to ability. This follows straightforwardly from the present account of ability, which requires that Henry be appropriately reliable throughout his modal environment, when in appropriate Si and Sh. Alternatively, the account requires that, relative to the modal environment that Henry is in, he could not easily be wrong regarding p, when in Si and Sh. These conditions are plainly not satisfied in Barn Facade. At least that seems right on plausible ways of specifying A(Se/Sh/Si/R/D) and E. That is, the problem in Barn Facade is not that Henry is in odd lighting conditions (he is in inappropriate situation), or that he has ingested a hallucinogen (he is in inappropriate shape), etc. Rather, it is that there are deceptive objects in Henry’s environment. Relative to Henry’s environment, he is unreliable at identifying barns, and in that sense lacks the kind of ability required for knowledge. Why does Sosa think that Henry knows in this case? Put differently, why does Sosa think that Henry’s true belief is attributable to ability? Presumably because he is defining ability in terms of appropriate Se, Si, and Sh, and in Barn Facade Henry is in appropriate Se, Si, and Sh. In that respect, Sosa is in agreement with Pritchard. But what Sosa misses is that one can believe from appropriate Se, Si, and Sh, and yet still not believe from ability. Believing from ability requires believing reliably (or safely), and that condition is not satisfied relative to Henry’s modal environment. ..

Archie

Archie selects a target at random, skillfully fires at this target and successfully hits it because of his skill. However, unbeknownst to Archie there is a force-field around each of the other targets. According to Pritchard, “luck of this sort does not seem to undermine the thesis that Archie’s success is a genuine achievement . . . It is, after all, because of his skill that he is successful” (: ). 

Greco  suggests a more nuanced view which is consistent with the above. Namely, we can imagine practical contexts where Henry’s modal environment is defined narrowly enough that it includes only real barns and no barn facades. For example, we can imagine that the practical context is restricted to the one working farm in the area. In that case, it is suggested, it is plausible that Henry does know, and precisely because he does have the relevant ability, relative to the modal environment specified by the practical context. By contrast, in Barn Facade we naturally conceive the case so as to include barn facades in the relevant modal environment.



 

In response, we should agree that Archie’s success is indeed a genuine achievement. In our terminology, it is a success attributable to an appropriate archery ability A(Se/Sh/Si/R/D). Our account allows this because we evaluate for ability only in cases where Archie is in appropriate Si. That is, to determine whether Archie has an ability relative to a modal environment E, we evaluate only worlds within E where Archie is in appropriate Si. But plausibly, archery abilities are not specified so as to require reliability at hitting targets protected by force-fields. (Note that this is so in any plausible way of specifying archery abilities. That is, the point holds whether we define archery abilities in terms of what is typical, normal, practical, etc.) Accordingly, Prichard is right that “luck of this sort does not seem to undermine the thesis that Archie’s success is a genuine achievement,” and in fact the present account explains why. Now compare Archie with a different next case. Modified Archie. Imagine a sport called Archery*, where the goal of the contest is to first identify a “good” target (i.e. a target not protected by a force-field) and then shoot it. Suppose Archie lacks the ability to find good targets but picks out a good target by luck and shoots it. Or suppose he is properly skilled but running out of time. He randomly picks out a target, which lucky for him is a good one, and shoots it.

In Modified Archie, the nearby worlds where Archie shoots at force-field protected targets are absolutely relevant for evaluating whether Archie has the relevant Archery* ability, whether his success is SSS-safe, etc. (Note that this is so on any plausible way of specifying Archery* abilities.) .. Epistemic Twin Earth Let us agree that S knows. Whether Twin-S satisfies our conditions for knowledge centrally depends on how we understand the relevant modal environment E. I propose that we can consider different alternatives, and in each case our account yields a plausible result. First, suppose that the relevant modal environment is understood to include Regional (see Figure .). Presumably, this is how Kallestrup and Pritchard () intend the example, and in that case Twin-S lacks an ability to discriminate water from non-water relative to the relevant modal environment. Therefore, according to our account, Twin-S does not believe from ability and so Twin-S’s true belief is not attributable to ability. This is in juxtaposition with S, who does have an ability to discriminate water from non-water relative to their modal environment,

Knowledge-Producing Abilities Earth

Global: H2O

Epistemic Twin Earth

Global: H2O

Regional: H2O

Regional: XYZ

Local: H2O

Local: H2O Twin-S

Figure .



Epistemic Twin Earth

who does believe from ability, and whose true belief is attributable to ability. So, our account rules the right way for both S and Twin-S. But suppose we define Twin-S’s modal environment so as to include Local and exclude Regional. Presumably, we can imagine further details that would recommend understanding the example this way. For example, we can imagine that Twin-S is effectively prohibited from entering Regional, and so not easily would they find themselves there. In that case, Twin-S’s true demonstrative belief that that’s water is attributable to an ability relative to E, their belief is SSS-safe relative to E, and our account rules that they have knowledge. But when the case is understood so that Regional is excluded from S’s modal environment (Regional is not modally close), I contend that we should lose the intuition that Twin-S does not know. For consider, on this understanding of the case, Twin-S’s belief is safe. At the very least, Kallestrup and Pritchard () lose their diagnosis regarding why Twin-S does not know. ..

GOOD and BAD

Whether S has knowledge in GOOD and BAD will depend partly on which cognitive abilities are required for testimonial knowledge, and this, of course, is controversial. But whichever abilities are relevant for producing testimonial knowledge, the clear difference between GOOD and BAD is the modal environment described in the two cases. And so, keeping S’s Se, Sh, and Si fixed, however these are plausible defined, there will be a clear difference regarding how S performs throughout the modal space defined by the two cases. And this is sufficient to explain why S knows in GOOD and does not know in BAD, independently of how we think of the cognitive abilities relevant for producing testimonial knowledge.



 

. The Value of Knowledge In Section ., we saw that our preferred version RVE can neutralize a range of proposed counterexamples. Even better, it can plausibly diagnose what is going on in the examples, and thereby plausibly explain the difference between knowledge and non-knowledge in a range of cases. In this section, we turn to considerations regarding the value of knowledge in order to adjudicate between robust virtue epistemology and Pritchard’s “modest” alternative. To get the two views more clearly in mind, we begin by looking at Pritchard’s own account of knowledge-producing abilities, i.e. the kind of abilities involved in the production of knowledge. Kallestrup and Pritchard write that Abilities are relative only to the stable environment in which whoever has them is typically located. That is to say, abilities are possessed as long as they are reliably manifested in . . . the normal circumstances in which abilities are acquired through learning and sustained through practice. Temporary abnormal environments cannot rob a subject of an ability that she otherwise reliably manifests in the normal run of things. (: , emphasis added)

This suggests that, on Pritchard’s view, a cognitive disposition counts as a knowledge-producing ability just in case it is reliable in “typical” and/or “normal” circumstances. Using the terminology employed above, we have: S has a knowledge-relevant ability A(Se/R/D) = S has a disposition seated in inner seat Se to believe truths in range R of propositions p . . . pn, when in normal shape and situation and in normal environment with degree of reliability D.

This account of abilities, now wedded to Pritchard’s characterization of “modest” virtue epistemology, yields the following account of knowledge: S knows that p iff for some appropriate A(Se/R/D): . . . .

p is in R S has Se S’s belief is produced by an exercise of A(Se/R/D) S has a true belief that p because S’s belief is produced by an exercise of A(Se/R/D) . S’s belief is Se-safe.



If we follow Pritchard’s own comments on the safety condition, we should gloss condition  as follows: In almost all the close possible worlds (and in all the nearest worlds) where S's belief that p is produced by an exercise of Se, p is true. For example, see Pritchard . For present purposes, this further gloss on the safety condition is irrelevant, however.

Knowledge-Producing Abilities



Alternatively, S knows that p iff for some appropriate A(Se/R/D): . .

S’s belief that p is true because it is produced by an exercise of A(Se/R/ D); and S’s belief is Se-safe.

We saw above that, according to RVE, the ability condition on knowledge entails a safety condition on knowledge. MVE characterizes the ability condition in a way that does not entail safety, however, and so has to add a safety condition in addition to the ability condition. Accordingly, we get the following alternative accounts of knowledge: RVE S knows that p iff for some appropriate A(Se/Sh/Si/R/D) and E: . S has a true belief that p because it is produced by an exercise of A(Se/ Sh/Si/R/D) in E; and thereby . S’s belief that p is SSS-safe. MVE S knows that p iff for some A(Se/R/D): . S has a true belief that p because S’s belief is produced by an exercise of A(Se/R/D); and in addition to this . S’s belief that p is Se-safe. This is not a merely notational difference. For one, the two views yield different extensions. To see this, consider how the two differ regarding the evaluation of their respective safety conditions. On RVE, one looks to all the close Bp-worlds that are also SSS-worlds. More exactly, one looks to all the close worlds where S’s belief that p is produced by an exercise of Se, and S is in the appropriate Sh and Si. If p is true in a sufficient number of those worlds, then S’s belief that p is SSS-safe. On MVE, one evaluates the safety condition by looking to all the close worlds where S’s belief that p is produced by an exercise of Se, independently of whether S is in appropriate Sh and Si. If p is true in a sufficient number of those worlds, then S’s belief that p is Se-safe. That is why Pritchard thinks that Archie’s hitting the archery target is an achievement (it satisfies the ability condition) but is not safe (it easily could have missed). It does not matter for safety, on Pritchard’s view, that force-fields constitute abnormal and therefore inappropriate conditions for the exercise of archery abilities. And of course, MVE is compelled to treat analogous epistemic cases in the same way. More on this follows.



 

A second difference between the two views is how they conceive the value of knowledge. More exactly, the two views differ regarding where the value of knowledge is seated. There would seem to be three options here, as follows. a.

All value comes from the ability condition broadly conceived, i.e. from the success-from-ability condition on the SSS-characterization of ability (and entailing SSS-safety). b. All value comes from the ability condition narrowly conceived, i.e. from the success-from-ability condition on the Se-characterization of ability. c. Value is divided between an independent narrow ability condition and an independent safety condition. Robust virtue epistemology opts for (a), modest virtue epistemology must opt for (b) or (c). With the two views more clearly in mind, we may now turn to adjudication. Regarding extension, MVE is too strong. Why should knowledge require success outside appropriate shape and situation? In other words, why should knowledge require Se-safety, as opposed to mere SSS-safety? For example, suppose that S uses their excellent reasoning powers to solve an equation that is well within the range of their capability. Suppose also that S could have easily been drunk, and that if they were drunk they would make mistakes. In other words, S could have easily been in bad shape, and if they were, they would make mistakes. In that case, S’s belief will be SSS-safe but not Se-safe. But here SSS-safety seems sufficient for knowledge. It is irrelevant that S’s belief is not Se-safe as well. Consider a second example. Suppose that in good lighting condition, S uses excellent visual perception to judge that there is a person standing in the corner of the room. Suppose also that the room could easily be dark (someone could have easily shut off the lights, for example), and that if the room were dark then S would make a mistake. In other words, S could have easily been in a bad situation, and if they were, they would be mistaken. In that case, S’s belief will be SSS-safe but not Se-safe. But again, SSS-safety seems sufficient for knowledge. It is irrelevant that S’s belief is not Se-safe as well. 

For present purposes, it will not matter how we understand “appropriate” here. That is, these considerations in favor of SSS-safety over Se-safety do not depend on whether we think of appropriateness in terms of what is typical, what is normal, what is practically relevant, etc.

Knowledge-Producing Abilities



Next, we may turn to considerations regarding the value of knowledge. First off, we consider the “final” value of knowledge, or why knowledge is valuable for its own sake. In this regard, it is important to note that both RVE and MVE can adopt the following explanation of knowledge’s final value. . .

In general, success from ability has final value. Knowledge is a kind of success from ability.

Therefore, .

Knowledge has final value.

Where the two views differ, we have seen, is in how they understand the notion of ability in premises  and . On the robust view, ability is understood broadly, in terms of SSS-ability. The view, then, is that the value of knowledge derives from the value of success from SSS-ability more generally. On the modest view, ability is understood narrowly, in terms of Se-ability. The view is that the value of knowledge derives from the value of success from Se-ability more generally. This difference in the two views is closely tied to a second. On the robust view, success from SSS-ability is thereby SSS-safe. That is, the ability condition entails a kind of safety condition, and that figures into the explanation of value. According to the view, then, one reason that success from ability has final value is that such success is SSS-safe, and in that sense not mere lucky success. On the modest view, however, success from ability does not entail safety, and hence its final value is unrelated to safety. Accordingly, we get the following positions on the value of knowledge: RVE on the final value of knowledge: Knowledge has final value because it is an instance of success attributable to SSS-ability, and therefore SSS-safe success. MVE on the final value of knowledge: Knowledge has final value (when it does) because it is an instance of success attributable to Se-ability and has that value independently of its also being Se-safe success. 



On Pritchard’s view, only “strong” achievements have final value, where strong achievements are those that either require great ability or that overcome significant obstacles. Pritchard : ff. This makes Pritchard’s account of the value of knowledge more complicated than what is presented here. See next note. Again, Pritchard’s own view is more complicated than this, since he thinks that only “strong” cognitive achievements have final value, and that not all knowledge qualifies as strong cognitive achievement. Accordingly, he does not think that all knowledge has final value. The discussion below abstracts away from this complication.



 

To adjudicate between the two views, then, we may now turn to considerations regarding the value of achievement more generally. Do we value “success-from-ability” conceived in the robust way or in the modest way? More specifically: Do we value success that is attributable to SSS-ability, and that is thereby SSS-safe success, or do we value success that is attributable Se-ability, and independently of whether it is Se-safe success? First off, we may concede that, at least in some cases and to some degree, we do value success that is from Se-ability but not SSS-ability. That is, we do sometimes value success that is from Se-ability but where S is not in appropriate situation and/or shape. But compare the value of that kind of success with the value of success from SSS-ability. For example, suppose that SAFE DRIVER is a capable driver, who drives home safely in good shape and in good conditions. Compare the value of their success with that of DRUNK DRIVER, who gets home without incident, but who could easily have crashed because they were drunk (i.e. in terrible shape). Or compare RECKLESS DRIVER, who also gets home without incident, but who could easily have crashed because they drove home in blizzard conditions (i.e. in terrible situation). Or consider a drunk driver in a blizzard, or a drunk archer in a hurricane (terrible shape and situation). Even if these characters perform successfully, we attribute this more to good luck than to ability, and for that reason we value their successful performance differently. Objection: In some cases, we value success in inappropriate shape or situation more than success in appropriate shape and situation. For example, we admire the performance of Babe Ruth, who (legend has it) sometimes played the game drunk, or badly hungover, and hit home runs anyway. But a plausible analysis of this kind of case is that we think Ruth had an ability to hit home runs even when drunk or hungover. That is, we attribute to Ruth an SSS-ability*, where the shape parameter included being in that sort of shape. Put differently, we think that Ruth’s home runs were better than SSS-safe (were Sh and Si are defined in terms of normal baseball parameters) – they were SSS*-safe (which includes the normal parameters but broadens them to include abnormal shape as well). Or that is how legend has it. Suppose legend is wrong. Suppose that Ruth had no such broader ability and so his home runs in bad shape were lucky. In that case, I contend, we should judge Ruth’s drunken success similar to that of the drunk driver – it is a good thing that he was successful, and perhaps there is some value in his successful performance, but it is not the same value as that of success from SSS-ability.

Knowledge-Producing Abilities



Considerations about final value, then, show that we value success from SSS-ability, and which is thereby SSS-safe, more than success from Se-ability in inappropriate shape and/or situation, and which is thereby not Se-safe. Considerations about the practical value of knowledge drive us to a similar conclusion. For consider what we want, practically speaking, from knowledge and knowers. In short, we want knowledge and knowers that we can depend on across relevantly close situations. That is, we want beliefs and sources of belief that are reliable across relevantly close situations. And which situations are those? Some theorists opt for typical situations, some for normal, and some for those that are relevant for practical purposes. But however one decides on this point, the sort of reliability that results will amount to an analogous version of SSS-safety. In effect, the robust view explains why knowledge is non-lucky, and in what sense.  Of course, the modest view also offers such an explanation, in terms of an independent safety condition. However, as we have seen, that anti-luck condition is too strong, and in any case grounds a less elegant explanation. All this suggests that we value knowledge, this time practically, because it is a kind of success from SSS-ability, and is thereby SSS-safe.

. Ampilism and Abilism We have seen that a guiding idea of virtue epistemology is that, in cases of knowledge, success is attributable to ability. Put differently: In cases of knowledge, success “manifests” competence. The central question of the chapter has been: What kinds of cognitive ability (or competence) are required for producing knowledge? Alternatively: How should we understand the notion of “cognitive ability” in the virtue-theoretic idea that knowledge is success attributable to cognitive ability? So far, we have been understanding the relevant notion of “success” in terms of true belief – both robust and modest virtue epistemology embrace the view that, in cases of knowledge, the truth of S’s belief manifests competence. John Turri has proposed that, in cases of knowledge, the safety of S’s belief manifests competence. A performance is ample just in case its safety manifests the agent’s competence. A performance is safe just in case it (i) succeeds and (ii) would not easily have failed. We then propose that knowledge is ample belief. (Turri : )  

 Greco  and Greco forthcoming. As in the preferred terminology of Sosa and Turri. Turri c considers the proposal, but does not endorse it.



 

Is this an alternative view to robust virtue epistemology? That depends on how one interprets Ampilism’s safety condition. If we interpret it as Sesafety, for example, then it is a different view but too strong. That is because, as we saw in Section ., Se-safety is too strong a requirement on knowledge. If we conceive Ampilism’s safety condition as SSS-safety, however, then the view does seem to reduce to the robust view. That is because, on the robust view, the ability condition on knowledge entails SSS-safety. On that view, S has a true belief that p because it is produced by an exercise of cognitive ability, and thereby S’s belief is SSS-safe. But this seems equivalent to saying that, in cases of knowledge, safety manifests competence. Ampilism holds that, in cases of knowledge, safety manifests competence. Abilism denies that ability grounds safety. Specifically, Turri (a, b, c) proposes that knowledge is true belief manifesting cognitive ability, but where “ability” is defined in a way that does not imply either safety or reliability. Rather, one has an ability (in the relevant sense) just in case one can produce a relevant outcome in a way that exceeds chance. I accept the following metaphysical thesis about powers in general: If a person possesses an ability/power to produce an outcome (of a certain type and in conditions of a certain sort), then when he exercises that ability/ power (in those conditions), he produces the relevant outcome at a rate exceeding chance. The basic intuition here is that abilities and powers are understood relative to a baseline of chance . . . You are enabled or empowered to produce an outcome to the extent that your prospect of producing it exceeds chance. (Turri b: )

In favor of this view of ability, Turri points to the existence and value of “lucky achievements,” or “lucky success-from-ability.” In this respect, he agrees with Pritchard above, who points to unsafe achievements in archery and elsewhere. In response, we may concede that a) there is a sense of “achievement” such that achievements can be unsafe, and b) achievements in that sense do have value of a sort. With Turri, for example, we can acknowledge achievement when a toddler takes a few steps, even if they cannot yet do so reliably, and we can agree that such an achievement is to be celebrated. We may also concede that there are two senses of “ability” – one associated with the “can” of ability, as when one says that Albert Belle can catch fly balls, without implying that he can do so reliably, and one associated with competence, as when one compares Ken Griffey Jr.’s ability to catch fly balls. Nevertheless, we should agree that “success from

Knowledge-Producing Abilities



ability” that is thereby safe and reliable has a different kind of value than does “success from ability” that is nevertheless unsafe and unreliable. That is why we have a different kind of admiration for Griffey than we do for Belle, and why we evaluate even their successful performances differently. When Griffey, as opposed to Belle, successfully runs down a fly ball, we think that this manifests competence, and is in that sense non-lucky. The present issue, then, is whether knowledge is best conceived as a) success attributable to ability-as-competence (RVE), or b) success attributable to “mere” ability, as opposed to ability-as-competence (Abilism). In favor of Abilism, Turri points to empirical data regarding folk judgments, which he interprets as allowing for unreliable and unsafe knowledge. Two theoretical considerations regarding the value of knowledge, however, count in favor of the robust view. First, the robust view is more in line with the Aristotelian idea that what is good in itself for human beings (human flourishing) is virtuous activity, in the sense of action from virtue. If knowledge is to be included in human flourishing, and thereby inherit the final value of human flourishing, knowledge must be understood as grounded in ability-as-competence as opposed to mere ability. Put differently, knowledge must be understood as competent (or virtuous) agency, as opposed to mere agency. Second, the practical value of knowledge and knowers is also bound up with the notion of ability-as-competence as opposed to mere ability. Hence, we value knowers not merely because they can get things right in relevant circumstances, but because they reliably get things right in relevant circumstances. Put differently, we value knowledge and knowers that we can depend on to get things right. Accordingly, the practical value of knowledge, as well as its final value, is best understood as grounded in competent agency, as opposed to mere agency. REF ERE NCE S Goldman, Alvin. . “Discrimination and perceptual knowledge,” Journal of Philosophy : –. Graham, Peter. . “Warrant, Functions, History,” in Naturalizing Epistemic Virtue. Cambridge, UK: Cambridge University Press, –. Greco, John. . “Safety in Sosa,” Synthese. Published online at https://link .springer.com/article/.%Fs–--z.  

 See Greco , . Cf. Greco . Thanks to Fernando Broncano-Berrocal, Allan Hazlett, Jesper Kallestrup, Chris Kelp, Jon Kvanvig, Duncan Pritchard, Joe Salerno, Ernest Sosa and John Turri for their comments on earlier versions of the chapter.



 

. “Post-Gettier Epistemology,” (Epistemologia Pós-Gettier), Veritas : –. . “Better Safe than Sensitive,” in Kelly Becker and Tim Black, (eds), The Sensitivity Principle in Epistemology. Cambridge, UK: Cambridge University Press. . Achieving Knowledge: A Virtue-theoretic Account of Epistemic Normativity. Cambridge, UK: Cambridge University Press. . “The Nature of Ability and Purpose of Knowledge,” Philosophical Issues, , The Metaphysics of Epistemology (Special issue of Nous): –. . “Knowledge as Credit for True Belief,” in Intellectual Virtue: Perspectives from Ethics and Epistemology, Michael DePaul and Linda Zagzebski (eds). Oxford: Oxford University Press, –. Kallestrup, J. and Pritchard, D. . “Robust Virtue Epistemology and Epistemic Anti-Individualism,” Pacific Philosophical Quarterly (): –. . “Virtue Epistemology and Epistemic Twin Earth.” European Journal of Philosophy (): –. Pritchard, Duncan. . “Knowledge and Understanding,” in Adrian Haddock, Alan Millar, and Duncan Pritchard, The Nature and Value of Knowledge: Three Investigations. Oxford: Oxford University Press. . “Knowledge, Luck, and Lotteries,” New Waves in Epistemology, D. H. Pritchard & V. Hendricks (eds). London: Palgrave Macmillan. Sosa, Ernest. . Judgment and Agency. Oxford: Oxford University Press, . . Knowing Full Well. Princeton, NJ: Princeton University Press. . A Virtue Epistemology: Apt Belief and Reflective Knowledge (Vol. I). Oxford: Oxford University Press. Turri, John. . “Knowledge Attributions in Iterated Fake Barn Cases,” Analysis (): –. a. “Vision, Knowledge, and Assertion,” Consciousness and Cognition  (C). Elsevier Inc.: –. doi:./j.concog.... b. “Knowledge and Assertion in ‘Gettier’ Cases,” Philosophical Psychology. doi:./... c. “Knowledge as Achievement, More or Less,” in Performance Epistemology: Foundations and Applications, Miguel Ángel and Fernández Vargas (eds). Oxford: Oxford University Press, –. a. “Unreliable Knowledge,” Philosophy and Phenomenological Research (): –. b. “From Virtue Epistemology to Abilism: Theoretical and Empirical Developments,” in Character: New Directions from Philosophy, Psychology, and Theology, Christian B. Miller, et al. (eds). New York: Oxford University Press. . “Manifest Failure: The Gettier Problem Solved,” Philosophers’ Imprint (): –.

 

Virtue Epistemology, Two Kinds of Internalism, and the Intelligibility Problem Jonathan L. Kvanvig

. Introduction A central issue in epistemology is the task of explaining how to go about making sense of the world and one’s place in it. The idea central to Empiricistic versions of Internalism is that such an explanation adverts, at bottom, to experience. Experiences constitute the basic evidence on which one’s view of the world depends, and as such, the view insists that the proper way to address the sense-making project is to start with the idea that the basic epistemological machinery involves psychological states of the individual in question. Such a position identifies evidence with psychological states, and thus embraces what we can call “Statism,” the view that evidence is some type of mental state, typically either experiences or beliefs. Virtue epistemologists start from a position of trying to understand what makes for a well-ordered cognitive life of the mind, and the answer they propose is that it is one on which the views and attitudes on display result from the intellectual virtues or conform to what virtuous people would make of the situations in which they find themselves. The virtues, and the people who exemplify them, thus provide models for how sensemaking is to be done. In both cases, however, one can be left with the feeling that what we wanted an explanation of has not been addressed. Suppose I have a set of experiences E, which the Statist identifies as the evidence, and then claims that because these experiences provide evidence for a vast array of opinions and attitudes, those opinions and attitudes are made intelligible by E. But such a claim doesn’t follow from the Statist assumption and isn’t explained by it. What we want to know about is something about the relation between evidence and attitude, and merely telling us what constitutes evidence doesn’t help. The same holds for virtue epistemology, since virtue epistemological assumptions don’t tell us about this relation either. We can think of a 



 . 

virtue as a function from certain features of the world, including character traits of the person in question together with whatever contextual factors might be relevant, to certain viewpoints about the world, a function that satisfies whatever requirements a virtue theorist might wish to insist on (e.g. reliability, stability, safety, sensitivity, etc.). Understood in this way, however, virtue theory doesn’t tell us what is in the functional box. A function is simply a “black box” specified in terms of inputs and outputs, and what we want to know about is inside the box. What we want, that is, is an explanation of why those views and attitudes are ones that make sense of the inputs to the function, why those views are made intelligible by the material that makes them epistemically appropriate. Contrast these two viewpoints with Propositionalism, the view that what makes certain attitudes appropriate to the situations in which we find ourselves is that those situations involve experiences and beliefs that have content and the content in question is evidence for the contents of the attitudes and opinions we form in those situations (when those attitudes and opinions are appropriate from a purely intellectual point of view). If these remarks are all there was to Propositionalism, it too would constitute a mere skirting of the intelligibility problem, for that problem – to repeat – arises out of a need to understand the central relation between our view of the world and the epistemological basis for that view. If the view is that the epistemological basis is evidence, then our question is about the evidential relation itself, and what we want to know is what it is about this relation that yields this intelligibility and sense-making that results from honoring the evidence. Propositionalism is designed to answer this question, but merely pointing out that evidence is a relation between propositional contents does not tell that story. So, what it says about evidence or epistemic grounds can’t be the end of the tale, on pain of making no progress on this central problem. The rough idea is that sensemaking results from seeing the connections between propositional contents that make up the story of evidence, but as we will see, this rough idea is not easy to refine into anything like an adequate response to the intelligibility problem. In what follows, then, I turn first to the problem at issue, showing that both Internalists and Externalists say nothing that directly addresses the problem in question. I then contrast two different versions of Internalism – Statism and Propositionalism – and show why the second may seem to have an easier time with the intelligibility problem. As we will see in the process, the metatheory involved in the Internalism/Externalism distinction does nothing to illuminate this central feature of an adequate

Virtue Epistemology, Internalism, and Intelligibility



epistemology, and as such should not be thought of as a fundamentally important way of carving up the possibilities in epistemology. It may be a distinction that isn’t quite as unnatural and gerrymandered as the distinction between grue and bleen, but neither is it as natural a distinction as that between blue and green. Moreover, given the difficulty of finding an adequate Internalist solution to the problem, one of the primary motivations for Internalism may lead us in the wrong direction in epistemological theorizing, for as we will see, the problem in question is precisely such a motivation. The end of the story might then just be good news for Externalist views, especially virtue epistemology, given the assumption I’ll make here that some version of it is the best hope for Externalism.

. The Sellarsian Intelligibility Problem The intelligibility problem plays a crucial role in Sellars’ use of a regress argument against Foundationalism. Sellars questioned whether and how basic beliefs could depend on, or be justified by, experience. He granted that experience could cause such beliefs and hence that such beliefs could causally depend on experience. But causal dependence isn’t justificational dependence, and it is justificational dependence that Foundationalism requires. Sellars formulated a dilemma for foundationalism at this point, for he maintained that experience couldn’t justify, or make rational or intelligible, the beliefs that it causes unless experience itself has propositional or informational content. He then insisted that if experience has such content, it can’t be a regress-stopper, but would instead be in need of justification itself. This last step ought to be resisted. Genes carry informational content, but don’t fall within the domain of epistemic normativity. Wishes and fears, hopes and dreams, have informational content, but also are not subject to epistemic appraisal. The contents in question can themselves be appraised epistemically, for we often have propositional justification for claims that we don’t believe or endorse. But the states in question are not the right kind of state to be subject to epistemic propriety or impropriety 

A brief word why. Externalism has two initially most promising versions, one reliabilist and one in terms of modal conditions such as safety or sensitivity. Solving certain problems for reliabilism moves one toward the kind of virtue views found in Ernest Sosa and John Greco, and addressing problems for modal epistemology pushes one toward virtue views as well, as we see in Duncan Pritchard’s work. To my eyes, it looks like you can begin climbing the Externalist mountain wherever you wish, and you’ll end up converging with the virtue path at some point as you near the peak.



 . 

(though, of course, they are subject to appraisal in terms of other normative notions, such as practical and moral appropriateness). In order for Sellars’ attempt to block the foundationalists’ way of stopping the regress of justification, an argument would be needed that epistemic appraisal applies not only to beliefs (and possible contents of belief ), but also to experience itself. Counting against such a claim is the wisdom embodied in ordinary language: “my visual experience of a white buffalo on the plains is rational” simply makes no sense. As ordinary language philosophers were wont to say, it’s a category mistake. Ordinary language, even if it embodies the wisdom of the ages (in some important and defensible sense), isn’t sacrosanct. So perhaps Sellars would have some room to argue the point further. Here’s one way, the best way I know. Start by pointing out that we should not assume the standard, Chisholmian view that the only objects of epistemic appraisal are doxastic items such as beliefs and possible contents of belief. There are other cognitive states that aren’t clearly reducible to belief that are equally subject to epistemic appraisal. First, there are degrees of belief, but more tellingly, there are assumptions, presuppositions, suspicions, conjectures, speculations, hunches, predictions, acceptances, confidence levels, and intellectual stances, and each of these things is subject to epistemic appraisal. You suspect that my latest counterexample to Plantinga’s theory of warrant misses its target, and your suspicion might be justified – “justified” in precisely the same sense of the term that your belief that Obama won’t be President ever again is (epistemically) justified. Perhaps these other cognitive states are all reducible to belief, but that would be quite surprising, at least to me. More likely would be a reduction of belief and these other states to degrees of belief, but even that would go beyond our current understanding of things. Think of the difference between Watson suspecting that the butler committed the murder, versus Holmes speculating the same thing. Those with an ear to hear will perceive a difference in attitude here, even though the cognitive element of degree of belief might be identical. I’m not sure what the difference is, but one possibility is that the latter involves a kind of theoretical disinterestedness that the former lacks: to suspect the butler may be to be affectively or attitudinally engaged with the issue, perhaps emotionally involved; but speculation can be idle. In any case, whatever the explanation, what seems clear to me is that the difference cannot be explained solely in terms of differences of degree of belief. 

See, e.g., Chisholm , , a, b.

Virtue Epistemology, Internalism, and Intelligibility



So, let’s suppose, at least provisionally, that epistemic appraisal applies more broadly than merely to the doxastic realm. Instead, we adopt the view that it applies to the cognitive realm more generally, leaving open that the cognitive realm includes things other than beliefs and degrees of belief. We thus limit the application of terms of epistemic appraisal, not to doxastic states and their contents, but rather to contents that are alethically evaluable and mental states with such contents, where the states display the direction of fit from mind to world that distinguishes cognitive states from affective ones, understood in terms of the opposite direction of fit: from world to mind. What we deny is that epistemic appraisal applies to the non-cognitive, whether that be actions or mental states that are either affective or conative. That domain – the non-cognitive – is where we find only normative appraisal of a different kinds, most obviously moral and practical normativity. Once we get to this point, however, Sellars can mount a challenge to the argument from ordinary language used above against the idea that experiences can be subject to epistemic appraisal. The challenge goes like this: “if epistemic appraisal applies more generally to the entire range of cognitive states, it ought to apply as well to experiences, since they too are cognitive states.” This challenge is worrisome, but not compelling. It supposes experiential states must be reckoned to be cognitive states, and that isn’t obvious. But even if we grant the point that experience is cognitive (as I’m inclined to think we should), perhaps that’s best thought of as a reason for thinking that extending the domain of epistemic appraisal to the entire domain of the cognitive is a mistake. Perhaps, instead, the arguments above only show that epistemic appraisal reaches into parts of the cognitive domain beyond the doxastic part of it, leaving open whether it applies to all cognitive regions or only some. On behalf of this more nuanced conclusion is, again, the argument from ordinary language: the language of epistemic appraisal seems completely out of place when applied to experience. A defender of Sellars has a further reply here. For the language of appropriateness is sensibly applied to experience. If you are sitting in a 

Cognitive states display a mind-to-world direction of fit in virtue answering to what the world is like in order to be correct. Affective states display the opposite direction of fit in virtue of the world needing to match the state in order for the state to be satisfied. Whether this loose metaphor can be given a decent theoretical explication is debatable, and whether it can be used to defend certain Humean views about the motivational impotence of cognitive states is also something I make no claim about here. For some of the relevant literature, see Humberstone , Sobel and Copp , Tenenbaum , Gregory , Frost , and Archer .



 . 

concert hall where an orchestra is playing, and you are having the experience as of a white buffalo on the plains, your experience is not appropriate to your surroundings. Your experience is abnormal, unfitting, out of place, and inaccurate. So, a Sellarsian can claim that such assessments show that the regress can’t be stopped by attempting to exclude experience from the domain of epistemic appraisal. Isn’t this point enough for Sellars’ argument? Not quite. Each of these assessments is evaluative in character, but the attempt to block the foundationalists’ regress stopper involves assessment that is different from the evaluative. The evaluative falls within the theory of value, not within the theory of obligation and permission that is the stuff of epistemic appraisal. The theory of value, epistemic or otherwise, is one thing; the theory of obligation another. This point remains true even if the theory of obligation we defend turns out to be teleological, defining the right in terms of the good. Sellars’ argument requires assessment for experience that goes beyond the domain of the good; it requires assessment in terms of the right. And the assessments above do not do that, so the ordinary language resistance to Sellars’ argument remains unscathed by the assessments just noted. The conclusion I think we should endorse is that Sellars’ attempt to start up the regress of justification again, in response to the foundationalists’ attempt to stop the regress by appeal to experience, fails. It fails because the language of epistemic normativity isn’t correctly applied to experience. Even so, there is a deeper concern raised by Sellars’ effort to undermine foundationalism. It raises the concern that there is no good sense-making story, but only a causal one, to be told about the connection between belief and experience. The result of a story that is merely causal is that there are two conscious, mental states that are connected causally, but the connection between the two states is not a matter of there being states that have informational content which makes forming the belief in question upon having the experience in question a sensible thing to do from the point of view of the agent in question. If I am appeared to F-ly, it makes sense from my perspective to believe that something is F; if it seems to me that p, then believing that p is a way of making sense of my experience of the world. But if the only story to be told is a causal one, none of these claims is true: there is no sense-making feature, there is no agreement, there is no 

A full discussion of this point would require careful attention to the fact that the language of “ought” in ordinary language functions as both a normativity indicator and an evaluative indicator. The mere fact that we find such language in certain places isn’t sufficient for showing that we are in the territory of the theory of obligation instead of the theory of value. Since I’ve written extensively on this issue in Kvanvig , however, I’ll sidestep it for present purposes.

Virtue Epistemology, Internalism, and Intelligibility



intelligible link between experience and belief. The connection, to use the term in its older, nineteenth-century sense, cannot be rationalized. Moreover, the epistemic value of experience should be accounted for, not simply in terms of some sort of direct and immediate causal contact with the world and the particulars that constitute it, but also in terms of its power to make manifest to the understanding what the world is like. In order to fulfill these functions, the connection between experience and belief cannot be merely a causal connection but must also be a connection that makes sense from the perspective of the agent in question, yielding a content for belief that is fitting from the point of view of the agent in question in response to the experience itself and its informational character and content. It is important to note that this intelligibility requirement does not threaten only (various versions of ) Externalism, but also threatens both of the important strands of Internalism mentioned here: Statism and Propositionalism. One might think the latter view immune to the problem, but careful reflection on the dialectic of Sellars’ problem shows that turning to Propositionalism is perhaps necessary for solving the problem, but nothing in that dialectic approaches the idea that turning to Propositionalism will solve the problem. As we will see, Sellars’ reflections pose a problem for everybody.

. Intelligibility and the Internalism/Externalism Distinction It might seem that all hope is lost for Externalists, once the problem of intelligibility is appreciated. For the problem is essentially a problem about acknowledging the role of the first-person perspective in a theory of epistemic appraisal, and Externalism is more at home from a third-person perspective, where we assess the standing of various cognitive states of a person from a point of view that might not be shared by the individual whose states are being evaluated. So, for example, if one wishes to assess whether a given belief is reliable – produced or sustained by processes or methods that generally get one to the truth (either conditionally or unconditionally) – one’s assessment doesn’t depend on connections intelligible from the cognizer’s perspective; and the same is true whether one’s assessment is done in terms of safety (e.g. Pritchard ) or sensitivity (e.g. Nozick ) or truth-tracking (e.g. Roush ) or proper function (e.g. Plantinga b). 

See Goldman  for details.



 . 

One might initially think that the same thing applies to virtue epistemology, but there is a glimmer of hope here that isn’t found in other versions of Externalism. John Greco (: ) articulates a virtue theory on which “A belief p has positive epistemic status for a person S just in case S’s believing p results from stable and reliable dispositions that make up S’s cognitive character,” but also attempt to specify a notion of subjective justification that is more responsive to first-person-perspective concerns. He claims that objective justification amounts to a belief being the result of dispositions that make a person reliable regarding that belief in the conditions in question, and subjective justification involves a belief being “the result of dispositions that S manifests when S is thinking conscientiously” (: ). This approach provides a luster of hope, but it fades on closer inspection. For the dispositions one manifests when thinking conscientiously might or might not make intelligible the connections between input and output that characterize the dispositions in question. An evil genius could implant a module that is triggered whenever conscientious thinking is detected, and results in whatever bizarre beliefs one wishes to imagine from any particular input experiences. Moreover, even if such cases don’t show that conscientious thinking is compatible with such opacity, we have no explanation of the intelligibility in question merely by noting that conscientious thinking guarantees it. If the problem were that easy to solve, any theory could posit the same kind of brute fact for its favored justificationmakers. As a result, we can’t secure a solution to the intelligibility problem by focusing on what occurs when a person is thinking conscientiously. One might think that Internalism has an advantage here, but that would be hasty at this point. If there is a version of Internalism with decent prospects for addressing the intelligibility problem, it will take some care to construct it. For the difference between Internalists and Externalist is a difference concerning the factors that make for positive epistemic status. Internalists insist that the ground or basis of such status is found in internal mental states of one sort or another, and Externalists deny this restriction. But neither viewpoint addresses directly what is involved in the intelligibility problem. A bit of formal representation will help to make this point obvious. Regarding positive epistemic status, there are (at least) three elements involved. First, there is the target of such status, which we might think of as the content of a belief, either actual or potential. Let us represent this item as “Bp” where p is the content in question. Second, there is the ground or basis for the epistemic status in question. In order to leave open

Virtue Epistemology, Internalism, and Intelligibility



whether this item is internal or external, let’s simply represent it as “Γ.” These are the relata of the positive epistemic status relation, J, so that if some ground is suitable as a basis for some proposition, we can represent this fact as JΓBp. Notice that the dispute between Internalists and Externalists concerns potential candidates for the Γ relatum. The intelligibility problem, however, is not about either relata but rather about the relation J itself. It is about what is involved in the relation between the relata that allows for the first relatum to be intelligibly related to the second relatum. So, settling the dispute about what are the appropriate instances of Γ will do nothing, by itself, when what we want is to be told what it is about the J relation that links Γ with Bp in such a way that the connection between the two is intelligible to the person in question. Note, especially, what kinds of positions don’t help here. Perhaps Γ makes Bp likely to be true, or modally safe or sensitive, or an instance of proper function or intellectual virtue or what a virtuous person would or might believe in such circumstances. Even if such claims are true, they don’t explain how Γ generates the sense-making feature that is required for addressing the Sellarsian intelligibility problem. Moreover, this point remains untouched if we insist that Γ is restricted to information that is in the form of a mental state or even if we insist that the information is accessible on reflection. As such, it is a bit surprising to notice that the kind of Internalism Sellars favored – one on which nothing can make for positive epistemic status without having propositional content (“Propositionalism” for short) – was embraced specifically to make sure that whatever generates positive epistemic status is the kind of thing that solves the intelligibility problem. But Propositionalism, by itself, does no such thing. Once we get clear on the formal representation of positive epistemic status, Propositionalists are embracing a theory about the appropriate values for Γ, when the intelligibility problem isn’t about what restrictions govern a decent account of values for Γ but rather about something involving J itself. Thus, what we need to consider is whether there are ways to supplement any of these positions so that they make some progress on the intelligibility issue itself. To that issue I now turn.

. Broken Promises, Shattered Hopes Imagine trying to solve the problem by stipulation: Γ is the right kind of thing for making a belief likely to be true, or safe or sensitive or a display of



 . 

virtue or proper function only if it makes sense from the perspective of the person in question how the belief in question is made intelligible, is revealed to be true (to that individual), by Γ. Such a ploy would address the intelligibility problem but not in a theoretically satisfactory way. For what we want to know is what it is about the things that can be values for Γ that generate such a sense-making feature. In order to answer that question, we must talk directly about what kinds of things can be values for Γ, rather than merely indicating that they have the desired feature. On this score, Internalists may seem to have a natural advantage, since their view is designed around the information contained in appropriate values for Γ, but as we shall see, this natural advantage can easily slip away. Consider the difference between Propositionalists and Statists concerning Γ. Both have the advantage of being able to claim that there are values for Γ involving propositional content. Because of this fact, both kinds of theory can claim that intelligibility arises because of the connection between the propositional contents in Γ and the propositional content involved in the target belief. A few simple examples show that no such promise can be sustained. Suppose something with the content the liquid in this glass is water is the value for Γ and the target belief has as content the liquid in this glass is H. Here, we arguably have identity of propositional content, and if any relation involving propositional content can make intelligible something involving the second proposition on the basis of something involving the first, one would expect that identity will make the connection as intelligible as anything could. Yet, believing the second on the basis of the first does not generate intelligibility in the absence of some further connection depending on the claim that water is H. Similarly, if in Lois’ collection of positive epistemic status generators is the claim that Superman has arrived, such a presence doesn’t make intelligible to Lois that Clark Kent has arrived. Such is the case even if the two sentences express precisely the same proposition. Propositionalism is the version of Internalism aimed directly at solving the intelligibility problem, since the point of that problem, in the hands of Sellars, was to show that an appeal to propositional content was necessary and that once in place, the regress of justification begins anew. According to Propositionalism, the distinction between there being evidence for a view and one’s having evidence for that view is the difference between one and the same content being encoded in a relevant cognitive state, or not. So, to have evidence for p is to be in a mental state that has some content that counts in favor of the truth of p, and it is in virtue of the content of the state that it epistemically supports the belief in question. It does so

Virtue Epistemology, Internalism, and Intelligibility



because it is the content or character of the state that provides the evidential connection to the content of the belief in question. In short, an experience or belief of a certain type confirms a belief with a certain content for a person at a given time and in a given context because the relevant typing of the experience or belief is in terms of its informational content, and that informational content is evidentially tied to the content of the supported belief. Note, however, that describing the view in this way is done without ever mentioning the notions of intelligibility or sense-making that are at the heart of the problem Sellars raises by distinguishing merely causal relations from justificatory ones. Perhaps the intent of Propositionalists was to design a theory to meet the Sellarsian strategy, but the execution of the plan reveals a theory more accurately described in terms of avoiding Statism, the view according to which the evidence is the mental state itself, rather than the content of the state. As described, Propositionalists claim that Statists can’t accurately characterize the distinction between the evidence itself and the having of it, but even if that point is well taken, Propositionalism so developed appears to have gotten derailed from the Sellarsian point that motivated it. The plan was supposed to involve embracing the mantra that only Propositionalism offers any hope of addressing the problem, and then characterize the view so that it is not only necessary but sufficient for addressing the problem. As we have seen, on the basis of the examples cited, the view developed as an alternative to Statism simply doesn’t solve the problem. We turn in the Section . to the issue of whether it can be supplemented to do so.

. Propositionalism and the Intelligibility Problem The examples used above as counterexamples to the idea that propositional content is sufficient for generating intelligibility between evidence and what it is evidence for will likely prompt the reader to think of Frege’s reasons for introducing modes of presentation, and that is clearly the correct way to try to supplement propositional content in a manner that gives some hope of addressing the intelligibility problem. The idea is to restrict Γ and the target of justification in such a way that justification ( J ) obtains only when intelligibility is preserved, i.e., only when opacity between epistemic support and target of justification is not present. 

For development of this objection, see Kvanvig a.



 . 

To such an end, let us grant that there is more to informational content than propositional content alone, and introduce guises as the additional informational element that encodes the modes of appearance that a triadic Fregean semantics requires. We can thus say that Γ is the evidence for the target of the J relation, but that it contains informational content and not merely propositional content. So, for each proposition that is evidence for the target of justification, there is a guise g for that proposition and only guised propositions count as evidence. To accommodate this idea formally, let us take Γ to be the set of such contents that are evidence for the person in question (sometimes taking it to include all evidence for anything, and sometimes taking it to include only items relevant to a particular proposition and the beliefs that have that proposition as content). We can then say that, if x2Γ, then x has the form g[e], where e is a proposition and g a guise of e. Similarly, the target of justification is not simply a proposition either, but rather a proposition under a guise: g[p]. Finally, we interpret the fundamental J relation – that of propositional justification – as a relation between guised propositional contents. None of this machinery is sufficient in itself to address the intelligibility problem, however, for guises themselves don’t automatically generate the needed contexts of obviousness or transparency concerning the holding of the justificatory relation J relation between the evidence g[e] and target g[p]. To see this point, note that knowing that Superman has arrived can help to justify the claim that Clark Kent has arrived, but only in given contexts (where it is known that the two names refer to the same person). To get the needed clarity of perspective, the guises must be targeted or indexed toward that which they support epistemically, and the target itself must be suitably indexed as well. The first index can be encoded as follows: instead of g[e], we needgp[e] in order to be included in Γ and thus be evidence (on our current assumption that the version of Propositionalism being developed is a version of Evidentialism). This point accords with a putative truism about evidence: evidence is that which points toward something (else). The suggested encoding specifies what the evidence points toward, and thus as learning proceeds the very same informational content can come to be encoded with additional pointings, accommodating the fact that we not only gain evidence through experience but also  

For details, see Salmon . By including such indexing, we avoid the problem of a person knowing both that Superman has arrived and that Superman is Clark Kent, but not having connected those two thoughts and thus believing that Clark Kent has arrived but without that belief counting as a sense-making episode for the person in question regarding the arrival of Superman.

Virtue Epistemology, Internalism, and Intelligibility



sometimes learn what to make of our evidence as well. But what of the second index, the one involving the target of justification? Note what we cannot do. We cannot say that the target is indexed to the evidence itself by representing the target as ge[p]. If we did so, it looks like we are endorsing the idea that one thing is evidence for a second only if the second is also evidence for the first. We could stop that impression by insisting that Γ contains only basic evidence, and thus that only one of the two indexed guises could be included in Γ. Such a restriction, however, would make it difficult to explain chains of reasoning, where targets of justification become evidence for future steps in the chain once they become justified by that reasoning. Perhaps, though, there isn’t a need for the second index. The idea would then be that something is evidence only when it has an indexed guise, where the index derives from what the guised evidence obviously or transparently supports. Moreover, targets of such justificatory support don’t themselves count as evidence until they are no longer merely guised propositions but propositions under indexed guises. Such a position sounds inordinately psychological for the context of propositional justification. After all, central to Propositionalism is the view that the evidence is one thing, and the evidence for you, or your evidence, or the evidence you have, is another. If we take targets of justification to be unindexed, we would then be presupposing some function on such targets that turns their guises from unindexed to indexed, and the only such function that seems to make any sense here is a psychological one: becoming aware of a guised proposition as epistemically relevant to some other guised proposition, and thus for the first to become a guised proposition indexed to the second. Such a maneuver would be guilty of failure to distinguish the evidence itself from the evidence one possesses. Statists would be happy with such a result, since they wish to identify the evidence with mental states themselves, thereby viewing the intrusion of psychology into the story as a benefit rather than a liability. Statism, however, has rather debilitating defects, including the above-noted difficulties addressing the intelligibility problem, so it is best at this point to view the intrusion of psychology here as unfortunate. We must thus find a way to avoid that impression. Let us pause for a moment to take stock of where we are. We first noted that Propositionalism alone doesn’t solve the intelligibility problem unless Fregean guises are introduced into the picture. But such guises alone don’t 

I’ve argued for this conclusion in a number of places, most recently in Kvanvig a.



 . 

solve the problem either, unless the guises are epistemically indexed to the propositional contents that are the targets of the justification relation. Furthermore, the targets themselves would seem to be in need of similar indexing in order for them to function correctly in chains of reasoning. But how are we to characterize the indexing of the targets, without resorting to psychological language that provides fodder for Statist resistance to Propositionalism? Perhaps we might view the target of justification as variably indexed, distinguishing between what can be in Γ and what cannot (among justified informational contents) by whether or not the indexing is variable. On this approach, the evidence is always of the form gp[e], thereby involving no variable in the indexing, whereas the targeted proposition can be variably indexed, having the form gx[p], where x is a variable for some indexing proposition or other, with no restrictions on what it might be other than those imposed by the J relation itself. For example, if the J relation is asymmetric, then x6¼e, and if the J relation is nonsymmetric, but not asymmetric, we can allow that x=e only in cases that count as counterexamples to asymmetry. But targeted propositions need not be variably indexed, and when guised propositions with non-variably indexed contents are targeted they can both be justified and become suitable members of Γ, capable of playing a role in justifying other guised propositions. Note here that guises are being modeled in a way that is not psychological. What it is to be evidence involves only informational content, and what it is for something to be evidence for you involves accessing propositional contents either in the form of experience or belief in such a way that the guises for the propositions are present in the mode of awareness or attention. They are, after all, modeled after Fregean modes of presentation. Moreover, the propositions targeted can also become evidence for you in reasoning to further conclusions when the guised propositions are specifically rather than variably indexed. Just what psychological processes generate that result is not part of the theory of propositional justification but would be relevant when attempting to say what it is to base one’s belief on one’s evidence for that belief. On a version of Propositionalism that counts as a version of both Evidentialism and Empiricism, basic evidence is always in the mode of experience, with content in the form of indexed guises that direct the mind toward propositional targets. These targeted propositions also appear in the mode of indexed guises, where the index can itself be either variable or specific. Allowing both possibilities is essential for making sense of our intellectual lives, for we often can see that some proposition is true

Virtue Epistemology, Internalism, and Intelligibility



(as shown by our evidence), even when we can’t see any further implications generated by this newly discovered information. Note the progress, but also the difficulty that remains. The progress is that we’ve moved beyond a view that identifies informational content with propositional content, and we’ve also moved beyond a view that supplements propositional content with Fregean modes of presentation. Instead, we’ve seen that if such a framework provides hope of solving the intelligibility problem, we need not only such modes of presentation, but we need them to be indexed to that which they obviously or transparently support (from a purely theoretical or intellectual point of view). The difficulty that remains can be put in the form of a question for the summary in the last paragraph of what an Empiricistic Evidentialist might say. The question is this: What is the story for how propositions become justified for a person and then become part of that person’s evidence? In the story above, propositions get supported by one’s evidence, and sometimes they are supported under guises indexed variably and sometimes not. The psychological difference between the two cases is the difference between seeing where the new information takes one, and not so seeing. In the former case, it is easy to see what to say: in such cases, the proposition under the indexed guise becomes part of one’s new body of evidence. In the latter case, it does not. But then a psychological change happens in which a person comes to see the significance of the new discovery, which, under the model in question, replaces the variable index with one having specific content. Such a transition can be rational or not, and the Propositionalist needs to be able to tell a propositional story about that transition. As described, though, there is no obvious such story to be told. Note well that the story to be told has to advert to informational content alone. It is the Statist, not the Propositionalist, who can talk about transitions from one mental state to another. So the only resources here are to find a way to get from an informational content of the form gx[p] to another of the form g 0q ½p, where the latter is an informational state that allows propositional support for q from p that addresses the intelligibility problem. And yet there is nothing in the former informational state that makes the transition intelligible between a variably indexed content and a specifically indexed one. The best hope for Propositionalism regarding this issue begins with the idea that it is something about the guise itself, indexed in the way it is, that leads to another (or the same) guise indexed in a different way. Perhaps the seemingly insurmountable roadblock for Propositionalism arises because



 . 

we are not correctly assessing inputs into the system of justification. When we get a transition from gx[p] to g 0q ½p, the Propositionalist has no resources for explaining the connection, until we notice that the introduction of basic evidence into the system is, at least for Empiricistic Evidentialists, a function of experience itself. For others, sources of basic evidence might arise in other ways. We need not concern ourselves with this variability in the land of Propositionalism but can focus instead on the creation of evidence through whatever sources there are. The transition from one state of information to another might be modeled as inference or support from evidence, where the indexing makes transparent that there is such a connection. But transitions also occur from one total state of information to another through the introduction of basic evidence. In such a case, a person comes to “see” that g 0q ½p is the more perspicuous rendering of gx[p]. That is, the person, in the mode of experience, sees a connection between two states of information, and it is the content of this seeing that provides the rational basis for the transition in question, a transition that satisfies the intelligibility requirement in virtue of the appropriateness of the visual metaphor. Upon seeing the connection and attending to it, the transition from gx[p] to g 0q ½p is as straightforward as are inferences by modus ponens (though, to be clear, the analogy is not meant to endorse that the transition should be modeled inferentially). Understood in this way, Propositionalism needs a special kind of intellectual seeing in order to address the intelligibility problem adequately, and such seeing is likely to cause problems for Empiricistic version of the view. For the kind of seeing in question appears to be a special case of rational intuition of just the sort from which Empiricists recoil. It is worth noting, however, that if the path followed is correct, the intelligibility problem provides a watershed issue for Internalists, one that is so central to the view that to give up on it as insoluble is to remove a plank essential to the epistemological seaworthiness of the view. In short, the mode of resistance to attacks on Internalism from Virtue Epistemologists has to involve some kind of appeal to the intelligibility problem, and only Propositionalism has a way of telling the Internalist story in a way that has any hope of solving that problem. 

An objection here: there is no such thing as an object of thought in which the Fregean guise is nonspecific. So, if there are guised propositions variably indexed, they cannot be the informational contents of thought, but rather should be said to be sets or classes of the informational contents that are possible contents of thought. I think this point is correct, but can be ignored for present purposes, since it would merely complicate the account in the text, without changing the basic contours of what is needed for a Propositionalist solution to the intelligibility problem.

Virtue Epistemology, Internalism, and Intelligibility



It is worth noting in passing at least that this appeal to rational intuition does not make the resulting view elitist in any way. Such intellectual seeing of the sort needed is part of the basic architecture of mentation itself, simply a matter of recognizing that one thing shows another thing to be the case, points in the truth-indicating way toward something beyond itself, and even animals and small children reveal such recognitional capacities. The threat to Empiricism that results from the intelligibility problem is thus not a threat that undermines the possibility of knowledge or justification for even the least advanced cognitive agents. For those who cannot stomach the resulting loss of Empiricism, it is easy to see the attraction of virtue epistemology. Without intelligibility strictures, virtue epistemology yields all the promise of reliabilist Externalism as well as something close to Internalist justification in the form of conscientious belief formation and revision. Where such a view falls short, no version of Empiricism can salve the sore and no version of Statism, whether Empiricist or not, helps either. The moral of the story is either to embrace Propositionalism or explain why the intelligibility demand is incoherent or problematic in some other way.

. Conclusion Propositionalism, in Sellarsian hands, was supposed to be designed to allow the sensibility of trusting our senses when it comes to believing on the basis of experience. Our senses allow us to gather evidence, and the evidence we thereby have is evidence for what we believe because of the content of the experience. We thus trust our senses by trusting what they tell us. Propositional content alone, however, cannot live up to the billing. Propositional content for the evidence we have, whether in the form of doxastic states or experiential states, may only opaquely signify what they are evidence for. There is, in this way as in other ways, both the evidence and what we make of it. The trick for Propositionalism is to prevent this conjunction from devolving into a multiplicity of mental states, and the way to avoid that is Fregean. On a Fregean conception, what we make of the evidence is central to the individuation of modes of presentation for the propositional evidence in question, since modes of presentation are theoretical postulates designed to explain cognitive significance, one aspect  

Again, see Greco  for details. For details on other ways in which there is such a difference, see Kvanvig .



 . 

of which is the role that the evidence plays in defeasible reasoning (as well as in perceptual recognition that is sometimes modeled in terms of reasoning but is better thought of as not involving inference at all). Not even modes of presentation are enough to solve the Sellarsian problem, however, and in seeing what else is needed, we uncover yet another argument against Empiricism. More to the point, we uncover the need for modes of presentation epistemically indexed to what they support, and we uncover the need for distinctively intellectual experiences that link some informational states of the sort described with other such states. It is only in this way that the Propositionalist has something helpful to say about the intelligibility problem. Justification is not a story about the evidence and what it supports, but rather a story about the evidence and the way it is presented that makes for both cognitive and epistemic significance. The Propositionalist thus not only embraces Fregeanism about informational content, but expands on it, yielding the only view with any hope of solving the intelligibility problem. None of the above should blind us to the burden still faced by Propositionalism, which I have not addressed here. As envisioned, Propositionalists insist that a connection between two different informational contents can occur both in doxastic form (in the form of a conditional) and in the form of experience. If the connection between the two were doxastic, the intelligibility problem would not be solved; if the connection is in the form of experience, it can be. Central to such an approach, then, is that when the transitions in question are embodied in cognition itself, the transition cannot be rationally accounted for when in doxastic form. As presented, I treat this implication as a pleasing even if unanticipated consequence. Is such treatment mere bluster in the face of an obvious objection to the view? Nothing in the above responds to the complaint that this consequence is not an unpleasant one, so the Propositionalist has to do more work to sustain the view that an adequate account of epistemic justification or rationality must solve the intelligibility problem. Completing the project thus demands more work, but today has already had enough trouble of its own.



It is worth noting that John Pollock , , defends something similar, arguing for a theory of meaning on which cognitive role plays a definitive role. There is not space to investigate the issue fully now, but the Fregean picture endorsed here is able to utilize certain advantageous elements of a conceptual role semantics without embracing such a point of view. Refusing to do so is a benefit, since such semantical theories have well-known disadvantages (for details, see, e.g., Fodor , Block , Whiting ).

Virtue Epistemology, Internalism, and Intelligibility



REF ERE NCE S Archer, Avery. . “Reconceiving Direction of Fit,” Thought: A Journal of Philosophy (): –. Block, Ned. . “Conceptual Role Semantics,” in Routledge Encyclopedia of Philosophy, Edward Craig (ed.). Boston, MA: Routledge, –. Chisholm, Roderick. . Perceiving. Ithaca, NY: Cornell University Press. . Theory of Knowledge. Englewood Cliffs, NJ: Prentice-Hall. nd edn. a. “Firth and the Ethics of Belief,” Philosophy and Phenomenological Research (): –. b. Theory of Knowledge. Englewood Cliffs, NJ: Prentice-Hall. rd edn. Fodor, Jerry A. and Ernest LePore. . Holism: A Shopper’s Guide. Cambridge, MA: Blackwell Publishers. Frost, Kim. . “On the Very Idea of Direction of Fit,” Philosophical Review (): –. Goldman, Alvin. . “What is Justified Belief?” In Justification and Knowledge, George Pappas (ed.). Boston, MA: D. Reidel, –. Greco, John. . Putting Skeptics in Their Place. Cambridge, UK: Cambridge University Press. Gregory, Alex. . “Changing Direction on Direction of Fit,” Ethical Theory and Moral Practice (): –. Humberstone, I. L. . “Direction of Fit,” Mind (): –. Kvanvig, Jonathan L. . Rationality and Reflection. Oxford: Oxford University Press. . “McCain on Propositionalism,” in Believing in Accordance with the Evidence: New Essays on Evidentialism, Kevin McCain (ed.). New York: Springer Publications. Nozick, Robert. . Philosophical Explanations. Cambridge, MA: Harvard University Press. Plantinga, Alvin. . Warrant and Proper Function. Oxford: Oxford University Press. Pollock, John. . Knowledge and Justification. Ithaca, NY: Cornell University Press. . Contemporary Theories of Knowledge. Totowa, NJ: Rowman and Littlefield. Pritchard, Duncan H. . Epistemic Luck. Oxford: Oxford University Press. Roush, Sherrilyn. . Tracking Truth: Knowledge, Evidence, and Science. Oxford: Oxford University Press. Salmon, Nathan. . Frege’s Puzzle. Oxford: Oxford University Press. Sobel, D. and D. Copp. . “Against Direction of Fit Accounts of Belief and Desire,” Analysis (): –. Tenenbaum, Sergio. . “Direction of Fit and Motivational Cognitivism,” in Oxford Studies in Metaethics, Russ Shafer–Landau (ed.). Oxford: Oxford University Press, –. Whiting, Daniel. . “Conceptual Role Semantics,” in Internet Encyclopedia of Philosophy. www.iep.utm.edu/conc-rol/. Accessed January , .

 

Knowledge Is Extrinsically Apt Belief Virtue Epistemology and the Temporal Objection Anne Meylan

. The Account of Knowledge and the Categorial Claim Let me start with a few quotations that nicely formulate the core of Ernest Sosa’s virtue epistemology. If a performance does succeed through the exercise of that competence in its proper conditions, then it is an apt performance, one creditable to the performer. Knowledge is just a special case of such creditable, apt performance. (Sosa : ) Belief is considered a kind of performance, which attains one level of success if it is true (or accurate), a second level if competent (or adroit), and a third if true because competent (or apt). Knowledge on one level (the animal level) is apt belief. (Sosa : ) Performance whose success manifests the relevant competence of the performer avoids thereby a kind of luck. According to competence virtue epistemology, knowledge is a special case of that. Knowledge of a sort is belief whose correctness is attained sufficiently through the believer’s epistemic competence, belief that is thus ‘apt’. (Sosa : )

These quotations contain two main claims regarding the fundamental nature of (animal) knowledge. First, Sosa tells us what, according to him, the components of knowledge are, what the correct account of knowledge is. The Virtue Epistemological Account of Knowledge An instance of knowledge consists in a successful (true) because competent belief, that is, in an apt belief.

The second claim concerns the metaphysical category to which knowledge belongs. For lack of a better term I shall call this ‘the categorial claim’. 

Sosa distinguishes two kinds of knowledge: animal and reflective. This chapter focuses exclusively on animal knowledge. For sake of brevity, I won’t mention this henceforth.



Knowledge Is Extrinsically Apt Belief



The virtue epistemological categorial claim A belief that constitutes an instance of knowledge is a performance (or an achievement).

What the categorial claim asserts is that a belief that constitutes an instance of knowledge belongs to the same category as, for instance, my shooting at a target, my baking of a tarte Tatin or the playing of a Beethoven sonata. A belief that constitutes an instance of knowledge is also something that the believer achieves or carries out. That the virtue epistemological account of knowledge and the categorial claim constitute two distinct claims even though their respective formulations are often entangled is crucial to bear in mind throughout this chapter. The present chapter aims to achieve two purposes and is, accordingly, divided into two parts. The goal of the first part (Sections . – ..) is to raise an objection against the categorial claim. Matthew Chrisman (, , ) has also recently argued against the categorial claim. As I shall also try to show, my own objection – which I call the temporal objection – has the advantage of blocking some replies that could be made on behalf of the virtue epistemologist to Chrisman’s objection. For this reason, the temporal objection is more compelling. The second part of this chapter (Sections .. and ..) is devoted to showing that abandoning the categorial claim is not as costly as virtue epistemologists might initially think. The virtue epistemological account of knowledge can still be maintained even if the categorial claim is rejected.

. The Temporal Objection As noted, I dub the main objection that I would like to raise against the categorial claim ‘the temporal objection’. It goes like this: The temporal objection . beliefs are states; . performances (like achievements) have temporal parts; . states do not have temporal parts; . states are not performances (or achievements); (,) C. Thus, beliefs are not performances (or achievements). (,) The temporal objection is largely inspired by an objection to the categorial claim which has been raised by Matthew Chrisman in a series of recent



 

papers (, , ). Here is a hopefully faithful reformulation of Chrisman’s objection: *. *. *. *. C.

Beliefs are states; Performances (like achievements) are active entities; States are not active entities; States are not performances (or achievements); (,) Thus, beliefs are not performances (or achievements). (,)

Premise * is, of course, crucial and is common to Chrisman’s and the temporal objection. Chrisman’s own way of backing up premise * is by relying on certain linguistic features of the verb ‘to believe’ that suggests it is a stative verb, viz. a verb that denotes a state (and not an event or a process). Here is an example of a linguistic feature that, according to several linguists and philosophers, helps to distinguish verbs denoting states from verbs denoting events or processes. The criterion of the verb’s aspect Verbs that denote events or processes can take the (present, past or future) continuous, while verbs that denote stakes, most often, cannot.

Even though ‘I was baking a tarte Tatin’ and ‘I am shooting at a target’ are correct English sentences, it is not proper English to say ‘I am believing that you are going to miss your train’. In most cases, the English verb ‘to believe’ – as with the English verbs ‘to desire’, ‘to love’, ‘to want’ – cannot properly be conjugated in either the present or the past or the future continuous. This gives us reason to think that ‘to believe’ denotes a state rather than an event. This is, put briefly, one first way of supporting the first premise of Chrisman’s objection to the categorial claim. 



 



Chrisman’s  paper provides, to my knowledge, the first carefully articulated argument supporting premise *. The other crucial premise is premise *, which is more extensively defended in his  paper. See Chrisman : . In his  article, Chrisman relies on the dynamicity of performances to build his argument. That is, premises * and * say that performances are dynamic entities while states are not, rather than that states are active entities while states are not. This nuance does not have any influence on what I say below. Kenny : ; Mourelatos : ; Vendler : –. In some circumstances, the verb ‘believe’ seems to admit of the continuous. For instance, it sounds correct to say: ‘At this time he was still believing that she would be the perfect employee.’ Hence, the criterion of the verb’s aspect does not constitute an infallible test. Thanks to M. Chrisman and Ch. Kelp for having drawn my attention on this point. Another linguistic feature of verbs that is supposed to ground their classification in the category of the stative verbs is the following: when Φ denotes a stative verb, ‘S has Φ-ed’ implies that ‘S still Φ-s.’ Only stative verbs are supposed to warrant this kind of implication. In contrast, my having acted foolishly, for instance, does not imply that I am still acting foolishly now (even though this might

Knowledge Is Extrinsically Apt Belief ..



Chrisman’s Objection: Weaknesses

Chrisman indubitably makes ingenious use of linguistic data to cast doubt on Sosa’s categorial claim. However, his objection leaves at least two replies open to the proponent of virtue epistemology. Let me present each of them in turn. First, as Mourelatos has emphasised, the criterion of the verb’s aspect allows exceptions. There are contexts in which a verb that is usually a stative verb can take the (present, past or future) continuous. For instance, it is grammatically correct to say ‘I am understanding more about quantum mechanics as each day goes by,’ and thus, to conjugate the verb ‘to understand’ in the present continuous even though ‘[t]he special affinity of “know” or of “understand” for state contexts is beyond doubt’ (Mourelatos : ). Or consider the verb ‘to crave’. It commonly takes the present continuous, for instance in the sentence ‘I am craving ice-cream,’ even though to crave something is to be in a certain state (namely, a state of intense desire or need). The fact that the linguistic criterion of the verb’s aspect admits exceptions leaves open the possibility that certain beliefs are not states but, as the virtue epistemologists say, performances. However, since the verb ‘to believe’ cannot take either the present or the past or the future continuous in most cases, it remains the case that this verb has ‘special affinity’ with state contexts. For this reason, I do not think that the fact that the criterion of the verb’s aspect allows exceptions is of great help to virtue epistemologists. There is a second reply that is available to the proponents of virtue epistemology, which is also more convincing. One of the premises that ground the inference from premise * to the conclusion that beliefs are not performances in my reformulation of Chrisman’s objection is the following: Premise *: States are not active entities – viz. entities that involve some form of agency; Premise * is controversial. As Chrisman (: –) himself recognises, Matthew Boyle has recently argued that there are active states and even, more importantly, active beliefs. Now, this is certainly a line of defence

 

be true). See Kenny : . This criterium does not, however, deliver the right verdict in the case of ‘to believe’. ‘S has believed that p’ does not necessarily imply that ‘S still believes that p.’ Thanks to M. Chrisman for having drawn my attention on this point. The fact that the verb ‘to crave’ commonly takes the present continuous has been suggested to me by Maria Scarpati. See, e.g., Boyle .



 

that the proponents of virtue epistemology might consider embracing in order to rebut Chrisman’s objection. The advantage of my objection (the temporal objection) over Chrisman’s objection is that it does not leave the aforementioned two replies available to the virtue epistemologists. The temporal objection does not leave the first reply available simply because it does not rely on any linguistic view regarding verb aspects. And it does not leave the second reply available because it does not rely on the controversial premise *. As we shall see now, the temporal objection rather relies on Helen Steward’s influential idea that performances have temporal parts (premise  of the temporal objection) while states do not (premise  of the temporal objection). ..

The Temporal Shape of States vs. Performances

That states – for example, the state of being liquid – should be distinguished from events – such as sunsets, wars and train arrivals – is widely recognised by metaphysicians of all stripes. The difficult question is what really distinguishes them. In her  book, Helen Steward provides a very convincing answer to this question. According to Steward, the ontological distinctions between states, events and processes depend on what she calls their temporal shapes. Temporal shapes capture the way states, events and processes respectively fill time. States – Steward says – do not fill time in the way events and processes fill it. The temporal shape of states is not characterised by a particularly long amount of time they would fill. It would be wrong to think that states are simply entities that last longer than events. There are long events – e.g. the whole building of the Eiffel tower – and short states – e.g. the state of being at  degrees, which your cooking water is for a brief instant. According to Steward, the temporal shape of states contrasts with the temporal shape of events as follows. 



Sosa seems, in fact, to adopt a view of this kind in his most recent works, e.g. Sosa : , n. . Chrisman also anticipates that virtue epistemologists might try to circumvent his objection by insisting that there are active states, that is, by rejecting premise *. This is why Chrisman spends some time in his  article arguing that there are, pace Boyle, no beliefs that are also active. If I understand Chrisman correctly, his main worry with Boyle’s view is conceptual. On the one hand, we cannot conceive of a state that is also dynamic (and not static). On the other, we cannot conceive of an active entity (action, performance, achievement, etc.) that is not also dynamic. Thus, we cannot conceive of a thing that is both a state and an exercise of agency. It is a conceptual truth that states cannot be active. Another advantage of my own way of arguing against the categorial claim is that it does not rely on the controversial assumption that grammar is a reliable guide for metaphysics.

Knowledge Is Extrinsically Apt Belief



States, unlike events, do not unfold; they do not occur. I suggest, then, that states are like continuants in lacking temporal parts. (Steward : )

Take the event consisting in the whole building of the Eiffel tower that lasted from  to . When the Parisians were only able to see the four pillars of the Eiffel tower, the event consisting in the whole building of the Eiffel tower had not taken place yet. What had already occurred at this moment – and at every other moment before  – was only a temporal part of the event consisting in the whole building of the Eiffel tower. Consider now the case of Gustave, a man who was happy from  to . At every moment during this period, e.g. in May , what had already occurred is not only a part of Gustave’s state of happiness. Unlike the building of the Eiffel tower, his state of happiness is already present in its entirety in May , and at every moment during these two years. His happiness is not something that unfolds during these two years in the way the construction of the Eiffel tower unfolds. In contrast to the building of the Eiffel tower, Gustave’s happiness does not unfold and has no temporal parts. As I already mentioned, the temporal objection relies heavily on Steward’s categorisation of entities in terms of temporal shapes. Obviously, premise  – namely, the claim that performances have temporal parts – and premise  – namely, the claim that states do not have temporal parts – are immediate expressions of her views. Additionally, and importantly, Steward’s categorisation in terms of temporal shapes also brings strong support to premise , namely, to the claim that beliefs are states. Take Gustave’s belief that he is a good scientist, a belief that he holds from 





Suppose that Gustave is increasingly happy during these two years of construction. You might wonder whether such an evolution of his happiness does not threaten the aforementioned claim according to which his state of happiness is present in its entirety at every moment during these two years. I do not think so, and here is why. For someone to be more and more happy is for them to be in a state of happiness at degree n at t and in a state of happiness at degree n+ at t. Even though including degrees of happiness, these two states are present in their entirety at t and t+ respectively. As for the whole evolution – the metaphysical entity that consists in the subject going from being happy at degree n to being happy at degree n+ – this is something that takes place in time, that has a beginning and an end is, therefore, an event and not a state. I am grateful to the editors of this volume for having pushed me to clarify this point. What occurs and might be said to unfold is the acquisition and then the loss of his happiness. But these – the acquisition and the loss of happiness – are events and not states. They should not, therefore, be confused with Gustave’s happiness. Since Steward’s categorisation primarily deals with states, events and processes and not with performances or actions, this might not appear immediately obvious. But performances or actions constitute a sub-kind of events (or processes, as Steward might prefer to say). If events (or processes) are essentially characterised by their possessing temporal parts, performances or actions also have temporal parts.



 

 to . How does this belief fill this period of time? Unlike a train trip or the building of the Eiffel tower, Gustave’s belief has no temporal parts. At every moment during these two years, the belief that he is a good scientist is present in its entirety. Beliefs do not fill time in the way events or processes fill it. Beliefs fill time in the way states fill it. That is, beliefs have the temporal shape that essentially characterises states. Steward’s classification of fundamental metaphysical entities in terms of temporal shapes provides a helpful metaphysical vindication of the first premise of the temporal objection. Let me, at this point, recap the main points that I have tried to make in this chapter’s first part. There is a very strong reason to doubt the virtue epistemological categorial claim according to which a belief that constitutes knowledge is a performance. I agree with Chrisman on this. Where our views differ is on what is problematic about the categorial claim. Contrary to Chrisman, I do not think that what is mainly problematic about the categorical claim is that it attributes a form of agentivity to beliefs that is not characteristic of states. The problem faced by the categorial claim has rather to do with states’ temporal shape. States are essentially characterised by a different temporal shape than performances. Given that – as I have argued – beliefs are states, they cannot also be performances.

. The Amendment: Knowledge Is Extrinsically Apt Belief ..

Aptness: Intrinsic and Extrinsic

In virtue epistemologists’ theories, the virtue epistemological account of knowledge is very often complemented with the categorial claim – viz. the view that a belief that constitutes knowledge is a performance or an achievement – as if the latter was a natural or necessary corollary of the former. This is illustrated by the quotations from Sosa provided at the very beginning of this chapter. Nevertheless, if the temporal objection presented in Section . withstands scrutiny, virtue epistemologists are, in fact, forced to abandon the categorial claim. How bad is this? The purpose of the second part of this 



The aforementioned criterion of the verb’s aspect is linked to Steward’s way of distinguishing between states and events in terms of temporal shapes. Indeed, the reason why verbs denoting states do not, most often, admit the continuous is that states ‘do not take up times at all; they have not temporal parts’ (Steward : ). But this is also the case with the virtue epistemological approach defended by John Greco. See Greco , .

Knowledge Is Extrinsically Apt Belief



chapter (Sections . – ..) is to make clear that this is not as costly an amendment as one might initially think. Virtue epistemologists can happily maintain their account of knowledge – namely, the view that knowledge is apt, i.e. true because competent, belief – while rejecting the categorial claim. In order to understand why the categorial claim is not, in fact, a necessary corollary of the virtue epistemological account of knowledge, we need to start with two preparatory remarks regarding normative properties in general. First, philosophers usually admit that whether an entity possesses a given normative property depends on whether it possesses certain other properties. A dish can be disgusting in virtue of its colour or its consistency; your driving behaviour can be forbidden in virtue of the speed you reach, etc. In these examples, the properties in virtue of which the entity possesses normative properties are descriptive (or natural). But it is also possible for an entity to possess a normative property in virtue of its possessing another normative property. Think of someone whose smartness depends on their being quick-witted. Second, it is commonplace to distinguish between two kinds of normative properties: intrinsic and extrinsic. An intrinsic normative property is a normative property that an entity possesses in virtue of its intrinsic properties only.

A beautiful chair is very often (even if, arguably, not always) beautiful because of the intrinsic properties of the chair (its shape, for instance). The normative property of being beautiful is, thus, an intrinsic normative property of the chair. 





See Chrisman :  for a more pessimistic conclusion. According to him, the rejection of the categorial claim forces virtue epistemologists to focus on the formation and maintenance of beliefs rather than on the state of believing. Consequently, it seems that the only viable virtue epistemological account is, on Chrisman’s view, an account of the formation or maintenance of knowledge. In other words, according to Chrisman the virtue epistemological account of the state of knowing falls with the categorial claim. This is precisely what Section . denies. Sometimes the expression ‘normative property’ is used to denote what I prefer to call ‘deontic normative properties’, namely, properties such as being obligatory, forbidden, permissible, etc. In this chapter, the expression ‘normative properties’ refers to a general class of which deontic normative properties are members. The relation between a normative property and the properties in virtue of which a thing possesses that normative property is very often considered to be a relation of supervenience. See Lemos ; Sosa ; Mulligan . This also seems to be Moore’s view (, ) even though Moore never makes use of the term ‘supervenience’. The remaining question – which still divides philosophers – is whether the supervenience relation is a reductive relation, that is, whether normative properties can be reduced to non-normative properties. I leave this problem aside here.



  An extrinsic normative property is a normative property that an entity does not possess in virtue of its intrinsic properties only but also in virtue of its extrinsic (or relational) properties.

Instrumental goodness – i.e. the goodness that an entity possesses in virtue of the goodness of its effects – is the paradigmatic example of an extrinsic normative property. It is a normative property that an entity possesses in virtue of its causal relation with some other good entities. With this in mind, let us go back to aptness. As Sosa often says, aptness is a normative property. In this respect, aptness contrasts with descriptive (or natural) properties, e.g. the property of weighing  grams, of being brick-red, of being triangular. Such properties do not, by themselves, make a normative difference to the way we evaluate the entities that possess them. Hence, what is true of normative properties in general is true of aptness. First, qua normative property, aptness is a property that an entity possesses in virtue of some other properties. In the case of aptness, the other properties in virtue of which an entity is an apt entity are the properties of being successful and of being competent. Second, qua normative property, it is possible to conceive of two kinds of aptness: intrinsic and extrinsic aptness. An entity E is intrinsically apt iff its being successful and its being competent – that is, the properties on which its being apt depend – are all intrinsic properties of E. An entity E is extrinsically apt iff its being successful and its being competent – that is, the properties on which its being apt depend – are not all intrinsic properties of E.

All this results from what has previously been said regarding intrinsic and extrinsic normative properties in general. .. The Specificity of Intrinsic Aptness Intrinsic aptness possesses a characteristic that other intrinsic normative properties do not. Suppose pleasantness is an intrinsic normative property. It is an intrinsic normative property that both states and performances are 

To be precise, I should say that aptness is a property that an entity possesses in virtue of its being successful because competent, this latter complex property being itself a property that an entity possesses only if it is, on one hand, successful and, on the other, competent. Finally, aptness is, as I state in the body of the text, a property that an entity possesses in virtue of its being successful and its being competent. And these are, in fact, the properties that matter for my purpose in this chapter.

Knowledge Is Extrinsically Apt Belief



susceptible to example. There are pleasant states and pleasant actions. Nothing like this with intrinsic aptness. Intrinsic aptness is a normative property that only actions, performances and achievements – what I have previously called ‘active entities’ – can exemplify. Or so I shall argue. Whether an entity is intrinsically apt depends both on its being intrinsically successful and on its being intrinsically competent. There are, thus, two possible explanations as to why intrinsic aptness is a normative property that only active entities can exemplify. Either this is because the property of being intrinsically success is itself a normative property that only active entities can exemplify. Or this is because the property of being intrinsically competent is itself a normative property that only active entities can exemplify. Accordingly, I shall call the first plausible explanation ‘the Success-Explanation’ and the second ‘the Competence-Explanation’. I consider each of them in turn below. According to the Success-Explanation, the reason intrinsic aptness would be a property that only active entities are susceptible to exemplify would be as follows. a.

the exemplification of intrinsic aptness by an entity E depends on the exemplification by E of the intrinsic property of being successful; b. only active entities can be intrinsically successful. But it is not obvious that statement b. is right. Indeed, one might think that certain events and processes – things like the processes of photosynthesis or food digestion – can be intrinsically successful even though it seems odd to claim that food digestion or the process of photosynthesis are performances of people or plants respectively. To see this, just note that an intrinsically successful entity might simply be one that achieves its constitutive or functional aim. Under this understanding, when a plant converts light into energy, the process at work is an intrinsically successful process of photosynthesis. But, as just said, it is not obvious that such a process is an active entity. Thus, the best explanation for why only active entities can be intrinsically apt is probably not the one provided by the SuccessExplanation. The best explanation is thus not that only active entities can be intrinsically successful. 



For instance, this distinguishes intrinsic aptness from intrinsic value-properties such as being beautiful. Events/processes that are not also active – for instance sunsets – cannot be intrinsically apt but they can be intrinsically beautiful. What I state here is that events and processes that are not also active can be intrinsically successful. A further question is whether states can be intrinsically successful. Sosa thinks that this is the case. See Sosa : , n. . Chrisman objects to this idea that states cannot be intrinsically successful



 

Alternatively, as I just said, one could argue that the reason why only active entities can be intrinsically apt is that only active entities can be intrinsically competent. The Competence-Explanation is, in my view, the right explanation. Consider things like archers’ shots, chirurgical gestures, mental calculi, etc. They can be conceived as intrinsically competent only if they are identical to (or constitute) actualisations of certain competences. Non-active events and states – things like sunsets and health – cannot be intrinsically competent. And this is because they are never identical to actualisations of certain competences. Only actualisations of competences can be conceived as intrinsically competent. This, I believe, has to do with the notion of competence itself. In the same way that the surface of my desk cannot be popular because the notion of being popular does not apply to objects’ surfaces, only actualisations of competences can be intrinsically competent because being intrinsically competent only apply to actualisations of competences. Now, since any actualisation of a competence is necessarily an active entity, only active entities can be intrinsically competent. And this is exactly what the Competence-Explanation predicts. However, non-active events and states – things like sunsets and health – can be extrinsically competent, that is, competent in virtue of the relation they have to a particular actualisation of a competence. For instance, your being healthy might well be the causal result of some of your actions (eating a balanced diet, working out etc.) all of which are actualisations of some of your competences. But the state of being healthy is not, in itself, the actualisation of one of your competences. It is only the result of your competence to stay fit. The existence of extrinsically competent

 



because states are by definition a-telic. It is, according to Chrisman (: ) ‘highly unclear what it would mean for some state to achieve an end’. Because I consider it even more obvious that states cannot be intrinsically competent, I do not discuss the suggestion that states cannot be intrinsically successful further in this chapter. Note, however, that if Chrisman is right, we have one reason more to conclude, as I do below, that the kind of aptness that beliefs can exemplify is extrinsic aptness. See Ho : , for a similar view. You might object that persons, animals and perhaps plants can be ‘competent’ even though persons, animals and plants are not active entities. They are obviously not performances or achievements. When one says of a person and even perhaps – even this is a bit far-fetched – of a plant that she/it is competent, the use of ‘competent’ is dispositional. What we mean is that the person has certain (beneficial) capacities, that can be actualised on some occasions. But clearly Sosa does not make a dispositional use of the term ‘competent’ when he states that an instance of knowledge consists in a successful (true) because competent belief. It does not want to say that the belief has certain capacities. For this reason, I do not need to discuss the dispositional use of the term ‘competent’ in this chapter. I am grateful to Ch. Kelp for having pressed me to clarify this point. As J. Greco mentioned to me, it seems, however, right to say that being healthy is an achievement. I think this comes from the fact that the term ‘achievement’ is sometimes used to denote the state of affairs that results from a certain (admirable) activity or performance and not the activity or performance itself (as, for instance, when I contemplate the birthday cake that I bake for my

Knowledge Is Extrinsically Apt Belief



states – even though there are no intrinsically competent states – constitutes the way out for virtue epistemologists. As we shall see, the existence of extrinsically competent states allows virtue epistemologists to preserve the virtue epistemological account of knowledge while rejecting the categorial claim (as the temporal objection requires). .. The Pitfall to Avoid Section .. was devoted to showing that aptness comes in two varieties: intrinsic and extrinsic. Ignore this for a moment and suppose that the only conceivable kind of aptness is intrinsic aptness. Under this hypothesis, the categorial claim – namely, the view that a belief that constitutes knowledge is a performance or an achievement – really is a necessary corollary of the virtue epistemological account of knowledge – namely, of the view that an instance of knowledge consists in an apt belief. Under this hypothesis, indeed, everything that is apt is intrinsically apt and, as I argued in section .., everything that is intrinsically apt is an active entity (an action, a performance, an achievement, etc.). One possible explanation as to why virtue epistemologists take the risk of complementing their account of knowledge with the much more controversial categorial claim might thus be as follows. They have not paid sufficient attention to the fact that aptness, like many other normative properties, comes in two varieties: intrinsic and extrinsic. Once it is acknowledged that aptness can be an extrinsic normative property of a belief that constitutes knowledge, the necessity of supplementing the virtue epistemological account of knowledge with the categorial claim vanishes. More precisely, in order not to have to claim that a belief that constitutes knowledge is a performance or an achievement, virtue epistemologists should adopt the metaphysical picture in Figure ., in which the intrinsic properties are represented with full lines and the extrinsic properties with dot lines. This metaphysical diagram in Figure . does not need the categorial claim to be true. Nothing requires that a belief that constitutes knowledge be a performance or an achievement. Indeed, what is intrinsically



daughter and think to myself – most often ironically – ‘what an achievement!’). Being healthy really is an achievement according to this specific way of using the term ‘achievement’. Thanks to the audience of the nd Leuven Epistemology Conference (– December ) for having helped me to understand this better.



  Knowledge

Extrinsically competent

+

Intrinsically successful

Intrinsically competent

Actualisation of a cognitive competence, e.g. inferring that p, perceiving that p, etc.

Figure .

Belief Because

Metaphysical picture of intrinsic and extrinsic properties

competent – and is, therefore, as I have argued, necessarily an active entity – is not the belief that constitutes knowledge. In this visualisation, the belief that constitutes knowledge is rather extrinsically competent. It is competent in virtue of the relation (represented by the bold arrow) that holds between it and the intrinsically competent actualisation of a cognitive competence. Only the actualisation of the cognitive competence is intrinsically competent here. While it is, as we previously saw, impossible to claim that beliefs are intrinsically competent (since beliefs are not active entities), there is nothing problematic in attributing the property of being intrinsically competent to the actualisation of a competence. Indeed, when someone actualises or exercises one of their competences, they really are achieving something: they infer, they perceive etc. The actualisation of a cognitive competence is an active entity. To recap, this metaphysical picture has the crucial advantage of exempting the virtue epistemological account of knowledge from the categorial claim and, in so doing, of avoiding the temporal objection. How does it achieve this? By making clear that a belief that constitutes knowledge is only extrinsically competent, and therefore only extrinsically apt. According to



In this rendering, the property of being successful is represented as an intrinsic property of the belief. As I explain in n. , I do not intend to settle the question of whether states can exemplify the intrinsic property of being successful. Indeed, I find the fact that states cannot be intrinsically competent more persuasive. Furthermore, the fact that states cannot be intrinsically competent is sufficient to conclude that the kind of aptness exemplified by beliefs that constitute knowledge is extrinsic aptness.

Knowledge Is Extrinsically Apt Belief



this approach, a metaphysically sounder way of formulating the virtue epistemological account of knowledge is as follows: An instance of knowledge consists in a successful (true) because extrinsically competent belief, that is, in an extrinsically apt belief.

. Conclusion I hope to have achieved two things in this chapter. First, I hope to have added fresh support to Chrisman’s case for rejecting the categorial claim. Virtue epistemologists cannot maintain the claim that a belief that constitutes knowledge is a performance. And the fundamental reason for this is that an entity that essentially has no temporal parts cannot be identified with an entity that essentially has temporal parts. Second, I hope to have shown that the rejection of the categorical claim is not as costly as virtue epistemologists might initially think. Contrary to what Chrisman seems sometimes to claim, it does not require that virtue epistemologists abandon the idea of analysing knowledge in terms of apt belief. All that is needed is that they make clear that the aptness in question is extrinsic. Put briefly, knowledge is not a performance but it might well be an extrinsically apt belief.

Acknowledgements I am very grateful to the editors of this volume, to Matthew Chrisman and to Robin McKenna for their insightful comments on earlier versions of this chapter. A second debt of gratitude is to the audiences of the Quod Libeta Conference at the University of Geneva ( November ), of the nd Leuven Epistemology Conference (– December ), and of the Research Colloquium organised by Filippo Ferrari at the University of Bonn ( February ). This research has been kindly supported by the Swiss National Foundation (grant number: PPP_).  

See n. . One remaining question is whether the suggested metaphysical diagram is able to secure an account of the value of knowledge. Aptness is, in this metaphysical diagram, an extrinsic property. The specific value that accrues to knowledge in virtue of being an apt belief is, accordingly, an extrinsic value. More precisely, the specific value that accrues to knowledge in virtue of being an apt belief belongs to a pretty uncommon kind of extrinsic value. The additional value of apt beliefs is one that they derive from the value of their causes. (Note that in the most common kind of extrinsic value – i.e. the instrumental kind – the relation is the reverse.) The suggestion that the specific value that accrues to knowledge belongs to this unusual kind of extrinsic value is not unproblematic. See Meylan  for a discussion of this issue.



  R E F E R EN C E S

Boyle, M. . ‘Active Belief’, Canadian Journal of Philosophy () (Iss. Supp.): –. Chrisman, M. . ‘The Normative Evaluation of Belief and the Aspectual Classification of Belief and Knowledge Attributions’, Journal of Philosophy (): –. . ‘Epistemic Normativity and Cognitive Agency’, Noûs (): –. . ‘Performance Normativity and Here-and-Now Doxastic Agency’, Synthese: –. Greco, J. . Achieving Knowledge. Cambridge, UK: Cambridge University Press. . ‘A (Different) Virtue Epistemology’, Philosophy and Phenomenological Research (): –. Ho, T.-H. . ‘Epistemic Normativity as Performance Normativity’, Theoria : –. Kenny, A. . Action, Emotion, and Will. London: Routledge and Kegan Paul. Lemos, N. . Intrinsic Value: Concept and Warrant. Cambridge, UK: Cambridge University Press. Meylan, A. . ‘The Value Problem of Knowledge: An Axiological Diagnosis of the Credit Solution’, Res Philosophica (): –. Moore, G. E. . Principia Ethica, revised edition Baldwin, T. (ed.). Cambridge, UK: Cambridge University Press. . Ethics. London: Williams & Norgate. Mourelatos, A. . ‘Events, Processes, and States’, Linguistics and Philosophy  (): –. Mulligan, K. . ‘Values’, in The Routledge Companion to Metaphysics, Poidevin R., Simons P., McGonigal A. and Cameron R. (eds). London: Routledge, –. Sosa, E. . ‘Nature Unmirrored, Epistemology Naturalized’, in Knowledge in Perspective: Selected Essays in Epistemology. Cambridge, UK: Cambridge University Press, –. . A Virtue Epistemology: Apt Belief and Reflective Knowledge, vol. . Oxford: Oxford University Press. . ‘Knowing Full Well: The Normativity of Beliefs as Performances’, Philosophical Studies (): –. . Judgment and Agency. Oxford: Oxford University Press. Steward, H. . The Ontology of Mind: Events, Processes, and States. Oxford: Oxford University Press. Vendler, Z. . ‘Verbs and Times’, The Philosophical Review (): –.

 

Explaining Knowledge Alan Millar

. Introduction This discussion is exclusively about propositional knowledge – knowledge that such-and-such. Two ideas are central to the treatment of such knowledge in mainstream epistemology at least since the s. The first is that the central aim is to provide a highly general account, and preferably one that is expressible as a reductive analysis in that it would explain what knowledge is in terms that do not implicate the concept of knowledge. At least part of the motivation for this approach is the assumption that anything short of a reductive account would leave knowledge itself unexplained or inadequately explained. Accordingly, reductive projects that make justified belief central to the analysis aim to explicate justified belief without invoking the concept of knowledge. The other idea is that in our study of particular kinds of knowledge, for instance perceptual knowledge and knowledge from testimony, we should be guided by the general conception of what knowledge is. I take this approach to epistemology to be fundamentally mistaken both in aiming for a reductive account and in aiming to explain particular kinds of knowledge in terms of a highly general account of knowledge. Here I provide glimpses of an alternative perspective on which knowledge is explanatorily prior to justified belief and the focus is on giving substantive accounts of various kinds knowledge. Knowledge in general is to be illuminated in terms of those substantive accounts. Of course, something unites the kinds. Knowledge is cognitive contact with (grasp of ) a truth or fact. What this amounts to emerges only by bringing to light the diverse modes of contact that characterize different kinds of knowledge.  

For a similar attitude on the issue of generality, see Roessler  and . I take this general approach in Millar . Hilary Putnam (: ) speaks of cognitive contact with the world, Linda Zagzebski (: ) of cognitive contact with reality, John McDowell (: ) of cognitive purchase on an objective fact, and Charles Taylor (: sec. ) of contact with the real in knowledge.





 

To account for knowledge of a certain kind in keeping with the general approach I have described we must explain how subjects who have knowledge of that kind are in cognitive contact with a fact. In Sections . and . I consider simple perceptual knowledge, focusing on a case in which you come to know – concerning objects at which you are looking – that they are of a certain kind from the way they look. In Section . I discuss what I call knowledge from a perceived indicator, focusing on cases such as seeing black streaks on a road and from their appearance coming to know that a vehicle has braked hard there. I show that such knowledge can be more akin to simple perceptual knowledge than one might initially think. When it is, it deserves to be regarded as a species of perceptual knowledge. My treatment of knowledge of both of these kinds accords a central role to perceptual-recognitional abilities – general abilities to recognize perceived things as being some way from the way they appear relative to some sense-modality. My general approach to epistemology is much indebted to the work of John McDowell (e.g. , , , ). The approach has a kinship with widely known varieties of virtue epistemology, notably those advanced by Ernest Sosa (e.g. in a, b, , , , ) and John Greco (e.g. in , , , ), but whereas the latter explicate knowledge in terms of abilities, competences, or dispositions the outputs of which are beliefs that might or might not be true, it is crucial for the position set out here that recognitional abilities are conceived as abilities the exercise of which is nothing less than the acquisition of knowledge. The broad perspective within which I am working is ‘knowledge first’ in that it takes knowledge to be explanatorily prior to justified belief. Part of my aim is to show that epistemology from such a perspective can be informative and explanatory.

. Perceptual Knowledge Imagine you want to buy some fruit. Shopping in a supermarket, in circumstances that are in no way unusual, you see some loose pears in a display box. You see, and in that way know, that they are pears. I claim 



I speak of general abilities to indicate that the abilities in question might be exercised on indefinitely many occasions. Such abilities contrast with specific abilities to do something on a particular occasion. See Mele , Whittle , and Maier . For the view that seeing-that P entails knowing that P, see Warnock /: section II, Sellars : , Dretske : –, Armstrong : , Williamson , Stroud , , and French .

Explaining Knowledge



that we best understand what that knowledge amounts to by thinking of it as acquired by the exercise of a general ability to recognize pears as pears visually. Such an ability is an ability to tell, so come to know, of pears that they are pears, from the way they look. Exercising this ability in the situation described is coming to know, concerning the objects in the box, that they are pears. The upshot is that you see that, and in that way know that, they are pears. This tells us how it is that you are in cognitive contact with the fact that these objects are pears. Cognitive contact is not invoked here to define or explicate what it is to see the pears to be pears. We already have a working understanding of what it is to do that. In the light of that understanding we learn something about what cognitive contact is by learning that seeing that the objects are pears is exemplary of a way in which you can be in cognitive contact with a fact. That is merely a step on the way to providing a deeper general understanding of visual-perceptual knowledge – of that kind of cognitive contact – in terms of a developed account of recognitional abilities and their exercise. (For more on the metaphor of cognitive contact, see Section ..) The explanation of your knowledge in this particular case is informative because you might have acquired the knowledge by reading a label on the box, or by hearing from an assistant, or by some other means. It adds to our understanding to know that you gained the knowledge in the way you did, rather than by other means. Note too that the explanation is neither circular nor ad hoc. It is not circular because the general ability to tell – come to know – concerning pears, that they are pears from the way they look is constitutively independent of the knowledge that is acquired in this case. It is not ad hoc because the general ability is not postulated to account only for the case in hand. It can be exercised in diverse situations. Moreover, we know how to assist young children in acquiring this ability. Roughly speaking, if their language is English, we apply the term ‘pear’ in their presence, sometimes drawing their attention to those applications, and to differences between pears and other fruits. We correct their applications when they go wrong and encourage them when they are right. We also know how to test whether they have acquired the general ability. They must show that, by and large, when they judge things to be pears they do so correctly, so that mistakes are rare and usually explicable as due to distraction, carelessness, or some cognitive impairment. They must not be overly cautious in making such judgements – too often withholding 

See, in particular, Millar , a, , , a, b, , and my contribution (chs. –) to Pritchard, Millar, and Haddock .



 

judgement in conditions suitable for recognition. Passing such tests is a highly reliable indicator that a person has the ability. So when we explain your coming to know of the pears that they are pears we are not postulating a mysterious power, but rather an ability of which we have some understanding, which most of us have, and know we have, and expect most others to have after a suitable learning process. It should not be in dispute that we have abilities such as that described or that there are analogous abilities corresponding to other sensemodalities. There are abilities to recognize a singing voice as that of Billie Holliday from the way it sounds and abilities to recognize light switches as light switches from the way they feel. There are abilities visually to recognize people as specified individuals. There are abilities to recognize features as belonging to feature-kinds. We might recognize the colour of a garment as green, or the texture of a scarf as that of silk. Since perceptual knowledge is to be understood in terms of such abilities, a central task of epistemology is to shed further light on them. It might be suggested that perceptual-recognitional abilities bear on how we acquire knowledge rather than on what it is to have the knowledge that is acquired by their exercise (cf. Hyman : ch. .). You come to know concerning the things in the box that they are pears in the manner described, but, as I noted earlier, you might have come to know this by other means. It is not possible, then, to give a wholly general account of what it is to know concerning the things in the box that they are pears in terms of the exercise of your visual-recognitional ability but, though clearly true, this is no objection to my approach. Knowledge in a given case is individuated not just by the content of the truth known but by the way in which it is known. To explain adequately your knowledge concerning the things in the box, we have to say how you are in cognitive contact with the fact that they are pears. In this case your mode of contact is recognition of the things as pears from the way they look. It is constitutive of that recognition that it is the exercise of the relevant general recognitional ability. So, to characterize your knowledge in the given case we must advert to that ability. This does not tell us in wholly general terms what it is to know concerning the things in the box that they are pears. It is not meant to, but it is, nonetheless, relevant to a more general understanding because it tells us that cognitive contact with the fact in question is something that can be effected in the manner specified. We shed light on knowledge in general in terms of the form of contact with a fact that is exemplified by the case. Note too that our characterization of the cognitive contact exemplified by the case is explanatory of what is widely regarded as an implication of

Explaining Knowledge



knowing – that it is no accident that your belief on the matter is true given how the belief was acquired or sustained. It is no accident that you form a true belief to the effect that these objects are pears, given that you have a general ability to effect such recognition, and that you are attending to the pears in conditions conducive to recognizing them to be pears from the way they look. In the next section I highlight and expand upon a number of key considerations relating to perceptual-recognitional abilities of the sort I have illustrated.

.

The Metaphysical Shape of Perceptual-Recognitional Abilities

A. In acquiring perceptual knowledge you recognize a thing as being some way from the way it appears relative to some sense-modality. When you recognize a thing to be some way you tell, that is, come to know, that the thing perceived is that way from its appearance rather on the basis of an assumption as to its appearance. Of course, if suitably developed subjects were looking at a horse and recognizing it to be a horse, they would very likely also take in at the level of belief or judgement that the thing they see has the look of a horse, but it is not constitutive of one’s simply recognizing a horse as a horse that one does so. Whether one does so depends on whether one has the conceptual resources to grasp facts pertaining to appearances. Children exercise such an ability at an early learning stage when they have not yet mastered the notion of a thing’s having the look of a horse. B. Appearances – ways things appear – are ways things look, or feel, or sound, or taste, or smell. If you can recognize things as being some way, there must be an appearance that is distinctive of being that way. An appearance is distinctive of being of some kind, or having a certain property, or being some specified individual or feature if and only if, by and large, a thing that had that appearance would be of that kind, or have that property, or be the specified individual or feature. The ability to recognize things perceptually as being of some kind, say, is an ability to tell of things of that kind that have an appearance that is distinctive of being of that kind that they are of the kind. To exercise the ability the environment must be favourable to such exercise in that, at least there, the appearance in 

I have sometimes formulated the requirement by saying that there must be an appearance that is nearly enough distinctive, but it is more in keeping with our ordinary ways of thinking to allow that an appearance can be distinctive of a kind even if a few things not of the kind have the appearance.



 

question is distinctive of being of the kind. Examples of the fake barn type draw attention to the possibility that an appearance that is distinctive of being of some kind in one region might not be distinctive of being of that kind in another region, where too many things that are not of the kind also have the appearance and are readily perceivable. The distinctiveness requirement explains why you could not recognize a pear as a pear from the way it looks, and in that way know that it is a pear, in a circumstance in which pears are interspersed with ringers for pears that you might easily come across. In such a case the objects you see have a visual appearance – a look – that is characteristic of pears in that, by and large, things that are pears would have that appearance, but in the envisaged fake pear set-up this appearance would not be distinctive of pears, since too many things that have it are not pears. C. Notwithstanding that we recognize things as being some way from the way they appear, recognition can have rich content (Millar ). This has been implicit throughout the discussion thus far. We recognize people as specified individuals, not just as objects possessing superficial properties such as those of shape and colour. We recognize objects as being, for instance, trees, houses, birds, or cars. Recognition is no less phenomenologically immediate for having rich content. It simply strikes you that a certain colleague is present. You know from what you see that she is there. Even if some salient feature contributes to eliciting recognition, it does so in the context of the overall Gestalt presented. Analogous points hold for recognizing things as being of some kind. From what you see as you look at the pears it strikes you that these things are pears. You see that they are pears in an act that is both perceptual, in that it involves you seeing the pears, and intellectual in that it is a matter of bringing these pears under a concept – applying to them the concept of pears. D. As with other abilities, recognitional abilities are exercised only in episodes in which a subject does, or is doing, the thing that the ability is an ability to do. You don’t exercise the ability to read English unless you are reading English. You don’t exercise the ability to swim unless you are swimming. Of course, there are difficult feats, hard performances, that one pulls off only from time to time, like throwing a dart into the bull’s-eye. Experts are unusually good at doing such a things, but if we say, without qualification, that they have the ability to throw the dart into the bull’seye, we obscure an important difference between the ability that accounts for the expert’s being good at hitting bull’s-eyes and abilities to bring off 

The perspective outlined in this paragraph is further developed in Millar b and , ch. .

Explaining Knowledge



easy performances, such as speaking English or swimming. In cases of failure to bring off an easy performance there will nearly always, perhaps always, be an explanation of the failure. The failure might be due, for instance, to a temporary impairment on the subject’s part or environmental circumstances that are distracting or otherwise disabling. The same does not apply to throwing a dart into the bull’s-eye. Even experts pull off this feat only occasionally and no explanation is called for of their not doing so on some occasion when they try to, since it is not to be expected that they succeed whenever they try, in the absence of impairment or some other impediment. That is a clue to the fact that in this case the ability the expert darts player has is a success-rate ability (Millar b). It is an ability to achieve an impressive level of success (getting the dart into the bull’s-eye) in a sequence of trials. Since that is what the ability is an ability to do, it is exercised only when the agent achieves such a level of success over a sequence of trials. It is not exercised in a single successful throw. Since recognizing pears as pears from the way they look is what the ability to recognize pears as pears from the way they look is an ability to do, the ability is exercised only in such acts of recognition. Clearly, we are fallible with respect to our abilities in that we are liable on occasion not to do the thing that the ability is an ability to do, even though we act in a manner that is directed at doing so. Though you have a general ability to walk, if you had been drugged you might fail to walk on some occasion when you try to. Notwithstanding your ability to recognize pears as pears visually you would fail to recognize pears as pears if you were in a perceptual Gettier set-up – one in which pears are at the location where you take pears to be but you are looking, not at these pears, but at some visual depiction of pears that makes it look as if pears are at the location in question. E. The basic way to characterize an ability to recognize things as being some way from the way they appear is to say that it is an ability to tell of things that are that way that they are that way from the way they appear. Some recognitional abilities are more enabling than that, for they consist in an ability to tell of things one perceives whether (whether or not) they are that way from their appearance. You might, for instance, have a general ability to tell of objects you see whether they are swallows, going by the 

 

‘Easy’ and ‘hard’ are semi-technical terms here. It is not implied that in the ordinary sense it is always easy to bring off an easy performance. The point is that once you acquire the ability failures to exercise it are almost always due to some impediment. An impressive success-rate might be any rate within a range achievable only by the very skillful. That said, a single successful throw might be explicable in terms of the thrower’s exercising an ability to perform a routine that raises the chance of success.



 

way they look. That is possible because it is possible to tell of things that have the look of swallows that they are swallows and tell of things that lack that look that they are not swallows. If you have this ability there is something that you are good at doing. Your being so entails that, with a high degree of reliability, if you were to make judgements as to whether things you see are swallows, going by the way they look, you would make true judgements. That is to say, you would satisfy two reliability conditions: () by and large, if you were to judge a thing to be a swallow from the way it looks you would be right (a type- reliability condition); () by and large, if you were to judge a thing not to be a swallow from the way it looks you would be right (a type- reliability condition). Satisfying conditions () and () is compatible with being excessively cautious in the judgements you make as to whether objects you see are swallows. If you were excessively cautious you would hardly count as being good in the relevant way. So we need to introduce a certain preparedness condition to the effect that in circumstances that are favourable to telling whether an object is a swallow going by the way it looks, and in which you aspire so to tell, by and large, you would make a judgement as to whether it is a swallow (that is, you would either judge it to be a swallow or judge it not to be a swallow). F. The conditions spelled out in the previous paragraph are plausible because swallows have an appearance that is not only distinctive of being a swallow but also characteristic of being a swallow in that, by and large, if a thing were a swallow it would have that appearance. Because this is so it is possible to tell whether a thing is a swallow from the way it looks. For one can tell of swallows that they are swallows from their having the distinctive appearance of a swallow, and tell of things that are not swallows that they are not swallows from their lacking that appearance. An appearance can be distinctive of being some way without being characteristic of being that way. The visual appearance of a smart phone is distinctive, but not characteristic, of being a phone. (Smart phones don’t look much like, for instance, typical domestic landline phones of the s.) The visual appearance of a garden variety rose is distinctive, but not characteristic, of being a rose. (Dog roses lack the luxuriant blooms of garden variety roses.) Clearly, we should accommodate the fact that people could 

It is a sufficient condition of knowing whether a thing one perceives is an F that one knows that it is an F. To know that visually you must have at least a general ability to tell of Fs that they are Fs from the way they look. It is not a necessary condition of having such an ability that you have a general ability to tell of things you perceive whether they are Fs. See further paragraph F below.

Explaining Knowledge



recognize smart phones as phones even if they were familiar only with smart phones, and lacked a general ability to tell by looking at a device whether it is a phone. Similar considerations obviously apply to the ability to recognize garden variety roses as roses. G. As we have seen, there can be significant differences in appearance among things that belong to a kind. Accordingly, it can be that some abilities have, as I shall put it, a limited spread. The spread of an ability to recognize things as being of some kind is the range of things of the kind that one is able to recognize as being of that kind. This range might not comprise the full range of things that are of the kind. If the spread of your ability to recognize roses as roses from the way they look is limited to garden variety roses then, if you were presented with a dog rose, you might not tell (come to know) that it is a rose by looking at it. And, of course, it is not possible to tell from a thing’s lacking the appearance of a garden variety rose, that it is not a rose. So, if your ability is thus limited to garden variety roses, it is not a general ability to tell of things you see whether they are roses, from the way they look. Still, there is a preparedness condition that you would satisfy if you had the ability: in circumstances favourable to telling of garden variety roses that they are roses, and in which you aspire so to tell, by and large, you would judge such roses to be roses, going by the way they look. You would not fail to meet this condition simply on account of withholding judgement as to whether things you see are roses when they lack the appearance characteristic of garden variety roses. H. Even if you are good at telling of garden variety roses that they are roses, but poor at telling of dog roses that they are roses, it could be that your circumstances are such that rarely if ever would you come across a rose that lacked the appearance characteristic of garden variety roses, and be prompted to make a judgement as to whether it is a rose. In those circumstances you would rarely if ever judge a rose not to be a rose, and so could satisfy the relevant type- reliability condition. Indeed, since in those circumstances lacking the appearance distinctive of garden variety roses is a reliable indicator of not being a rose, it could even be that you count as having an ability to tell whether a thing is a rose, from the way it looks. I. Our ascriptions of abilities to people underdetermine the character of those abilities. People of widely diverse abilities in playing tennis may all be said to be able to play tennis. People may be said to be able read even though they differ among themselves significantly with respect to speed and comprehension. The considerations advanced above suggest that we should countenance the possibility that among those who count as being able to recognize things that are some way as being that way, there can be



 

differences with respect to the level of their ability that are due variations in the spread encompassed by the ability.

. Knowing from a Perceived Indicator I turn now to another kind of knowledge that involves perception and responsiveness to distinctive ways of appearing. It is a kind of knowledge from a perceived indicator. I focus on a case in which you tell, so come to know, on seeing black streaks on a road that a vehicle has braked hard there. It is widely held that all knowledge, or at least all empirical knowledge, is evidence-based. Though I shall not argue the point in detail here, it seems to me that this view masks epistemologically significant differences. There is a clear sense in which knowledge can be based on evidence. In the clear sense your knowledge that P is evidence-based if and only if there is evidence that P (distinct from the fact that P) and you know that P on the basis of this evidence. Your knowledge that P is based on evidence only if the evidence supplies you with a reason to believe that P that is adequate for knowledge, and you believe that P for that reason. Perceptually recognizing pears as pears is not an acquisition of evidence-based knowledge in this sense. In recognizing the pears to be pears you simply see that they are. You might also have evidence that they are pears constituted by the fact that they have the look of pears, but that would be a bonus, albeit one that could have an important knowledge-sustaining role. I think we need to take seriously the idea that in the indicator case in hand, your knowledge that a vehicle has braked hard is a hybrid of perceptual and evidence-based knowledge. It is constitutively evidence-based because you tell that a vehicle has braked hard on the basis of the fact that there are black streaks of a certain appearance on the road. That is essential to its being the knowledge that it is, but it is also perceptual knowledge because by perception not only do you see the streaks, and see that they look a certain way, you also see that they have a certain significance. I elaborate on this in what follows. The presence of the streaks at a place on the road constitutes evidence that a vehicle has braked hard at that place in virtue of indicating that this is so. Indication is factive: the marks indicate that a vehicle has braked hard only if a vehicle has braked hard. There is also a modal element to 

I have discussed this kind of knowledge in earlier work. See Millar a, ,  and my contribution to Pritchard, Millar, and Haddock .

Explaining Knowledge



indication: there would be these marks only if a vehicle had braked hard. It is because the marks indicate that a vehicle has braked hard that they constitute strong evidence that a vehicle has braked hard. Indeed, they constitute clinching (conclusive) evidence in view of which it can be settled (established) that a vehicle has braked hard. What makes the evidence clinching is not just that the presence of streaks of that sort reliably indicates that a vehicle has braked hard, but that these particular streaks do indicate that a vehicle has braked hard. (In the circumstances, these very streaks would be there only if a vehicle had braked hard. Nothing else would have caused them to be there if a braking vehicle had not caused them to be there.) That is why their presence can supply you with a reason adequate for knowledge to believe that a vehicle has braked hard, but this, of course, raises the question as to what makes such a reason available to you. This is where recognition of significance comes into the picture. Recognition figures in two ways in the case in hand. You recognize the streaks on the road as streaks of a certain familiar appearance, though you would be hard put to describe their appearance with any specificity (cf. Austin –: –). You might well think of them as having the appearance of skid-marks or tyre-marks. You also recognize the indicative significance of those streaks – that they indicate, and so can show or reveal, that a vehicle has braked hard. It might initially seem odd to talk of recognition in connection with indicative significance. You see the streaks but not the event that caused them to be there. Yet much of what we see to be so implicates a relation to something that is not then perceptible. We recognize a blemish on someone’s arm as a scar even though its being a scar entails that the skin has been cut, or otherwise injured. We recognize a white coating on grass as frost even though its being frost entails that it has recently been freezing. There is nothing odd, then, about recognizing that the black streaks on the road indicate that a vehicle has braked hard. That they indicate this is an implication of their being skid-marks and we can certainly see that such marks are skid-marks (cf. Dretske  on secondary epistemic seeing). It seems plausible that we could have learned to think of the streaks as skid-marks in a manner that is analogous to learning that birds that look like these are robins, or that fruits that look like these are pears. In cases of the latter sort, the work of learning goes into discriminating robins from other birds and pears from other fruits. It does not depend on having evidence for a covering generalization linking an appearance with being of some kind. I suggest that, similarly, the ability perceptually to recognize



 

phenomena as having a certain indicative significance – as being signs that such and such – does not depend on having evidence for a generalization linking the phenomenon with what it indicates. Acquiring the ability is like learning a technique or skill. The test for having it is reliably manifesting the skill. It does not address whether you have evidence for a suitable generalization. Similar considerations plausibly apply to many others cases, including measuring a patient’s temperature with a thermometer, telling the approximate level of fuel in a car’s fuel tank from the fuel gauge, and telling that it has been raining from the (right kind of ) wetness of streets and pavements. What matters for learning to do these things is getting the hang of the right technique.

. Recognitional Abilities and Justified Belief I have explicated perceptual knowledge in terms of the exercise of recognitional abilities, not in terms of justified belief. Standard ways of accounting for the justification for belief that perception can yield are in my view highly problematic. (I explore this and develop a different way of thinking in Millar b, a, and , ch. .) Consider the good case in which you recognize pears to be pears from the way they look and a corresponding bad case in which, faced with things that look just like pears, you incorrectly judge those things to be pears. The problematic perspective under consideration incorporates a certain Parity Thesis according to which cases related in this way can be on a par with respect to justified belief. It is assumed to be constitutive of the good case that one has a justified belief and that whatever factors make for justified belief in that case could be realized also in the bad case. Suppose that our good and bad cases are so related, according to some account of justified belief that incorporates the Parity Thesis. Following this line of thought, it would seem natural to suppose that what reason achieves in the good case is no less achieved in the bad case. A corollary is that even in the good case reason does not, as it were, reach out to (make contact with) the fact you come to know. It is open to those advancing such a view to draw on a perspective on which the justification arising from perception is taken to be provided by experiences gained in perception, these being conceived to be metaphysically independent of what currently obtains or occurs in one’s surroundings. It is irrelevant to the justification that the experiences are acquired in an episode of perceiving our surroundings. So this perspective can accommodate the idea that you could be no less justified in believing there to be pears before you in a case in which you

Explaining Knowledge



perfectly hallucinate a box of pears than you are in our imagined good case in which you see the pears, recognizing them to be pears. If one accepts the way of thinking about justified belief that I have just sketched, one might be puzzled by how it could be thought that the work of reason is different in our good and bad cases, for in both cases you form a belief on the basis of a visual experience in circumstances in which whatever other requirements there are for justified belief are supposed to be satisfied. This reaction works with an overly restrictive conception of the reach of reason. Recognizing the pears as being pears from the way they look is clearly a rational act. Since it puts you in cognitive contact with the fact in question, we could say that it is an act by which reason reaches right out to and embraces the fact itself. The metaphors of cognitive contact, and of reason reaching out to a fact, have bite because they serve to mark a contrast with a picture under which knowledge is conceived as something acquired by an act – the formation of a justified belief – that in itself falls short of the acquisition of knowledge, but which can contribute to such acquisition when further, independent, conditions are satisfied. On this latter picture, reason’s work goes no further than forming a belief in a manner and in circumstances that ensure that the belief is justified, but that is no guarantee that knowledge has been acquired. When, by contrast, we think of knowledge as cognitive contact with, or grasp of, a fact, the fact known is partially constitutive of the act whereby it is known, so that the knower is not, as it were, at some remove from the fact. In our favoured case, the act in question is recognizing the pears to be pears. It is not performed in the bad case. There is no good reason to deny that reason is at work in the bad case, so long as we acknowledge that the way in which it is at work differs from that in the good case. In the bad case there is no failure of due epistemic diligence on your part – had there been perceptible signs that you were, or might be, in a fake-pear set-up you would have noticed this. So it is not just that you are blameless in the bad case, for in judging as you did you were reasonable. In respect of reasonableness, the good and bad cases are on a par. My claim is that, nonetheless, the cases are fundamentally different because in one you grasp a fact in a rational act and in the other you do not. In both cases your thinking was directed at recognition. In the bad case recognition  

This picture, I take it, is the target of McDowell . I agree with Christoph Kelp () that simply invoking blamelessness in this context would not go far enough.



 

was not achieved. This sort of thing is bound to happen from time to time, since we are fallible with respect to our recognitional abilities as we are with respect to all of our abilities. We are not bound to exercise them whenever we act in a manner directed at doing so. Under the conception of justified belief that I have been criticizing, perception of mind-independent things (objects, scenes, events, features) is merely the contingent source of the experiences that are the real sources of justification for beliefs that perceptions yield. Since our good and bad cases are taken to be on a par with the respect to justification, it is thought to be irrelevant to the justificatory power of the experiences in the good case that those experiences are implicated in an episode of seeing pears. On mainstream ways of thinking, this is taken to be a datum that must be accommodated. It is an embarrassment from a perspective that accommodates the, by no means outlandish, idea that rationality is in part a matter of having a grip on reality. How then from this perspective are we to think of justified belief? Those with the requisite conceptual resources can have reasons to believe things they see to be so that are constituted by facts about the visual appearance of the objects they see. Believing for such reasons serves to sustain belief in what is seen to be so. Another type of reason is also routinely available to those with the relevant conceptual resources. I look through a window of my house and say, ‘It’s raining.’ Someone says that it’s not supposed to rain today and asks me whether I’m sure. I reply somewhat impatiently, ‘I can see that it is raining.’ In saying this I both explain how I know, and give a reason for thinking, that it is raining. For another case, suppose that I am asked whether a certain colleague was at a meeting that I attended yesterday. No one is going to be confused or baffled if I say, ‘Yes, I saw that she was there.’ (Perhaps it would be a little more natural to say, ‘Yes, I saw her there’, or even just, ‘Yes, she was’. The first of those statements would in a suitable context imply that I saw that she was there and so recognized her to be there. The second could do so as well in conjunction with information about a way in which I would have been able to know of the colleague’s attendance.) That I saw that this colleague was there is clearly a reason for me to think now that this colleague was there. In the absence of any reason to doubt my report it is likely to be taken by others to give them a reason to think that this colleague was there. So far, we have common sense. A task of epistemology is to make sense of () how we have first-person access to the kind of reason under consideration and () how such reasons figure in a satisfying account of the relation between knowledge and justified belief.

Explaining Knowledge



It is a step towards addressing () to acknowledge that developed human beings have the ability not only to recognize things they perceive to be some way, but also to tell of such things that they are things they see to be that way. On seeing the pears, I know that I see them and that I see them to be (see that they are) pears. Although the latter knowledge is reflective, in that it is about my knowing something, it is not well described as introspective, since knowledge as to what one sees is gained by looking outward. We have a general ability to tell, of things we see to be some way, that we see them to be that way. To see what this amounts to it will help to reflect more closely on simply telling that the things in the box are pears. I tell from the way the pears look that they are pears. The way they look is a feature that the pears have irrespective of whether anyone is actually looking at them. Given that the environment is favourable, it is a feature that reveals them to be pears. In order for me to tell from their look that they are pears, their look must not just reveal them to be pears, but also be manifest to me. The manifestation of this look to me elicits my recognition of them as pears. Now, clearly the look of the pears does not reveal that they have the relational property of being seen by me. Nonetheless, thanks to my having a general ability to tell of things I see that I see them to be the ways I recognize them to be, the manifestation of the look to me enables me to tell that I see the pears to be pears. In this circumstance, I tell that I see the objects in question to be pears in a rational act that is an immediate response to their manifesting the look of pears to me. In other words, their manifesting that look to me elicits not only visual recognition of the pears as pears, but also the knowledge that I see that they are pears. In looking at the pears, then, I am not puzzled or taken by surprise by coming to believe that they are pears; I know why I so believe, since I know that I see that they are pears. Moreover, the fact that I see them to be pears constitutes a reason to believe that they are and one that sustains me in believing that they are. It could easily seem to be an objection to this line of thought that the reason in question is not one in the light of which I came to believe that the things in the box are pears. I came to believe that they are pears, not because of something else I believed but simply because I saw them and, thanks to my having a general ability to do so, recognized them to be pears. But it is also true that in seeing that they are pears I simultaneously acquired the belief that I saw that they are. Although the latter belief did not lead me to believe that they are pears, it sustains that belief so that, among other things, I’d be liable to resist giving up the belief that they are pears so long as I believed that I saw them to be pears. It is commonplace



 

for beliefs acquired simultaneously to be such that one sustains the other even though it did not generate the other. The question now (question () above) is how the account just set out fits into a satisfying perspective on the relation between perceptual knowledge and justified belief. The issue might seem to be complicated by the fact that under the basic conception of perceptual knowledge with which I am working we do not account for such knowledge in terms of justified belief. One’s seeing that P is a way of knowing that P that is explicated in terms of the exercise of an appropriate recognitional ability, not in terms of justified belief. Nonetheless, the account has it that, for those with the requisite conceptual resources, there is justification for believing what one sees to be so that consists in believing it to be so for a reason constituted by the fact that one sees it to be so. The same applies to justification derived from taking in that a thing perceived has a certain look. A pleasing feature of this approach to perceptual justification is that it enables us to think of such justification in keeping with the reasons conception of justified belief. On this conception to be justified in believing something amounts to having an adequate reason to believe it and believing it for that reason. This enables us to conceive of all justified belief, even that provided by perception, as being provided by reasons constituted by truths. This way of thinking is consonant with what seems to me to be the entirely sensible, and indeed common sense, idea that justification worthy of the name, by contrast with the reasonableness of which I spoke earlier, is provided by what we know. Such a view can seem to be objectionable because one assumes that knowledge must be explicated in terms of justified belief. In my thinking, it is the latter view that should be abandoned. On such an approach we are not compelled to think of knowledge as a primitive for which no substantive account can be given, for we may examine the diverse forms that knowledge takes along the lines I proposed in Section ., and which I illustrated and commented upon in Sections ., ., and .. Even so, it might be thought to be an objection to the view of perceptual knowledge that I am proposing that, although it accommodates the thought that human beings, with the requisite conceptual resources, who know that P in virtue of seeing that P, have a justification for believing that P, it also represents it to be possible that one should know that P in virtue of seeing that P while lacking such justification. (For this concern, see Kelp .) It might assuage the worry here if we introduce a broad notion of right believing. Right believing is either believing in virtue of knowing, or believing for an adequate reason

Explaining Knowledge



constituted by truths one knows. It is a purely terminological matter whether we take justified belief to be right believing thus broadly conceived, or instead take it to be belief justified in keeping with the reasons conception, but note that either way knowledge is explanatorily prior to justified belief. (Clearly, it is constitutive of knowledge based on assumptions that the assumptions provide one with a reason to believe, thus a justification to believe, that is adequate for knowledge, but that justification itself depends on knowledge of the truth of the assumptions. See Millar [a, ] on this.) It might be suggested that when one sees that P the fact that P is itself a reason for one to believe that P and can be a reason for which one so believes. There is, though, a plausible constraint on believing that P for a good reason, that generates a problem for such a view. A reason for one to believe that P is such that if one were to believe that P for that reason then it should be possible for one to understand why one believes that P in the light of the truth that constitutes that reason. If the truth that Q constitutes a reason for me to believe that P then it should be possible for me to make sense of my believing that P for that reason. As applied to the case in which I know that my colleague is at the meeting on seeing her there, the suggestion under consideration has it that the truth that my colleague is there can be a reason for which I believe that she is there, and one which would justify me in so believing. The problem is that this truth is not one in the light of which I can make it intelligible to myself that I should now believe that my colleague is there. Suppose I were to ask myself, ‘Why do I believe this?’. The answer, ‘Because she is there’, does not provide the sought-for explanation. The fact that she is there entails nothing as to how I relate to it, so it is not surprising that citing it does not suffice to give me an explanation of why I believe that she is there. Against this it might be said that it does give such an explanation since in the circumstances the fact that my colleague is there explains why I believe that she is, but this rests on confusion. In the circumstances, I have an ability to tell when my colleague is present. The fact in question explains why this ability is exercised when I see this colleague there, but that explanation pertains to how I came to know that 



I do not assume that all right believing is knowing. One can rightly believe something for a reason constituted by truths one knows, even if these truths do not provide a reason adequate for knowledge. At the Joint Session of the Mind Association and Aristotelian Society at Cambridge in , Lucy O’ Brien, in response to my Inaugural Address, asked me why I did not countenance such a view. Kurt Sylvan advanced it in his presentation to the Virtue Epistemology conference at KU Leuven in September . On neither occasion did I have a clear idea of why it would not do.



 

my colleague is there. It does not explain what reason I have to believe that she is there. Of course, there is such an explanation: the fact I see that she is present gives me a reason so to believe.

. Abilities, Reliability, and Fallibility Under the conception of recognitional abilities that I have been outlining, these abilities are exercised only in doing what the ability is an ability to do. I have argued elsewhere (Millar b, ) that this is in keeping with a plausible perspective on abilities in general. One might object to this approach because one thinks that it cannot adequately account for our fallibility. In reliabilist epistemology our fallibility in forming true beliefs is captured in terms of the idea that the mechanisms of belief-formation that account for the acquisition of knowledge reliably produce true beliefs but sometimes do not. The kind of virtue epistemology that grew out of, and reacted against, reliabilist epistemology (e.g. Sosa a, b, and Greco ) focused not on reliable mechanisms but on agents who are reliable in virtue of having abilities or competences that are taken to account for the acquisition of knowledge, but which are sometimes manifested in acquiring false beliefs. From this perspective it might seem that recognitional abilities as I conceive of them are infallible, but recognitional abilities do not admit of assessment in such terms. The question is whether we are fallible in relation to our abilities, and the correct answer to that is, obviously, that we are. If Φ-ing is an intentional action, our fallibility in relation to an ability to Φ is a matter of our being such that we would not always succeed in Φ-ing on every occasion on which we intend to Φ. It does not follow from this that the ability to Φ can be exercised by doing something that falls short of Φ-ing. In other words, it does not follow that there can be defective exercises of such an ability. The cases of failure may quite naturally be conceived simply as cases in which the ability is not exercised. There is nothing ad hoc or stipulative about this. For it seems entirely natural to hold that you exercise an ability only if you do, or are doing, what the ability is an ability to do. 

Reliabilists and virtue epistemologists are by no means alone in thinking that fallibility requires us to think that the modes belief-formation that result in our acquiring knowledge must sometimes result in false beliefs. John McDowell (, ) invokes defective exercises of rational capacities. Lisa Mirachi () invokes degenerate exercises of competences to know that are not acquisitions of knowledge. Clayton Littlejohn (, sec. ) thinks that the same ability can be exercised both in favourable environments and in unfavourable environments. He conceives of unfavourable environments as ones in which knowledge is not acquired through the exercise of the ability.

Explaining Knowledge



You exercise an ability to ski only when skiing. You exercise an ability to make a cake only if you are making a cake. You might be making a cake yet not make it (not complete the making of it) because you are interrupted and don’t get back to the task. As you are making the cake you are exercising the ability to make a cake, but if you don’t bring the task to completion then, of course, you don’t complete the exercise of the ability. (This issue is more fully explored in Millar b and , ch. .) Recognition is not an action, since it is not something one does intentionally. A recognition-directed judgement – a judgement aimed (directed) at recognition – is by contrast a mental act, an actualization of the capacities required to make such judgements (Geach ). Successful recognition-directed judgement just is recognition. To account for our fallibility with respect to recognition-directed judgement, in keeping with the perspective that I am commending in this discussion, we need a treatment of fallibility that is somewhat different from that which applies to abilities the exercises of which are actions. An aspect of one’s fallibility in relation, say, to being able to recognize a song thrush as a song thrush from the way it looks is not being bound to recognize a song thrush as a song thrush on every occasion on which one makes a judgement that a thing is a song thrush going by the way it looks. Sometimes one might judge wrongly. Another aspect is that one might fail to effect recognition either because of some impairment (e.g. blurred vision) or because although it is as if one is looking at a song thrush one is not (e.g. in a perceptual Gettier situation, or if undergoing a hallucination). Failures such as these are failures to exercise the ability. There is no need to posit defective exercises of the ability to account for them. Moreover, from the fact that those who have an ability to recognize Fs as Fs from ways Fs look might sometimes do so and sometimes not when making a judgement directed at recognition, it does not follow that this ability is what in Section . paragraph D, I called a success-rate ability. For, again, the cases of failure are not just cases in which the end aimed at – recognizing the thing perceived to be an F – is not attained; they are cases in which the ability is not exercised. 

I am grateful to participants at the Virtue Epistemology conference at KU Leuven in September  for general discussion and feedback on a presentation of some of the material set out here, and to Chris Kelp for organizing that conference. Discussions with participants at the Helsinki Epistemology Workshop in August , especially those sceptical of my view, helped to motivate and shape the present chapter. I am grateful to Markus Lammenranta for organizing that workshop. I thank colleagues at Stirling for the opportunity to discuss an earlier version of this paper, and John Greco for comments that prompted what I hope are improvements.



  R E F E R EN C E S

Armstrong, D. . Belief, Truth and Knowledge. Cambridge, UK: Cambridge University Press. Austin, J. L. –/. “Other Minds,” Proceedings of the Aristotelian Society, Supplementary Volume : –. Reprinted in Austin’s Philosophical Papers. rd edn. J. O. Urmson and G. J. Warnock (eds). Oxford: Clarendon Press, –. Page references are to this volume. Dretske, F. . Seeing and Knowing. London: Routledge and Kegan Paul. Fernández Vargas, M. A. (ed.). Performance Epistemology: Foundations and Applications. Oxford: Oxford University Press. French, C. . “Does Propositional Seeing Entail Propositional Knowledge,” Theoria : –. Geach, P. . Mental Acts. London: Routledge and Kegan Paul. Greco, J. . “Agent Reliabilism,” Philosophical Perspectives : Epistemology: –. . Achieving Knowledge: A Virtue-theoretic Account of Epistemic Normativity. Cambridge, UK: Cambridge University Press. . “A (Different) Virtue Epistemology,” Philosophy and Phenomenological Research : –. . “Knowledge, Virtue, and Safety,” in Fernández Vargas (ed.), –. Haddock, A. and Macpherson, F. (eds). . Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Hyman, J. . Action, Knowledge, and Will. Oxford: Oxford University Press. Kelp, C. . “Justified Belief: Knowledge First-Style,” Philosophy and Phenomenological Research : –. Littlejohn, C. . “Fake Barns and False Dilemmas,” Episteme : –. McDowell, J. . “Criteria, Defeasibility and Knowledge,” Proceedings of the British Academy : –. Reprinted in McDowell , –. . Mind and World. Cambridge, MA: Harvard University Press. . “Knowledge and the Internal,” Philosophy and Phenomenological Research : –. Reprinted in McDowell , –. . Meaning, Knowledge, and Reality. Cambridge, MA: Harvard University Press. . “The Disjunctive Conception of Experience as Material for a Transcendental Argument,” in Haddock and Macpherson (eds) , –. . “Tyler Burge on Disjunctivism,” Philosophical Explorations : –. . Perception as a Capacity for Knowledge. Milwaukee, WI: Marquette University Press. Maier, J. . “Abilities,” The Stanford Encyclopedia of Philosophy (Fall  Edition), Edward N. Zalta (ed.). http://plato.stanford.edu/archives/fall/ entries/abilities/. Accessed January , . Mele, A. R. . “Agents’ Abilities,” Noûs : –. Millar, A. . “The Scope of Perceptual Knowledge,” Philosophy : –.

Explaining Knowledge



. “Perceptual-Recognitional Abilities and Perceptual Knowledge,” in Haddock and Macpherson (eds) , –. a. “Knowledge and Reasons for Belief,” in A. Reisner and A. SteglichPetersen (eds). Reasons for Belief. Cambridge, UK: Cambridge University Press, –. b. “How Visual Perception Yields Reasons for Belief,” Philosophical Issues : The Epistemology of Perception: –. . “Scepticism, Perceptual Knowledge, and Doxastic Responsibility,” Synthese : –. . “Perceptual Knowledge and Background Beliefs,” In D. Dodd and E. Zardini (eds), Scepticism and Perceptual Justification. Oxford: Oxford University Press, –. a. “Perceptual Knowledge and Well-Founded Belief’, Episteme : –. b. “Abilities, Competences, and Fallibility,” in Fernández Vargas (ed.) , –. . Knowing by Perceiving. Oxford: Oxford University Press. Mirachi, L. . “Competence to Know,” Philosophical Studies : –. Pritchard, D., Millar, A., and Haddock, A. . The Nature and Value of Knowledge: Three Investigations. Oxford: Oxford University Press. Putnam, H. . “Sense, Nonsense, and the Senses: An Inquiry into the Powers of the Human Mind,” The Journal of Philosophy : –. Roessler, J. . “Perceptual Experience and Perceptual Knowledge,” Mind : –. . “Perceptual Attention and the Space of Reasons,” in C. Mole, D. Smithies, and W. Wu (eds), Attention: Philosophical and Psychological Essays, Oxford: Oxford University Press, –. Sellars, W. . Science, Perception and Reality. London: Routledge and Kegan Paul. Sosa, E. a. “Reliabilism and Intellectual Virtue,” in Sosa c, –. b. “Intellectual Virtue in Perspective,” in Sosa c, –. c. Knowledge in Perspective: Selected Essays in Epistemology. Cambridge, UK: Cambridge University Press. . A Virtue Epistemology: Apt Belief and Reflective Knowledge, Volume I. Oxford: Clarendon Press. . “How Competence Matters In Epistemology,” Philosophical Perspectives : Epistemology: –. . Knowing Full Well. Princeton, NJ: Princeton University Press. . “Epistemic Competence and Judgement,” in Fernández Vargas (ed.), –. Stroud, B. (). “Perceptual Knowledge and Epistemological Satisfaction,” In J. Greco (ed.), Ernest Sosa and His Critics. Oxford: Blackwell, –. . “Seeing What Is So,” in J. Roessler, H. Lerman, and N. Eilan (eds), Perception, Causation, and Objectivity. Oxford: Oxford University Press, –.



 

Taylor, C. . “Retrieving Realism,” in J. K. Schear (ed.) Mind, Reason, and Being-in-the-World: The McDowell-Dreyfus Debate. Abingdon: Routledge, –. Warnock, G. /. “Seeing,” Proceedings of the Aristotelian Society : –. Whittle, A. . “Dispositional Abilities,” Philosophers’ Imprint : –. Williamson, T. . Knowledge and Its Limits. Oxford: Oxford University Press. Zagzebski, L. . “What Is Knowledge?” in J. Greco and E. Sosa (eds), The Blackwell Guide to Epistemology. Oxford: Blackwell, –.

 

Anti-risk Virtue Epistemology Duncan Pritchard

. Introductory Remarks For more than a decade now I have been defending a particular kind of methodology as regards the theory of knowledge, which I call anti-luck epistemology. In essence, this argues that we should take seriously the platitude that knowledge excludes luck. This means both giving an account of the nature of luck and also articulating the precise manner in which knowledge is incompatible with it. Once one puts these two components together, the thought runs, then one will have formulated, on a principled basis, the anti-luck condition on knowledge. Such a methodology has many advantages, not least because it offers a more systematic way of approaching the problems posed by the theory of knowledge. For one thing, we are now not simply formulating a condition and then testing it against various counterexamples to see how it fares relative to competing conditions, in a piecemeal manner. For another, the condition we come up with will cover all cases where it matters that knowledge and luck are incompatible, as opposed to the more standard approach of trying to find an anti-Gettier condition and then seeing whether, hopefully, it will also have application to other kinds of cases that trade on luck (such as lottery cases, for example). In previous work I’ve made a number of claims about anti-luck epistemology. For instance, I’ve argued that it not only motivates the safety condition on knowledge over other competing modal conditions on knowledge (such as sensitivity), but also gives us an explanation of how  

See, for example, Pritchard , , , a, b, a, and Pritchard, Millar and Haddock : chs. –. Another aspect of the anti-luck epistemology methodology that I emphasise in my presentations of it is its appeal to the empirical literature on luck (and, relatedly, risk) ascriptions. See, for example, Pritchard . See n.  for a list of some of the relevant empirical literature. For a survey of some of this literature, see Pritchard and Smith .





 

this condition should be interpreted – an interpretation that avoids some familiar problems that have been levelled at it. I’ve also argued that antiluck epistemology not only provides us with an account of the anti-luck condition on knowledge, but that we can use this as a basis to offer a complete virtue-theoretic account of knowledge, one that does justice to both the platitude that knowledge excludes luck and the related platitude – the ability intuition, as I call it – that knowledge entails the manifestation of relevant cognitive ability. I describe the resulting theory of knowledge as anti-luck virtue epistemology. My concern here, however, is not with further defending and articulating anti-luck epistemology (or anti-luck virtue epistemology). Rather, I want to explain a recent shift in my position. For while I think that anti-luck epistemology and anti-luck virtue epistemology are broadly on the right lines, I have become convinced that both need to be refined in an important way. In particular, I now argue that the relevant notion that we should focus on as epistemologists (and not just as theorists of knowledge, as I explain below) is in fact epistemic risk rather than epistemic luck. Antiluck epistemology is thus to be replaced with anti-risk epistemology. Moreover, and crucially for our purposes, we will see that this also means offering a new virtue-theoretic account of knowledge: anti-risk virtue epistemology.

. Luck and Risk In order to understand the transition from anti-luck epistemology to antirisk epistemology, we first need to appreciate some of the subtle – though as we will see, important nonetheless – differences between luck and risk. Luck and risk are very closely related notions. Indeed, it is quite common for psychologists to study risk and luck ascriptions together. It is easy to 

  

See especially Pritchard , a. For example, I’ve claimed that anti-luck epistemology motivates a version of the safety principle that evades the dilemma posed for safety by Greco (). I’ve also claimed – e.g. in the works just cited – that it is better placed to handle a problem posed to safety by necessary, or at least modally stable, truths (although as we will see below, I think anti-risk epistemology is in fact on stronger ground in this regard). Versions of the safety principle had previously been defended by such figures as Sainsbury (), Sosa (), and Williamson (). For a comparison of safety and the competing sensitivity principle in the context of an anti-luck epistemology, see Pritchard (). See especially Pritchard a and Pritchard, Millar, and Haddock : chs –). See also Pritchard b. See especially Pritchard b. See also Pritchard a. See, for example, Teigen a. For discussion of this point, and further elaboration on the empirical literature in this regard, see Pritchard and Smith .

Anti-risk Virtue Epistemology



see why. Suppose you are a journalist working in a war zone and a sniper’s bullet whizzes past your ear, narrowly missing you. You were lucky not to be shot. There was also a high risk of you being shot. Or suppose you are on a plane that crash lands, albeit safely, into the Hudson River. You were lucky not to die. You were also at a high risk of dying. And so on. Cases like this show that, often at least, luck and risk go hand-in-hand. If it is lucky that such-and-such event failed to happen, then that tends to correlate with it being the case that there was a high risk of it happening. The close relationship between the two notions means that it is easy to overlook their differences. And yet, as we will see, the differences are in fact important, and important in a way that is particularly relevant to epistemology. I’ve argued at length elsewhere that we need a modal account of luck. This holds, roughly, that a lucky event is an event that actually occurs but which doesn’t occur in close possible worlds where the relevant initial conditions for that event remain the same. Narrowly avoiding being shot by a sniper’s bullet is lucky because there is a close possible world where, keeping the relevant initial conditions fixed (e.g. someone is still trying to shoot you, and so forth), you do get shot. In contrast, there is nothing lucky about not being shot in normal circumstances where there are no snipers around, since there are no close possible where one gets shot. Notice too how the modal account of luck can also account for how luck comes in degrees. Ceteris paribus, narrowly avoiding a sniper’s bullet by inches is luckier than avoiding it by several feet, since the non-obtaining of the target event is modally closer in the first scenario. Or consider, for example, a paradigm case of a lucky event: a lottery win. On the modal account this is a lucky event because there are very close possible worlds where the relevant initial conditions for that event remain the same (e.g. you continue to buy a lottery ticket, the lottery remains free and fair, and with the same long odds of winning, and so on), but you fail to win. The lottery win is an interesting case because it also illustrates why one wouldn’t want a straightforwardly probabilistic account of luck. This might be initially surprising, since a probabilistic account of luck, on the face of things, looks promising. Aren’t lucky events those events that are probabilistically unlikely to happen? So, for example, isn’t the lottery win a lucky event because it was a low probability event? This is 



See especially Pritchard : ch. , . Note that luck doesn’t just apply to events, but I will henceforth focus on events in order to simplify my discussion (and I will be making the same simplifying assumption in my discussion of risk). The modal account of risk that I describe below inherits this feature.



 

too quick, however, for consider the person who plays the lottery but loses. It is also true of them that they could very easily have won – all it would have taken, after all, is that a few coloured balls fall in a different configuration. And yet the odds massively support the possibility that they would lose. The crux of the matter is that there can be very low-probability events (in this case winning the lottery) that can nonetheless be modally close. But in such cases we still treat the event as being down to luck (bad luck, in the case of losing the lottery), something that is widely confirmed by the empirical literature on luck ascriptions. Indeed, isn’t that just why we play the lottery (but don’t place bets on scenarios with similar long odds but which are modally far-fetched, like the North Korean soccer team winning the World Cup)? The modal account of luck, while obviously not without its detractors, is certainly the leading view in the literature. Given the close connections between luck and risk, a modal account of risk would be similarly appealing, such that, roughly, our concern in making a risk assessment is the modal closeness of the target event (rather than, for example, its probability). But once one develops such an account, however, one starts to notice the small, but important, differences between the notions of luck and risk. The first big difference is that luck is about the non-obtaining of the target event. To have a lucky lottery win entails that you could have very easily lost the lottery drawing, in relevantly similar circumstances. To have luckily avoided being shot by a sniper while working as a journalist in a war zone entails that you could have easily been shot by that sniper. And so on. The important point is that our focus is on the closeness of the possible 







See, for example, Teigen , , , a, b, ; Tetlock ; Kahneman and Varey ; and Tetlock and Lebow . For a survey of some of this empirical literature, see Pritchard and Smith . The UK national lottery used to trade on this feature in its advertising campaign. They had the slogan ‘It could be you!’ and featured someone being picked out as a winner by a God-like finger. Clearly this is not the ‘could’ of probability, as in this sense it couldn’t be you, as the odds are so astronomically long. Rather it is the ‘could’ of modal closeness: if you play the lottery, then someone, just like you, will be a winner; all that is required is that a few numbered balls fall in a slightly different configuration. (Note too that the UK has very strict advertising laws, and it is unlikely they would have got away with such a campaign slogan had the probabilistic reading been the only one available). To be fair, this is largely because for a while it was the only full-fledged philosophical account of luck in the literature. For some detractors/proponents of rival (or at least distinct) views, see Riggs , ; Lackey ; Levy ; Broncano-Berrocal ; and Coffman . Interestingly, the dominant account of risk in the literature is probabilistic – see, for example, Hansson . See also Hansson . I argue against the probabilistic conception of risk in Pritchard b, where I also articulate the modal account of luck.

Anti-risk Virtue Epistemology



world where the target event that actually occurred failed to occur (while keeping the relevant initial conditions fixed). Risk is slightly different in this regard. When we make assessments of risk, we have in mind some specific risk event that we want to avoid. So, for instance, when flying we might make an assessment of risk relative to the risk event of the plane crashing. Or, when mountain climbing, we might make an assessment of risk relative to the risk event of falling off the mountain. On the modal account of risk, these assessments will be concerned with the modal closeness of the risk event (keeping all relevant initial conditions fixed), such that the closer it is, the riskier it is. But won’t that make the modal account of risk exactly like the modal account of luck? Interestingly, no. The reason for this is that we can vary our risk assessments by varying the risk event that we focus upon. So, for example, we may judge that in flying there is a far greater risk of the plane crashing at take-off and landing than there is while in flight. That is, we might judge that the former scenario is modally closer, and hence riskier, than the latter scenario (which it is, I gather). Changing our target risk event thus changes the level of risk involved in the target event (in this case, taking the plane). There is no limit to the range of risk events that we can focus upon, and indeed we can even imagine complex risk assessments that involve multiple risk events (though to keep matters simple I will just focus on assessments involving a single risk event). Notice the difference with luck. With luck, we first identify a target event, and then we consider the modal closeness of that event not obtaining. Once the event has been determined, then there is no further parameter that we need to consider in evaluating whether it is lucky. With risk, however, there is a further parameter, since we also need to select the risk event that we wish to focus upon. I think this point gets overlooked because of course we can change the direction of focus with luck too, by considering a different target event. Rather than focusing on the event of avoiding being killed by a sniper’s bullet, for example, we could focus on the event of avoiding being shot (but not necessarily killed). This will undoubtedly make a difference to how lucky we think the target event is. But the point is that with luck, but not risk, once you’ve selected your target event, there is no further parameter that needs to be fixed to determine whether the event is lucky (as there is with risk). This detail might seem so small as to be insignificant, but as we will see it is in fact very important to resolving a (hitherto unnoticed) difficulty facing anti-luck 

See Pritchard b for further discussion of the modal account of risk.



 

epistemology. It also enables an anti-risk epistemology to offer a broader set of epistemic assessments than anti-luck epistemology. There are further differences between luck and risk that we should flag. One interesting difference is that risk assessments are essentially forward looking, while luck assessments are essentially backward looking. When we judge that we were lucky to have avoided the sniper’s bullet, we are making this assessment from the perspective where things turned out well, luckily for us (as in ‘phew, I was lucky there’). In contrast, when we judge that we were at a high risk of being shot we are taking the forwards-looking perspective of considering our situation prior to the sniper firing, and responding to the clear danger that we are facing. As we will see, this has important diagnostic implications for our understanding of postGettier epistemology. A further point in this regard is that our interest in luck – bad luck, anyway (which happens in any case to be the main concern of epistemologists) – is usually motivated by a concern to eliminate risk, rather than vice versa. Why is it significant that one was lucky not to fall off the mountain? Well, ordinarily at least, because one was concerned that one was at such a high risk of falling (and dying). Why is it significant that one was lucky not to die in the plane crash? Well, ordinarily at least, because one was concerned at the high level of risk of dying in that plane crash. The point is that while luck and risk tend to go hand-in-hand, our deeper concern is with the latter rather than the former, such that our interest in (bad) luck typically signals a concern to eliminate risk. As we will see, this relationship between luck and risk also has important diagnostic implications that an anti-risk epistemology can exploit to its advantage. There are other differences between luck and risk, but many of them aren’t salient for our current purposes. I will just mention one, since it’s interesting in its own right, which is that luck is clearly double valenced, in that it comes in both good and bad forms (though, as just noted, in epistemology it is usually just the bad epistemic luck that we are interested in). Risk, in contrast, tends to concern negative outcomes, in that the risk event is something that we want to avoid (such as a plane crash). Even when risk adds value to an activity, as with extreme sports or aesthetic risk, the risk event is still something negative (serious injury, ruining the performance).  

This difference came to light in discussion with Jesus Navarro. I discuss ‘positive’ kinds of risk of this kind, with a particular focus on aesthetic risk, in Pritchard . I offer a more detailed description of the modal account of risk in Pritchard b. See also Pritchard c, where I apply the modal account of risk to epistemological issues in the philosophy of law.

Anti-risk Virtue Epistemology



. Anti-risk Epistemology So, how will an anti-risk epistemology differ from an anti-luck epistemology? We first need to understand something that unsettled me about anti-luck epistemology from the off (though, oddly, no-one else clocked this theoretical lacuna, so far as I know). Recall that the methodology of anti-luck epistemology involves putting together a theory of luck with an account of the specific way in which knowledge excludes luck. We’ve encountered the modal account of luck already. Sparing the reader the details, the specific way in which knowledge is incompatible with luck is veritic epistemic luck, whereby it is a matter of luck that one’s belief is true, given how it is formed. Given the modal account of luck, this means that we have a condition on knowledge such that the non-obtaining of the target event (true belief ) could very easily have occurred. But, strictly speaking, this entails that our focus shouldn’t be on the closest possible world where one believes falsely on the same basis, but rather the closest possible world where one fails to form a true belief on the same basis, which is a disjunction of possible worlds where one forms a false belief and where one fails to form a belief at all. It is the latter part of the disjunction that is troubling, as it doesn’t seem to be necessarily knowledge-undermining to form a true belief on a basis whereby one could very easily have not formed a belief at all on that same basis. Perhaps one is just a cautious believer? Indeed, isn’t being a cautious believer often a sign that one forms one’s beliefs in an epistemically conscientious way, and hence that one is more likely to be a knower? But how is one to restrict the view, in a principled fashion, according to anti-luck epistemology? My thought, way back in Pritchard (), was that the fact that one had initial conditions built into the modal account of luck, and hence had an independent rationale for a basis-relative account of the anti-luck condition (safety), would offer a response to this problem. The idea was that so long as one takes basis-relativity seriously, then this problem disappears because in the relevant cases one’s basis is never such as to allow mere non-belief. Consider, for example, a case in which one believes that one’s lottery ticket is a loser because one reads the result in a national, well-established, newspaper. We now need to examine the closest possible worlds where one 

Veritic epistemic luck is thus, as I put it, a malign form of epistemic luck, in that it is incompatible with knowledge, as opposed to benign forms of epistemic luck, like the evidential epistemic luck that one is lucky to have the evidence that one does. For the details, see Pritchard , passim.



 

has the same basis for belief as in the actual world. This is going to entail reading the lottery result in a national, well-established, newspaper. But if that’s right, then there won’t be a close possible world where the basis is the same but one fails to form any belief at all, as reading the result in the newspaper is going to lead one to form some judgement about whether one’s ticket is a winner or a loser. The problem, however, is that while it’s true that appealing to the basisrelativity of safety will deal with most cases, it does not deal with all of them. Consider, for instance, a subject who has a sound memorial basis for their belief that the Battle of Hastings was in . Suppose, however, that although our subject is accordingly confident of this belief in the actual world, there are close possible worlds where they is disposed to doubt themselves. It is just a psychological fact about this subject that their confidence in what they believe is quite variable, even though there is no epistemic basis for this variability (it is not as if, for example, they have reasons to doubt their memory). It will now be true of such a subject that in close possible worlds they will have the same basis for belief as in the actual world and yet will not believe the target proposition on this basis. It would thus follow, on the interpretation of safety motivated via anti-luck epistemology, that our subject lacks knowledge. But that doesn’t seem like the right result at all. After all, in all close possible worlds where they form a belief on this sound memorial basis their belief will be true. So why should their psychological reticence to form beliefs on this basis in some close possible worlds deprive them of knowledge? Indeed, as noted above, one could argue that, ceteris paribus anyway, cautious believers like our subject are to be preferred, from an epistemic point of view, to those who are far too quick to form beliefs. Could we get around this problem by stipulating that a basis for belief should be understood in such a way that it entails that the subject forms a belief? That would certainly fix the problem, but it is hard to see what motivation there would be for this requirement, other than the post hoc one of evading this difficulty. On any plausible rendering of the basing relation, one can surely be in possession of the basis and, in principle anyway, form any doxastic attitude about the target proposition. 



Note that it is important to the case that the cause of the self-doubt is purely psychological, as otherwise we would have grounds for thinking that the basis for belief is changing. This example is an adaption of Radford’s () ‘diffident schoolboy’ case, albeit to illustrate a very different point. There may be some versions of the epistemic basis relation that wouldn’t allow this, such as Swain’s () counterfactual account. But I think that this would be a problem for such a view, and hence, if true, would demand revision of the proposal. For more on the epistemic basing relation, see Neta

Anti-risk Virtue Epistemology



That same basis could just as much be a basis for disbelief in this proposition as belief, for example. But if that’s right, then the presence of the basis must also be compatible with not forming a belief in the target proposition, given that disbelief entails, but is not entailed by, a lack of belief in the proposition. We can evade this problem if we shift to an anti-risk epistemology. For our concern now is no longer with the modal closeness of the nonobtaining of the event in question (forming a true belief ), but rather with the modal closeness of the risk event. When it comes to knowledge, I think the risk event is very clear – it’s that the very same basis for belief leads you into error (i.e. false belief ). That is, we want a safe basis for belief that will ensure that we can’t very easily end up in error. So, this problem disappears according to an anti-risk epistemology, as we now have a principled way of focusing on the modal closeness of the basis leading to false belief, specifically, rather than either false belief or non-belief. In fact, the move to anti-risk epistemology also better motivates the way that safety-based theorists like myself respond to the problem posed by necessary propositions. The general contours of the puzzle is that one could form a belief in a necessary proposition in a completely haphazard way that has no epistemic credentials at all – guesswork, say – but that it would inevitably be safe nonetheless because there is no close possible world where one can believe such a proposition and believe it falsely (since there is no possible world where it is false). But this appears wrong, in that we would want to say that any belief formed in this way is unsafe; it just seems a matter of luck that one happened on the right answer. The way safety theorists like myself respond to this is to say that once we shift to a basis-relative formulation of safety – which, recall, was independently motivated in terms of an anti-luck epistemology – then our focus should not be on the particular proposition believed in the actual world, but





 and Korcz . See also Bondy and Pritchard , which makes the case for a new kind of epistemic luck regarding the basing relation, and motivates the proposal via appeal to anti-risk epistemology. In fact, they don’t have to be necessary for the problem to arise. So long as the truth of the proposition is modally stable across close possible worlds, then the same problem will arise, even if the proposition in question is contingent. It would be different if the world were engineered to guarantee you true beliefs within a certain domain (e.g. by a helpful demon). Now I would grant that your beliefs are safe, though they still aren’t knowledge. Indeed, the ‘temp’ case that I offer as part of a critique of (what I call) ‘pure’ (or ‘robust’) anti-luck epistemology (i.e. an account of knowledge that only has an anti-luck epistemic condition) is meant to illustrate that one can have beliefs that are bound to be true (and hence are not luckily true) but which do not amount to knowledge. For discussion, see Pritchard, Millar, and Haddock : ch.  and Pritchard a.



 

rather on the doxastic output of the basis in the actual world instead. This means that while there is, of course, no close possible world where one falsely believes the necessary proposition that one actually believes, the kind of haphazard basis described above will lead to lots of false beliefs in close possible worlds (just not false beliefs in the proposition actually believed). As such it will be an unsafe basis. This certainly fixes the problem. But the motivation for this formulation of safety from an anti-luck epistemology point of view is rather weak. Yes, this licenses a focus on a basis-relative account of safety, but notice that one could have such an account and nonetheless insist on it being a basis for belief in the proposition actually believed, so there is still a motivational lacuna in play here. In particular, what aspect of anti-luck epistemology is meant to motivate the loosening of this restriction so that our focus should be on the doxastic output of the actual basis more generally in close possible worlds, and not simply on whether that basis leads to a false belief in close possible worlds regarding the proposition actually believed? This lacuna disappears once we adopt anti-risk epistemology. For notice that we have explicitly formulated the risk event in terms of the basis leading us into error (false belief ). If that’s our focus, however, then of course we will not want to restrict our attentions only to error apropos the proposition actually believed, for then we will miss out on the wider errors that a bad basis can lead us to, as when a haphazard approach to beliefformation leads to lots of false beliefs in close possible worlds. Finally, notice that anti-risk epistemology enables a much wider range of epistemic assessments than anti-luck epistemology. This is because, as noted above, we can introduce a range of different risk events when making a risk assessment, including considering multiple risk events within a single assessment. When it comes to knowledge, as I just noted, it seems clear that there is a single risk event that concerns us, which is that our basis for belief will not lead us into error (false belief ) – that’s the risk event that we want to be modally distant in order for one to count as a knower. But once we turn our attentions away from knowledge, specifically, then we might have other risk events in mind, including multiple risk events. Think about good inquiries, for example. Sure, ending up with a false belief via a particular basis might well be one of the core risk events to avoid in a good inquiry, but it needn’t be the full story. It could be, for example, that the good inquirer also needs to come up with an answer 

I develop this line in a number of places. See, for example, Pritchard a, b. See also Pritchard b.

Anti-risk Virtue Epistemology



within a reasonable timeframe, such that another risk event, to weigh-up against the former, is that one altogether fails to resolve one’s inquiry. Sometimes a suboptimal solution is better than no solution at all, after all. One could regard the good inquirer as trading these two risk events off against each other, and so forming a judgement about what the right lines of inquiry should be. That is, one might be willing to be sanguine about a higher level of risk that one’s inquiry leads to error than is normal in order to be more confident that one will avoid the other risk scenario of failing to resolve the inquiry. This would, of course, add a pragmatic element into the nature of good inquiry, but I take it that this isn’t particularly controversial, unlike a pragmatic encroachment thesis in the theory of knowledge. On a traditional account of the nature of the virtues, after all, the intellectual virtues are entwined with the practical and moral virtues, and so one would not expect the manifestation of the former to be in isolation from the latter. More precisely, while a particular intellectually virtuous line of inquiry, once undertaken, might be uninfluenced by purely pragmatic factors, the question of which line of inquiry to undertake, and relatedly the criteria under which that inquiry is to be evaluated, may well be highly influenced by practical factors. In this way, one’s intellectual virtues can work in concert with one’s other virtues to promote one’s (overall) flourishing. Similarly, the question of which risk event or events are relevant to conducting an inquiry might be a purely practical consideration, even if the inquiry itself is undertaken in a purely intellectual spirit. Pragmatic encroachment about inquiry of this kind is nothing to be concerned about. Note too that the use of multiple risk events is especially salient once we think about inquiries as collaborative endeavours (as they often are). Consider a particular realm of scientific inquiry. It might be overall epistemically very beneficial that within that realm there are scientists taking varying degrees of epistemic risk, with some being very risk-adverse 



For more on pragmatic encroachment in the theory of knowledge, see Hawthorne ; Stanley ; and Fantl and McGrath . Note that Ballantyne ,  has argued that anti-luck epistemology is committed to pragmatic encroachment about knowledge – a claim that, if sound, would plausibly carry over to anti-risk epistemology. Ballantyne’s argument, however, depends on an earlier version of the modal account of luck that I offered which included a significance condition (e.g., Pritchard ). This extra condition is now dropped, for reasons that I articulate in Pritchard . Similarly, there is no significance condition in the modal account of risk that I offer either, and hence there is no basis (in this particular regard anyway) for thinking that either anti-luck or anti-risk epistemology is committed to a pragmatic encroachment thesis about knowledge. For an influential neo-Aristotelian account of the virtues along these lines, with a particular focus on the nature of the intellectual virtues and their role within the virtues more generally, see Zagzebski .



 

and so pursuing relatively safe lines of inquiry (such that they are likely to get the results they want, but those results are rarely going to be significant or surprising), and others taking large risks with very speculative lines of inquiry, with the potential of discovering important scientific truths (but more often than not discovering nothing of the kind). We can make sense of this very easily within an anti-risk epistemology, since what is taking place here is effectively a trade-off when it comes to the relative weights of different epistemic risk events. This last point demonstrates that while anti-luck epistemology delivers a theory of knowledge (in anti-luck virtue epistemology), anti-risk epistemology goes one stage further and offers us not just a theory of knowledge (anti-risk virtue epistemology, as explored in Section .), but also a broader way of approaching epistemic issues. In particular, we now have a richer way of evaluating epistemic standings in terms of a range of epistemic risk events, including even the possibility of multiple epistemic risk events, or even epistemic risk events considered from a specifically social perspective.

. Anti-risk Virtue Epistemology I noted above that I have argued elsewhere that anti-luck epistemology generates a theory of knowledge that I have termed anti-luck virtue epistemology. Anti-luck epistemology by itself cannot offer a complete theory of knowledge but can only deliver us the anti-luck condition on knowledge. In short, the reason for this is that such a condition is only going to give you a certain modal profile for one’s belief. But that one’s belief has this modal profile does not suffice to demonstrate that your cognitive success (i.e. true belief ) is in any significant way down to your manifestation of relevant cognitive agency, something that I take to be integral to our concept of knowledge, as illustrated by the ability platitude noted above. In particular, we can imagine cases where external factors engineer it to be the case that there is no luck involved in one having true beliefs, and so one’s beliefs are safe, but since one’s cognitive success has nothing to do with one’s manifestation of relevant cognitive agency, then one wouldn’t count as a knower. This means that there needs to be an ability, or virtue, condition on knowledge, a condition that demands that there is the salient explanatory link between one’s cognitive success and one’s manifestation of relevant cognitive agency. 

For more on this point, see Pritchard, Millar, and Haddock : ch.  and Pritchard a.

Anti-risk Virtue Epistemology



Going in the other direction, I’ve also argued that one cannot account for knowledge exclusively in terms of a virtue-theoretic condition. That is, one can’t simply subsume anti-luck epistemology within a broader virtuetheoretic account of knowledge, as a number of virtue theorists have tried to do (I call them robust virtue epistemologists). Very roughly, robust virtue epistemologists hold that knowledge is cognitive success that is because of the manifestation of relevant cognitive agency, where this means that the former is primarily attributable to the latter. The thought is that while Gettier-style cases clearly do involve virtuous agency and cognitive success, they don’t enjoy the relational quality of being a cognitive success that is because of one’s virtuous agency. In short, they are not cognitive achievements, where this means a cognitive success that is because of one’s manifestation of relevant cognitive agency (and which is a subclass of the broader category of achievements simpliciter – successes that are because of one’s manifestation of relevant agency). In any case, the claim is that once one adds this relational ingredient into the mix then virtue epistemology can deal with the problem posed by epistemic luck without needing to appeal to anti-luck epistemology. I maintain that this is mistaken. I originally argued for this point by appealing to a distinction between the kind of intervening epistemic luck that is familiar from standard Gettierstyle cases, and the environmental epistemic luck that one finds in specifically barn-facade-type cases. In standard Gettier-style cases something intervenes between one’s rational basis for the target proposition and the fact itself, though one’s belief is true nonetheless. So, for example, one reasonably thinks that one is seeing a sheep in the field, and there is a sheep in the field, but in fact what one is looking at is a big hairy dog that is obscuring from view the real sheep behind. Cases of environmental epistemic luck are different, however, in that nothing intervenes in this 





This is how Greco , b, , a, c understands the ‘because of’ relation, though see Greco  for a reworking of his view. The other main proposal in the literature in this regard is due to Sosa , , , , . This construes the ‘because of’ relation in terms of disposition manifestation. That a glass is shattered when hit, for example, could be because it is fragile, where this kind of explanation need not be in competition with a causal explanatory story (e.g. that so-and-so lost his temper and threw the glass at the wall). For specific discussion of Sosa’s account, see Pritchard a and Kallestrup and Pritchard . See also Zagzebski , , who treats the ‘because of’ relation as an indefinable primitive. In order to keep the discussion to a manageable length, I will be focusing on the causal-explanatory construal of the ‘because of’ relation. I introduced this distinction, and the associated terminology, in Pritchard a, b, c. I have further discussed this distinction in a number of places, including Pritchard, Millar, and Haddock : chs –, Pritchard a, a, and Kallestrup and Pritchard . This is Chisholm’s (: ) famous Gettier-style case.



 

way. In the barn facade case, for example, one really does see a genuine barn, it is just that one is in an environment that ensures that one’s belief is nonetheless unsafe. Here is the thing: environmental luck (epistemic or otherwise) is entirely compatible with one’s success (cognitive or otherwise) being primarily attributable to the manifestation of one’s relevant (cognitive or otherwise) agency. Consider shooting an arrow at a target. If there is merely intervening luck involved, such that you skilfully took the shot, and was successful, but the latter was just down to an intervention (e.g. a dog running on, grabbing the bolt, and putting it in the target), then this success is not primarily attributable to the manifestation of your agency. It is not, as we say, your achievement at all, even though you performed well. But now consider a parallel case of environmental luck. Everything is the same except this time the dog failed at the last moment to intercept the bolt. In close possible worlds the dog would have intercepted it, it’s just that he failed in the actual world. Your success is modally fragile, and thus unsafe. Nonetheless, isn’t this success now primarily down to your manifestation of relevant cognitive agency? After all, while something could have intervened – this is what ensures that the success is unsafe – in fact it didn’t. This means that environmental luck (epistemic or otherwise) is compatible with genuine achievements. It also means, in turn, that achievements (cognitive or otherwise) can be unsafe. But since knowledge demands safety, this entails that we cannot equate knowledge with cognitive achievements, as the proponent of robust virtue epistemology proposes. These days I make essentially the same point in terms of the notion of an epistemic twin earth case, which is effectively a way of sharpening up the earlier distinction between intervening and environmental epistemic luck. Imagine two counterpart agents, one on earth and one on twin earth, who are microphysical duplicates of each other, with identical causal histories. The environment that they are presently causally interacting with is also identical in every respect. In addition, the ‘normal’ environment – i.e. the sorts of things that they would normally be causally interacting with – is also identical for both subjects. Now suppose that both agents form, on the same basis, the belief that p. All that is different with regards to the two agents is that the agent on twin earth occupies a very different modal environment. For whereas the agent on earth is forming a belief that p in such a way that this same basis for belief will generate true beliefs across all close possible worlds, due to idiosyncratic features of their modal 

See Kallestrup and Pritchard . See also Pritchard a.

Anti-risk Virtue Epistemology



environmental the agent on twin earth is forming a belief that p in such a way that the very same basis for belief will generate false beliefs in close possible worlds. What is interesting about epistemic twin earth cases is that we have effectively kept fixed across the two subjects any possible factor that might be relevant to the manifestation of cognitive agency. After all, one can imagine that manifestations of agency – cognitive or otherwise – might be influenced by such factors as one’s actual causal environment, one’s normal causal environment, one’s causal history, or one’s microphysical nature. But no-one holds that manifestations of agency – again, cognitive or otherwise – are influenced by factors that are exclusive to one’s modal environment. And yet the agent on twin earth is forming their belief in such a way that it is unsafe, in contrast to the agent on earth who is forming an identical belief. Insofar as we grant that knowledge is incompatible with veritic epistemic luck (and hence cannot be unsafe), then we should be inclined to treat the agent’s belief on twin earth as not being knowledge. I take the forgoing to indicate that the anti-luck and ability platitudes about knowledge, noted above, in fact impose distinct constraints on a theory of knowledge, rather than, as is (implicitly) supposed, the same constraints. One can see why they might be thought to impose the same constraints. After all, in general at least, the reason why one’s true belief is due to luck (and hence fails to satisfy the anti-luck constraint) is that one’s true belief had nothing to do with one’s manifestation of relevant cognitive agency. Going in the other direction, the usual reason why one’s true belief was not because of one’s manifestation of relevant cognitive agency is that it was due to luck. But this prima facie account of how the two intuitions relate has been shown to be problematic. On the one hand, we find that there are cases where agents have beliefs that have the right modal profile to count as nonlucky (safe), but which don’t satisfy the ability constraint on knowledge. On the other hand, epistemic twin earth cases, and the phenomenon of environmental epistemic luck more generally, show that an agent’s cognitive success can be primarily creditable to their manifestation of cognitive agency and yet nonetheless be lucky (unsafe). The upshot is that we need an account of knowledge that recognises the way in which these overlapping constraints on one’s theory sometimes diverge. 

For further discussion of cases involving purely modal veritic epistemic luck, see Pritchard a: §.



 

This is where anti-luck virtue epistemology came in. Roughly, this argues that knowledge is safe cognitive success that is significantly attributable to one’s manifestation of relevant cognitive agency. Note that this is both in a sense stronger and in a sense weaker than robust virtue epistemology. The former, because the explanatory relation is now between safe cognitive success and the manifestation of relevant cognitive agency rather than just cognitive success. The latter, because we are now only demanding that this explanatory relation satisfy a significant level rather than the cognitive success being primarily attributable to the cognitive agency of the subject. Such a proposal can capture the way our two core platitudes intersect with one another. Sometimes even a high manifestation of cognitive agency, of a kind that would ordinarily easily suffice for knowledge, can fail to bring one knowledge due to purely environmental factors that make the cognitive success that results nonetheless unsafe. And sometimes a relatively low manifestation of cognitive agency, of a kind that wouldn’t ordinarily deliver knowledge, can suffice for knowledge on account of the fact that one is in the kind of epistemically friendly environment that ensures that the belief is safe nonetheless. As I’ve argued elsewhere, both kinds of scenarios are uniquely accommodated by an anti-luck virtue epistemology. Crucially, anti-risk virtue epistemology, which is just anti-luck virtue epistemology with an anti-risk construal of safety, will inherit all the benefits of anti-luck virtue epistemology, while facing none of the problems that we saw afflicted the motivation of safety on anti-luck grounds. In particular, the kinds of theoretical lacunae that we noted as regards antiluck epistemology will not infect anti-risk virtue epistemology. But this is not the only advantage to replacing an anti-luck virtue epistemology with an anti-risk virtue epistemology. In order to see this, we need to note that, unsurprisingly, the proponents of robust virtue epistemology are not convinced that they should abandon their view and endorse such a two-aspect proposal. As noted 



Robust virtue epistemology needed the stronger claim in this regard in order to deal with Gettierstyle cases, but we don’t face that hurdle, as of course we have safety built into the proposal from the off. This part of the proposal helps us to accommodate cases of knowledge where one’s cognitive success is primarily down to the cognitive agency of others, as is quite common in testimonial cases. See, for example, Pritchard a for discussion of both points. In particular, I claim that these two scenarios are both cases of epistemic dependence, where what I mean by that is how knowledge can be significantly influenced by factors outwith one’s cognitive agency. The first kind of scenario is negative epistemic dependence, whereby environmental factors prevent an otherwise high manifestation of cognitive agency from being knowledge. The second kind of scenario is positive epistemic dependence, whereby environmental factors enable an otherwise low manifestation of cognitive agency to count as knowledge. See, for example, Pritchard a for the details.

Anti-risk Virtue Epistemology



previously, epistemic twin earth cases bring the phenomenon of environmental epistemic luck, and the problem it poses for robust virtue epistemology, into sharp relief. Such cases effectively make it impossible to evade the problem posed by environmental epistemic luck to robust virtue epistemology. Interestingly, however, the response from robust virtue epistemologists on this score is increasingly not to try to evade the problem at all, but rather to meet it head-on by arguing that one can coherently allow that knowledge is compatible with environmental epistemic luck. Accordingly, they argue that knowledge can be compatible with veritic epistemic luck after all (of this particular variety anyway). You can see why they might take this line. After all, their concern is with cognitive achievements – i.e. cognitive successes that are because of cognitive agency – so if environmental epistemic luck doesn’t undermine cognitive achievements, then why care about it? Why not insist that knowledge is a cognitive achievement, and hence is compatible with environmental epistemic luck? Relatedly, surely everyone will agree that as veritic epistemic luck goes intervening epistemic luck is much worse, from an epistemic point of view, than environmental epistemic luck. After all, in the latter case, there is no mismatch between one’s reasons and the facts that one finds in intervening epistemic luck. This is where the shift to anti-risk epistemology, and thus to an anti-risk virtue epistemology, becomes important. One might be sanguine about one’s theory of epistemology ascribing knowledge in cases where there is a high level of epistemic luck. But can one really be sanguine about one’s theory of knowledge ascribing knowledge in cases where there is a high level of epistemic risk? Cases of environmental epistemic luck involve levels of epistemic risk that are on a par with cases of intervening epistemology luck, after all, and even robust virtue epistemologists are inclined to treat the latter as problematic from the perspective of ascribing knowledge. So how can they now consistently argue that we can be sanguine about one form of high epistemic risk if they themselves agree that corresponding levels of risk are elsewhere incompatible with knowledge? Moreover, recall our point that we tend to care about (bad) luck because we care about risk, rather than vice versa. This means, in the epistemic case, that we tend to worry about epistemic luck because we are worried about epistemic risk, rather than vice versa. Accordingly, it is not the high 

See, in particular, Sosa’s (: ch. ) discussion of the ‘jokester’ case, which has been very influential. I critically discuss Sosa’s reasons for ascribing knowledge in this case in Pritchard a.



 

levels of epistemic luck, specifically, that we should focus upon, but rather the high levels of epistemic risk. With the attentions of robust virtue epistemologists duly focused on the latter, it is hard to see how they could consistently argue that they are content to allow knowledge to be compatible with this high level of epistemic risk, particularly when elsewhere they maintain that such a level of epistemic risk is incompatible with knowledge. The upshot is that once anti-risk epistemology, and hence anti-risk virtue epistemology, is on the table, then it is far harder for proponents of robust virtue epistemology to shrug-off the high levels of epistemic risk involved in cases of environmental epistemic luck. Anti-risk epistemology also gives us a very specific handle on how to think about post-Gettier epistemology. Recall that we noted above that luck is essentially a backward-facing notion, in contrast to risk, which is by its nature a forwards-looking notion. This point is important to diagnosing the direction post-Gettier epistemology took. For notice that Gettier-style cases are by their nature examples whereby the agent was cognitively successful, and we are reflecting back on the unusual, happenstance, way in which they were cognitively successful. It is thus no surprise that our focus becomes on the role of luck in this cognitive success, and thus on epistemic luck. But if I am right that our interest in eliminating epistemic luck from knowledge is derivative on our interest in eliminating (high levels of ) epistemic risk, then this backwards-focused nature of Gettier-style assessments led us to focus on quite the wrong thing, particularly given the (subtle, yet important) differences between luck and risk, and hence between epistemic luck and epistemic risk. In particular, had we spotted this distinction earlier, and clocked the need to examine these cases from the forwards-looking perspective provided by risk, then we would have been better placed to resolve the problems in hand (i.e. not just the narrow Gettier problem of coming up with an anti-luck/risk condition on knowledge, but also the broader Gettier problem of developing an adequate theory of knowledge).

. Conclusion We have seen that the shift from anti-luck epistemology to anti-risk epistemology, while on the face of it a rather marginal change, in fact  

I first made this point in Pritchard b. See also Pritchard a. I distinguish between these two interpretations of the ‘Gettier problem’, and offer resolutions of these problems – albeit via an anti-luck, rather than an anti-risk, epistemology – in Pritchard a, b.

Anti-risk Virtue Epistemology



offers us important new theoretical resources, on account of the small – but crucial nonetheless – differences between the notions of luck and risk. Anti-risk epistemology better motivates certain epistemic moves that we need to make, thereby removing significant motivational lacunae that were present in anti-luck epistemology. It also offers a new approach to epistemic questions that allows us to broaden our scope of epistemic evaluations by focusing upon a range of different, and potentially multiple, epistemic risk events. But perhaps most importantly, we have seen that this change of direction enables us to offer a more compelling virtue-theoretic account of knowledge. As we have seen, anti-risk virtue epistemology inherits all the advantages of anti-luck virtue epistemology but none of its flaws. Crucially, it presents us with the means to better resist the current fashion of taking a more permissive line on unsafe beliefs by illustrating how this commits one to implausible claims about epistemic risk, while at the same time providing us with a better diagnostic handle on why post-Gettier epistemology panned in out in quite the manner that it did. So, while the shift from anti-luck epistemology to anti-risk epistemology may seem of minor philosophical significance at first blush, on close inspection we see that it delivers important theoretical gains. REF ERE NCE S Ballantyne, N. . ‘Anti-luck Epistemology, Pragmatic Encroachment, and True Belief’, Canadian Journal of Philosophy : –. . ‘Luck and Interests’, Synthese : –. Bondy, P., and Pritchard, D. H. . ‘Propositional Epistemic Luck, Epistemic Risk, and Epistemic Justification’, Synthese (Online first, DOI: ./ s---). Broncano-Berrocal, F. . ‘Luck as Risk and the Lack of Control Account of Luck’, Metaphilosophy : –. Chisholm, R. . Theory of Knowledge (nd ed.). Englewood Cliffs, NJ: Prentice Hall. Coffman, E. J. . Luck: Its Nature and Significance for Human Knowledge and Agency. London: Palgrave Macmillan. Fantl, J., and McGrath, M. . Knowledge in an Uncertain World. Oxford: Oxford University Press. 

Thanks to Jesus Navarro for helpful discussions on this topic, and to John Greco and Chris Kelp for detailed comments on an earlier draft. An earlier version of this chapter was presented at a workshop on contemporary issues in epistemology at the University of California, Irvine in March , and at a conference on ‘New Trends in Virtue Epistemology’ at the University of Glasgow in September . I am grateful to the audience for their questions on both occasions, and especially Michael Brady, who was the commentator on my paper at the Glasgow conference.



 

Greco, J. . ‘Knowledge as Credit for True Belief’, in Intellectual Virtue: Perspectives from Ethics and Epistemology, M. DePaul and L. Zagzebski (eds). Oxford: Oxford University Press, –. a. ‘The Nature of Ability and the Purpose of Knowledge’, Philosophical Issues : –. b. ‘Worries about Pritchard’s Safety’, Synthese : –. . ‘What’s Wrong with Contextualism?’, Philosophical Quarterly : –. a. Achieving Knowledge. Cambridge, UK: Cambridge University Press. b. ‘Knowledge and Success from Ability’, Philosophical Studies : –. c. ‘The Value Problem’, in Epistemic Value, A. Haddock, A. Millar, and D. H. Pritchard (eds). Oxford: Oxford University Press, –. . ‘A (Different) Virtue Epistemology’, Philosophy and Phenomenological Research : –. Hansson, S. O. . ‘Philosophical Perspectives on Risk’, Techné: Research in Philosophy and Technology : –. . ‘Risk’, Stanford Encyclopedia of Philosophy, E. N. Zalta (ed.). http://plato .stanford.edu/archives/spr/entries/risk/. Hawthorne, J. . Knowledge and Lotteries. Oxford: Clarendon Press. Kahneman, D., and Varey, C. A. . ‘Propensities and Counterfactuals: The Loser That Almost Won’, Journal of Personality and Social Psychology : –. Kallestrup, J. and Pritchard, D. H. . ‘Virtue Epistemology and Epistemic Twin Earth’, European Journal of Philosophy : –. Korcz, K. A. . ‘The Epistemic Basing Relation’, Stanford Encyclopedia of Philosophy, E. N. Zalta (ed.). https://plato.stanford.edu/entries/basing-episte mic/. Lackey, J. . ‘What Luck Is Not’, Australasian Journal of Philosophy : –. Levy, N. . Hard Luck: How Luck Undermines Free Will and Moral Responsibility. Oxford: Oxford University Press. Neta, R. . ‘The Basing Relation’, The Routledge Companion to Epistemology, S. Bernecker and D. H. Pritchard (eds). New York: Routledge, –. Pritchard, D. H. . ‘Epistemic Luck’, Journal of Philosophical Research : –. . Epistemic Luck, Oxford: Oxford University Press. . ‘Anti-luck Epistemology’, Synthese : –. . ‘Sensitivity, Safety, and Anti-luck Epistemology’, in Oxford Handbook of Scepticism, J. Greco (ed.). Oxford: Oxford University Press, –. a. ‘Apt Performance and Epistemic Value’, Philosophical Studies : –. b. Knowledge. London: Palgrave Macmillan. c. ‘Knowledge, Understanding and Epistemic Value’, Epistemology (Royal Institute of Philosophy Lectures), A. O’Hear (ed.). Cambridge, UK: Cambridge University Press, –.

Anti-risk Virtue Epistemology



a. ‘Anti-luck Virtue Epistemology’, Journal of Philosophy : –. b. ‘In Defence of Modest Anti-luck Epistemology’, The Sensitivity Principle in Epistemology, T. Black and K. Becker (eds). Cambridge, UK: Cambridge University Press, –. . ‘The Modal Account of Luck’, Metaphilosophy : –. a. ‘Anti-luck Epistemology and the Gettier Problem’, Philosophical Studies : –. b. ‘Risk’, Metaphilosophy : –. a. ‘Epistemic Dependence’, Philosophical Perspectives : –. b. ‘Epistemic Risk’, Journal of Philosophy : –. a. ‘Anti-risk Epistemology and Negative Epistemic Dependence’, Synthese (Online first: https://doi.org/./s–--). b. ‘Knowledge, Luck and Virtue: Resolving the Gettier Problem’, in Explaining Knowledge: New Essays on The Gettier Problem, C. Almeida, P. Klein, and R. Borges (eds). Oxford: Oxford University Press, –. c. ‘Legal Risk, Legal Evidence, and the Arithmetic of Criminal Justice’, Jurisprudence (Online first, DOI: ./..). . ‘Aesthetic Risk’, Think : –. Pritchard, D. H., Millar, A., and Haddock, A. . The Nature and Value of Knowledge: Three Investigations. Oxford: Oxford University Press. Pritchard, D. H., and Smith, M. . ‘The Psychology and Philosophy of Luck’, New Ideas in Psychology : –. Pritchard, D. H., and Whittington, L. J. (eds). . The Philosophy of Luck. Oxford: Blackwell. Radford, C. . ‘Knowledge – By Examples’, Analysis : –. Riggs, W. . ‘Why Epistemologists Are so down on Their Luck’, Synthese : –. . ‘Luck, Knowledge, and Control’, Epistemic Value, A. Haddock, A. Millar, and D. H. Pritchard (eds). Oxford: Oxford University Press, –. Sainsbury, R. M. . ‘Easy Possibilities’, Philosophy and Phenomenological Research : –. Sosa, E. . Knowledge in Perspective: Selected Essays in Epistemology, Cambridge, UK: Cambridge University Press. . ‘How to Defeat Opposition to Moore’, Philosophical Perspectives : –. . A Virtue Epistemology: Apt Belief and Reflective Knowledge. Oxford: Oxford University Press. . Reflective Knowledge: Apt Belief and Reflective Knowledge. Oxford: Oxford University Press. . Knowing Full Well. Princeton, NJ: Princeton University Press. . Judgment and Agency. Oxford: Oxford University Press. Stanley, J. . Knowledge and Practical Interests. Oxford: Oxford University Press. Swain, M. . Reasons and Knowledge. Ithaca, NY: Cornell University Press.



 

Teigen, K. H. . ‘How Good Is Good Luck?: The Role of Counterfactual Thinking in the Perception of Lucky and Unlucky Events’, European Journal of Social Psychology : –. . ‘Luck: The Art of a Near Miss’, Scandinavian Journal of Psychology : –. . ‘Luck, Envy, Gratitude: It Could Have Been Different’, Scandinavian Journal of Psychology : –. a. ‘Hazards Mean Luck: Counterfactual Thinking and Perceptions of Good and Bad Fortune in Reports of Dangerous Situations and Careless Behaviour’, Scandinavian Journal of Psychology : –. b. ‘When the Unreal if More Likely Than the Real: Post Hoc Probability Judgements and Counterfactual Closeness’, Thinking and Reasoning : –. . ‘When a Small Difference Makes a Large Difference: Counterfactual Thinking and Luck’, in The Psychology of Counterfactual Thinking, D. R. Mandel, D. Hilton, and P. Catellani (eds). London: Routledge, –. Tetlock, P. E. . ‘Close-Call Counterfactuals and Belief-System Defenses: I Was Not Almost Wrong but I Was Almost Right’, Journal of Personality and Social Psychology : –. Tetlock, P. E. and Lebow, R. N. . ‘Poking Counterfactual Holes in Covering Laws: Cognitive Styles and Historical Reasoning’, American Political Science Review : –. Williamson, T. . Knowledge and Its Limits. Oxford: Oxford University Press. Zagzebski, L. . Virtues of the Mind: An Inquiry into the Nature of Virtue and the Ethical Foundations of Knowledge. Cambridge, UK: Cambridge University Press. . ‘What Is Knowledge?’, Blackwell Guide to Epistemology. J. Greco and E. Sosa (eds). Oxford: Blackwell, –.

 

Responsibilism within Reason Kurt Sylvan

.

Introduction

According to ambitious responsibilism (AR), the virtues that are constitutive of epistemic responsibility should play a central and fundamental role in traditional projects like the analysis of justification and knowledge. While AR enjoyed a shining moment in the mid-s, it has fallen on hard times. Part of the reason is that many epistemologists – including fellow responsibilists – think it paints an unreasonably demanding picture of knowledge and justification. While AR’s defenders have responses to this worry, they tend either to collapse AR into a much less ambitious view, or to threaten virtue’s explanatory force in AR’s analyses. I agree that such objections undermine AR’s existing versions. But I think it would be premature to draw the curtains on the view. My goal is to show that the stock objections only threaten the periphery of certain versions of AR, and to develop a novel version that avoids them. With this goal in mind, here is the plan. I will begin in Section . by clarifying the core commitments of AR and explain how influential responsibilists have added to these commitments in optional ways. In Section ., I will rehearse the standard objections to AR, explaining why they only target optional accretions. I’ll then turn in Section . to develop a version I call Kantian Responsibilism (KR). KR is a two-level view, consisting of (i) a high-level analysis of epistemic normativity in responsibilist terms, and (ii) a first-order account of the conditions under which these terms apply. According to KR’s first tier, epistemically virtuous thought is thought that 

This is my shorter and less loaded name for what Baehr () calls strong conservative responsibilism. Note that Baehr usefully distinguished weak and strong as well as autonomous and conservative responsibilism. Weak versions hold that responsibilist virtue properties play an important but auxiliary role in epistemology, while strong versions hold that they play a central role; autonomous versions hold that traditional projects in epistemology ought to be replaced by independent inquiry into intellectual virtue, while conservative versions maintain that virtue epistemology can contribute to traditional projects.





 

manifests respect for truth; because I hold that manifesting certain reasonssensitive dispositions is necessary and sufficient for respecting truth, KR’s second tier takes epistemic virtues to coincide substantively with reasonssensitive dispositions. After unpacking KR in Section ., I show in Section . how it answers the objections to AR. I close in Section . by making some broader points about KR’s virtues, especially when compared with reliabilist virtue epistemology.

. AR: Core and Periphery Let’s begin by recalling how responsibilism developed in response to the early reliabilist virtue epistemology of Ernest Sosa. When first formulated by Sosa (), virtue epistemology (VE) was offered as a contribution to traditional projects concerning the nature and structure of knowledge and justification. In this formulation, VE was not sharply distinguished from the simple reliabilism of Goldman (); indeed, the title of Sosa’s (: ) key section was ‘Reliabilism: an ethics of moral virtues and an epistemology of intellectual virtues.’ This feature prompted a demand for a stronger break with tradition, with Code (: ) suggesting in responsibilism’s first articulation that we should ‘offer a different approach to epistemology, based upon an entirely different set of expectations from those that have long constituted the fundamental motivations of foundationalism and coherentism,’ and hold that traditional epistemology’s ‘methodologies, but also . . . the presuppositions and expectations that underlie them, are misguided’. Kvanvig () echoed Code’s thought more seditiously, recommending that we ditch traditional epistemology for the independent study of epistemic virtue. Early responsibilism’s radicalism was tempered in mid-s by Zagzebski’s landmark Virtues of the Mind, which offered accounts of knowledge and justified belief meant to rival Sosa (). Zagzebski () remains AR’s high watermark. Unfortunately, however, this watermark has been maintained partly because many virtue epistemologists have rejected Zagzebski’s project. Some have returned to the autonomous VE of Code and Kvanvig. Others have sought compromises. Baehr () advocates a ‘weak conservative’ responsibilism that agrees on the importance of traditional epistemological projects yet holds that responsibilist virtue has a merely auxiliary role in them. Some reliabilist virtue epistemologists, on the other hand, now deny that there is a real tension between their projects and 

See Roberts and Wood .

Responsibilism within Reason



responsibilists’, recommending a merger. For example, Greco () invokes both responsibility and reliability in his account of justification, while Sosa () suggests that responsibilist themes follow from reliabilist VE properly understood. While the hope of defending an undiluted responsibilism as ambitious as Zagzebski’s has disappeared, I think this pessimism rests on mistaking peripheral features of certain versions of AR for core features. None of the objections to Zagzebski () undermine AR when core and periphery are distinguished, as we’ll see. .. The Core As I understand it, AR’s fundamental claim is simple: (ARC) Responsibilist virtue properties should play a central and fundamental role in the analysis of epistemically normative properties.

Three parts of this claim require unpacking, however, raising three questions for AR: Q: What are ‘responsibilist virtue properties’? Q: What does it take to play a ‘central and fundamental role’? Q: What counts as an epistemically normative property? I think many objections to ARC rest on false assumptions about how we should answer Q–. On Q. To understand what ARC says, we need minimal answers to Q– that existing defenders of ARC and I would both accept. Proceeding backwards from Q, we can start with the truism that epistemically normative properties are a special case of normative properties, where ‘normative’ is understood broadly to include the normativity associated with standards of correctness that fall short of all-things-considered significance, like norms of prescriptive grammar and etiquette. Normative properties across domains then come in various flavors. There are evaluative properties, including thin ones like goodness and badness, and thick ones like courageousness, shrewdness, and grubbiness. There are also deontic properties, most obviously including permissibility and obligation, 

‘Normative’ is sometimes used more narrowly for standards we ought objectively to care about; see, e.g., Kolodny  and Broome . It is unclear whether epistemology is normative in this sense. Early responsibilists like Code () took it to be a virtue of their view that it made epistemology normative in this sense, but I remain unsure.



 

but also justification in my book. Finally, there are what Zimmerman () calls hypological properties – blameworthiness, excusability, and praiseworthiness. Normative properties and relations are either paradigmatic members of one of these categories, or ones analyzable in terms of these. Epistemically normative properties and relations are ones whose normativity is tied to the constitutive aims or norms of believing. Note that when epistemically normative properties and relations are understood in this way, it is not so clear that knowledge is normative. Knowledge might well be the norm of belief, in the sense of being what we need in order to believe permissibly. But something can be the norm of something else without being normatively constituted. Pleasure might be the norm of desire, for example, but it is just a psychological state. Similarly, it is compatible with treating knowledge as the norm of belief to hold that knowledge is just a generic factive mental state that has as species states like seeing that p and hearing that p. Since this view is not obviously false, we shouldn’t include knowledge among the paradigmatically normative items. If knowledge were analyzable in terms of justification, it would be normative in our sense, by having as a constituent a paradigmatically normative property. But this claim is neither obvious nor an essential part of the traditional epistemology to which AR contributes. The tradition includes a strain of non-normativism about knowledge: Goldman () and Dretske () were surely key moments in traditional epistemology, even if they involved broke with a yet older orthodoxy. On Q. Let’s consider now how one might neutrally answer Q. By talking about a ‘central and fundamental role,’ defenders of AR have in mind (a) that responsibilist virtue properties do the bulk of the analytical work, with other properties playing a merely enabling role and (b) that these properties cannot be replaced without loss of explanatory power by other properties. 



Alston () recommended viewing justification as a merely evaluative property. But this insistence involved two mistakes. First, Alston assumed that voluntary control is a precondition for deontic assessment, which has been rightly questioned by doxastic compatibilists like Hieronymi () and McHugh (). Second, even if Alston were right that voluntary control is a precondition for deontic assessment, we shouldn’t deny that justification is a deontic property: we should rather stop talking about justification and replace it with talk about epistemic goodness. Teleologists would prefer the ideology of aims and their opponents would prefer the ideology of norms, broadly understood to include constitutive standards and not just deontological injunctions. This distinction between teleological and non-teleological theories crosscuts the distinction between VE and non-virtue-theoretic epistemologies. Some virtue epistemologists (e.g. Sosa and Greco) are teleologists. Responsibilists inspired by Aristotle may also count as teleologists. But responsibilists can reject teleology.

Responsibilism within Reason



To see what is meant by (a), we can contrast versions of AR with theories that partly invoke responsibilist virtue, like Greco’s (). According to Greco (: ), a belief is fully epistemically virtuous iff it is both a manifestation of an objectively reliable disposition and epistemically responsible, where epistemic responsibility consists in manifesting the dispositions constitutive of being motivated to believe the truth. This account is not purely reliabilist. But it is not a version of AR, since reliability is equally foregrounded. Reliability is also fundamental on Greco’s theory – it isn’t treated as significant merely because it follows from a more basic responsibilist requirement. This point matters, since it is often forgotten that Zagzebski () agreed that full epistemic virtue entails reliability, but maintained that reliability’s significance derives from a responsibilist virtue requirement. Hence, while Zagzebski and Greco agree that manifestations of reliabilist and responsibilist virtues are equally necessary for some central epistemic statuses, Zagzebski doesn’t treat the former as fundamentally explanatory. So only Zagzebski qualifies as a defender of AR. Greco () is, however, closer to AR than Sosa (). Sosa’s latest view is the mirror image of Zagzebski’s, seeking to derive responsibilist themes from a unified competence-theoretic framework. On Q. Whether Sosa’s reduction of responsibilist virtue to competencetheoretic virtue works partly depends on what counts as a responsibilist virtue property. Here some room for variation among responsibilisms has been strangely overlooked. According to the standard answer to Q, responsibilist virtue properties are character traits of the sort ascribed to persons when we deem them (e.g.) open-minded thinkers. I have elsewhere argued that responsibilists ought to ditch the emphasis on character traits in general and rich traits like open-mindedness and intellectual courage in particular, and I build on this point below. But there is a simpler objection to this answer to Q: it omits any mention of epistemic responsibility! Perhaps previous responsibilists implicitly assumed that the best way to analyze responsibility will be via the manifestation of character traits, and jumped to articulating AR characterologically to forestall questions about epistemic responsibility. But whatever the reason, I don’t think the characterological gloss belongs in a neutral characterization of responsibilism. It reflects a substantive and, I think, false theory about epistemic responsibility. 

See Sylvan .



 

A more natural and neutral characterization would appeal directly to epistemic responsibility, leaving various substantive theories open: responsibilist virtue properties will be virtue properties that can be instantiated only by subjects with the capacity for epistemic responsibility. There is then an independent question about what counts as a virtue property. Here again previous responsibilists jumped the gun in assuming that virtue properties are fundamentally instantiated by persons rather than acts and attitudes. In ethics, there is an opposing tradition represented by Thomson (), Hurka (), and Harman () according to which virtue terms apply most fundamentally to acts and attitudes, where applications to persons get analyzed in these more fundamental terms. While Hurka is not a virtue ethicist, Harman and Thomson uphold a novel version of virtue ethics. VE can make the same move, as I noted in Sylvan (). The move gains plausibility from two points. The first is Hurka’s () point that a person who isn’t virtuous can sometimes perform virtuous acts and merit high praise for them. The second point is that analyzing the act’s virtuousness as its being the sort of act a virtuous person would perform is unhelpfully indirect as well as extensionally inadequate. Even if the virtuous person would have performed this sort of act, this particular act might be performed for vicious reasons. Of course, one might add that the act be done for the sorts of motives the virtuous person would have. But the particular motives could have been acquired only in virtue of past viciousness. Instead of adopting these unhelpfully indirect analyses, I suggested in Sylvan () that it is at least as plausible to appeal directly to virtuous acts and attitudes, and characterize virtuous people as people who tend to do virtuous acts or form virtuous attitudes. I won’t repeat my arguments for this view. Here what matters is that virtue properties cannot be neutrally characterized as traits or global dispositions, since this characterization prejudges the substantive question of whether person-attaching virtue properties are fundamental. How can we give a more neutral characterization? One straightforward way involves linguistic ascent. There is a recognizable class of virtue terms, like ‘conscientious,’ ‘open-minded,’ and ‘courageous.’ These terms apply not just to persons, but also acts, attitudes, and motives independently of what we think about the person generally. With these points in mind, we can give the following neutral characterization of responsibilist virtue properties: 

Cf. Sylvan : .

Responsibilism within Reason



Responsibilist Virtue Properties (RVP): A responsibilist virtue property is a property ascribed by a virtue term (e.g. ‘conscientious’) whose felicitous use presupposes that the relevant subject has the capacity for epistemic responsibility.

Core AR is then the view that upholds ARC, with responsibilist virtue properties understood according to RVP, and the notions of epistemic normativity and a ‘central and fundamental role’ understood in the foregoing ways. ..

The Periphery

This neutral characterization makes it easy to see how existing versions of AR add peripheral assumptions about (i) epistemic responsibility, (ii) the scope and priority of virtue properties, and (iii) which specific responsibilist virtue properties should figure in the analysis of epistemic normativity. We should realize that there might be better versions of AR that drop these assumptions. While Zagzebski is not AR’s only defender, I’ll use her version to illustrate some accretions wrongly regarded as representative of AR. Similar points apply to other existing versions of AR, but I focus on Zagzebski because her version is best known, and its reception strongly affected AR’s generally. Zagzebski makes five key peripheral choices. First, she understands virtues as stable character traits. Second, she understands virtuous epistemic motivation as the desire for knowledge or, in other work, the love of truth. Third, she assumes that there is a tight connection between responsibility and character traits. Fourth, she holds that epistemic virtue is a kind of moral virtue. Fifth, Zagzebski’s responsibilist analyses of knowledge and justified belief are indirect in a way that threatens virtue’s explanatory force: for example, Zagzebski doesn’t require that justified beliefs actually manifest virtues, but just that justified believers believe the sorts of things and have the sorts of motivations that virtuous people would. Some of these choices can be observed in Zagzebski’s (: ) account of virtue: A virtue, then, can be defined as a deep and enduring acquired excellence of a person, involving a characteristic motivation to produce a certain desired end and reliable success in bringing about that end.



 

This account of virtue illustrates Zagzebski’s emphasis on character traits. It also illustrates her emphasis on a certain sort of motivation. Her characterization is already questionable by invoking a motivation to produce a desired end, which seemingly excludes non-consequentialist and nonHumean virtues such as respect from counting. As Scanlon (: –) and Parfit (: –) observe, respect for life needn’t require desiring or seeking to preserve life in every case (consider assisted suicide and voluntary euthanasia). Perhaps it is necessarily true that there is some end or other that one ought to desire to produce whenever one manifests respect for a value V. But this fact is not what makes respect for V a virtue: if one ought to desire that end, it is because one ought to respect V, not vice versa. Zagzebski’s characterization of the motivational requirement on epistemic virtue is yet more problematic. In Zagzebski (: –), the relevant motivation is the desire to acquire knowledge, though Zagzebski () later emphasized the desire to acquire truth. As we’ll see in Section ., these ways of understanding the motivational requirement on epistemic virtue lead to serious objections: when the motivational requirement is understood in these ways, it is unclear how knowledge and justified belief can be helpfully analyzed in terms of epistemic virtue. The fourth and fifth features of Zagzebski’s view emerge later in the book. Realizing that it is too demanding to expect every case of justification to involve a manifestation of virtue in her sense, Zagzebski (: , my emphasis) adopts a counterfactual analysis: A justified belief is what a person who is motivated by intellectual virtue, and who has the understanding of his cognitive situation a virtuous person would have, might believe in like circumstances.

As we will see in Section ., this indirectness creates an explanatory problem: it is hard to see how we can helpfully explain the actual justifiedness of your belief in terms of the would-be virtuousness of some counterfactual person who may have reasoned differently. Zagzebski’s (: ) account of knowledge may seem more direct: Knowledge is a state of belief arising out of acts of intellectual virtue. 

Her stress on character traits seems a consequence of a more fundamental view about the relationship between character and responsibility, which is dubious for reasons familiar from the literature against ‘true self’ accounts of moral responsibility (see, e.g., Wolf ). Zagzebski :  tells us that ‘the kind of responsibility we think of as distinctively moral and the praise and blame that accompanies it are associated with traits that are acquired gradually in the course of forming habits.’

Responsibilism within Reason



But in reading this claim, we must remember her account of acts of intellectual virtue: An act of intellectual virtue A is an act that arises from the motivational component of A, is something a person with virtue A would (probably) do in the circumstances, is successful in achieving the end of the A motivation, and is such that the agent acquires a true belief . . . through these features of the act. (: )

The resulting account of knowledge faces an explanatory problem like the one that faces the account of justification owing to the indirect, counterfactual nature of the second clause. A final controversial feature of Zagzebski’s view is its claim that intellectual virtue is a kind of moral virtue. When combined with her analyses of justification and knowledge, this claim suggests that a belief cannot be justified or knowledgeable if it is not a belief a morally virtuous person would have. It is not clear why we should agree. One might have thought there can be epistemically justified but false moral beliefs. It will be hard to agree that the morally virtuous person would have them. Merely requiring that they might have them doesn’t give us a sufficiently precise yardstick for sorting justified and unjustified beliefs. (Of course, one might deny that there can be justified false moral beliefs, or that fully morally virtuous people could form epistemically unjustified beliefs. But it would be nice not to make AR rest on such claims.)

. The Standard Objections to AR Are Objections to the Periphery We can now see why the standard objections to AR are only objections to peripheral features, such as the foregoing five in Zagzebski’s view. Let’s back up and recall the objections. Some of the most persuasive cases against AR have appeared in the last decade – especially in Alfano (), Baehr (), Dougherty (), Olin and Doris () and Sosa (). But the central dilemma was already clear in early responses to Zagzebski () – especially Alston (), Greco (), and Kornblith (). Since these papers were published in , I’ll call this dilemma the ‘YK Dilemma.’ Alston () pressed the first horn. He found it unclear how Zagzebski’s account accommodates ordinary cases of knowledge and justified 

See Zagzebski : –.



 

belief. Suppose you inadvertently come to know there’s a red book on the table by catching a glimpse of it and automatically forming the belief. One might doubt that you manifest any motivation to acquire knowledge or truth. Extending Alston’s point, one might find it implausible that any of the distinctive components of Zagzebski’s analysis are necessarily satisfied by perceptual knowledge and justified belief. There are ways out for Zagzebski, but they invite the dilemma’s second horn. Consider justified belief first. Zagzebski might insist that even if you aren’t virtuously motivated or exhibit virtuous character in believing, your belief is the sort a person with such character and motivation would have. And all Zagzebski requires for justification is that this is true. But while this point shows that the case is not a direct counterexample, it raises a question about the explanatory significance of Zagzebskian virtue, as Greco () and Kornblith () stressed. It is unclear how we can explain your belief’s justification by appealing to some counterfactual person’s virtues. Zagzebski’s account of knowledge doesn’t so clearly invite this worry, since it imposes an actual motivational requirement. But even if this requirement is met by ordinary virtuous subjects, one might find it explanatorily superfluous. There are more and less compelling versions of this objection. I don’t agree with Greco () that some purely reliabilist feature does the explanatory work instead. When it comes to the kind of knowledge that entails justification, there is an important difference between clairvoyance and blindsight cases and ordinary perceptual cases, and I think no pure reliabilist story explains this difference. Moreover, Greco invokes reliability of character to address clairvoyance cases, opening his explanation up to situationist objections and simpler worries in the cases that motivated Harman, Hurka, and Thomson to deny that virtuous action requires virtuous character. Still, a version of this objection is decisive against Zagzebski for three reasons. First, Zagzebski’s account of the motivational requirement is too strong. Greco () rightly insists that even if it is part of the mechanics of adult knowledge acquisition, it isn’t part of the constitutive story. Second, the second clause of Zagzebski’s account of knowledge still employs a counterfactual, and it is explanatorily superfluous. Third, as Greco () noted, no particular forms of Zagzebskian motivation (e.g. openmindedness) play a constitutive role even when present. There is hence a 

But it does invoke to a similar counterfactual fact via the second clause of Zagzebski’s analysis of an act of intellectual virtue, which invites a similar objection.

Responsibilism within Reason



dilemma for Zagzebski. On the one hand, Zagzebski invites extensional objections: there can be knowledge without acts of intellectual virtue. To avoid these objections, Zagzebski could emphasize the indirectness of her appeal to virtue. But then virtue becomes explanatorily superfluous. I take this dilemma to be the most important objection to Zagzebski, and see most other objections as variations on it. For example, Baehr () and Dougherty () invoke cases like Alston’s to suggest that Zagzebski fails to explain ‘low-grade’ knowledge, and Baehr (: –) adds that Zagzebski’s mention of virtue loses its explanatory significance if sidelined in some counterfactual clause. The situationist objection is also a variation: since we don’t have the relevant character traits, accounts that require them will be extensionally inadequate, and accounts that invoke them counterfactually make them explanatorily superfluous. Some of Sosa’s () objections are also variations. Sosa (: –) memorably remarks that Zagzebski’s motivational requirements aren’t met by ‘[h]edge fund managers, waste disposal engineers, and their receptionists, [who] can all attain much knowledge in the course of an ordinary workday despite the fact that they seek the truths relevant to their work only for their instrumental value,’ adding that ‘[t]hat is why they want them, not because they love truth.’ But it is worth separately noting two further objections in the surrounding text: Independently of all that, it remains that there is a distinctive dimension of epistemic assessment isolated from all such broadly ethical (or prudential) concerns. Moreover, within this epistemic dimension, love of truth plays a negligible role at most, if any at all . . . An assassin may even have no desire whatever for the truth on the location of his victim except only for the fact that it will make his crime possible . . . His search for truth, since agential, is subject to the full range of responsibilist assessment nonetheless. And his knowing the location of the victim in believing as he does . . . is still epistemically better than his merely believing correctly. (: –)

One implicit objection here is to Zagzebski’s blurring of moral and epistemic virtue: it is unclear why beliefs with the highest epistemic qualifications must have moral merit or be believed by someone with moral merit. A second implicit objection is to Zagzebski’s specific motivational requirement. Sosa allows that his assassin might exhibit responsibilist intellectual virtues. What he denies is that love of truth is an operative virtue, or a desire for truth in general. 

Baehr (: –) also makes this second point.



 

These objections are, I believe, compelling against Zagzebski. Versions of the YK dilemma’s first horn and the situationist objection undermine the appeal to character traits. Sosa’s points are compelling against Zagzebski’s motivational requirement and her claim that epistemic virtues are moral virtues. The YK dilemma’s second horn is decisive against Zagzebski’s indirect appeal to virtue. But as I’ve emphasized, these features of Zagzebski’s view are peripheral. It remains to be seen whether similar objections confront versions of AR that drop them. I think not, and I will now turn to develop a version of AR that avoids these objections and has independent attractions.

. Responsibilism within Reason My version of AR is a two-tiered view. The first tier offers a metaphysical analysis of epistemically normative properties in terms of responsibilist virtues I call virtues of respect for truth. Because my view is a version of AR, it denies that these virtues admit of analysis in non-responsibilist normative terms. This view is, however, compatible with the naturalist thought that these virtues are in turn grounded in non-normative properties. The first tier’s analysis is only intended as a view about the internal structure of the epistemically normative domain, leaving open how it is grounded in the non-normative. The question of what it takes to manifest the virtues of respect for truth can be answered in several ways. Besides answering it with an analysis of these virtues in non-virtue-theoretic terms, one could answer it by offering a piece of first-order normative theorizing about virtue-making characteristics (by analogy with first-order theories of right action that seek to determine what makes acts right). This observation brings us to the second tier, which is a theory of this kind. Notably, this theory appeals to notions reminiscent of reliabilist VE – in particular, to certain reliable (local) dispositions, and to a distinction among ways of manifesting these dispositions that mirrors Sosa’s adroitness/aptness distinction. But in contrast to Sosa, these are proximally dispositions to reliably conform to reasons, and only distally to hit the mark of truth. I call the overall view ‘Kantian responsibilism’ (KR). I use this name primarily because it invokes a notion reminiscent of Kantian ethics – viz. respect – but here I understand this as a responsibilist virtue. 

I take this view to be consistent with the more systematic Kantian view I defend in Sylvan , because I take the views to operate at different levels. The epistemic Kantianism in Sylvan  is a foundational first-order theory. Kantian responsibilism is a complementary theory working at two

Responsibilism within Reason ..



The First Part of KR: An Analysis of Epistemic Normativity

According to KR, the most fundamental properties in the epistemically normative domain are virtues of respect for truth. Two are especially crucial. On the one hand, manifesting full respect for truth requires forming doxastic attitudes that conform to epistemic reasons (of some sort) as a manifestation of a disposition to so conform, where epistemic reasons for (dis)belief are understood as indicators of truth and falsity. I’ll call this sort of respect heedfulness of the standard of truth (i.e. the standard that it is correct to believe p only if p is true). I don’t assume that heedful belief is factive: even if you form a false belief, you might be sufficiently heedful of the standard of truth if your evidence strongly supports it. I also don’t assume that heedful belief is transparent: you can be heedful of reasons without realizing that this is what you are doing, and indeed while believing that you’re failing to do so. Partly for that reason, manifesting full respect for truth also requires forming attitudes that cohere with one’s beliefs about what the reasons support. I will call this kind of respect conscientiousness relative to the standard of truth. If you follow your convictions about what is required for heeding the standard of truth, you are at least conscientious with respect to that standard. While full respect for truth requires heedfulness and conscientiousness, there may be cases in which it is impossible to manifest both. Whether there are depends on whether we can have justified false beliefs about what it takes to respect truth. If so, being heedful might require being unconscientious in some cases. These would be epistemic analogues of Arpaly’s () Huck Finn case. Just as the intuition is that Huck should heed the real moral reasons and not his false beliefs, so the intuition here assuming the case is possible is that one should be heedful at the cost of being unconscientious. (As it happens, I doubt whether it is possible to have justified false beliefs about what it takes to respect truth, but it is not crucial to resolve this issue here.) So far, I’ve been imprecise about the ontology of respect for truth. The reason is that evaluations of respect can have different objects, and there is a further question about which is most fundamental. A person who doesn’t have the character trait of being heedful or conscientious relative to the other levels: the level of analysis, and the level of factoral first-order theory. The ‘fundamental role’ played by virtue in AR is an analytically fundamental role, not a normatively fundamental role. Hence AR is consistent with there being a deeper normative story, such as epistemic Kantianism.



 

standard of truth might manifest heedfulness on some occasions or with respect to some subject matters. When we say that this person manifests some respect for truth, we might mean they manifest a local disposition to form attitudes that are respectful of a certain sort of evidence. Or we might simply mean that they on that occasion formed their beliefs in a respectful way. My preferred way of regimenting talk of respect is to treat respectfulness of attitude-formation as fundamental, and to see doxastic attitudes as respectful of truth to the extent that they manifest dispositions to form attitudes that are respectful of the evidence and one’s beliefs about what is probably true relative to it. We respect truth by respecting evidence and our beliefs about what it supports. We in turn respect the evidence not just by forming the attitudes that happen to conform to the evidence, but rather as a manifestation of a (typically local) evidence-sensitive disposition. We can use respect talk to pick out character traits and global dispositions, but these play no fundamental role in my view. We might also talk about an attitude of respect. But when I talk about manifesting respect, I don’t have in mind that one has a corresponding occurrent attitude, beyond what is minimally involved in manifesting a (typically local) reasons-sensitive disposition. When I turn to the substantive part of KR, I will say more about what it takes to manifest full respect, and in turn more about what is substantively required for various epistemic statuses. But for now, I’ll leave open some substantive details, which are part of the periphery (though one I now endorse). ... KR’s Account of Virtue We can now state KR’s analyses of various epistemic properties, beginning with the most obvious case: epistemic virtue. KR upholds the following general claim: KR-V: All epistemic virtue properties consist in (manifestations of ) forms of respect for truth. We can then distinguish attitude-attaching and person-attaching virtue properties, as well as between some varieties of the latter: KR-Vattitudes: A doxastic attitude’s being epistemically virtuous =df-metaphysical its being respectful of truth. KR-Vperson-weak: A person’s being epistemically virtuous in the weak sense =df-metaphysical their tending to have attitudes that respect truth.

Responsibilism within Reason



KR-Vperson-strong: A person’s being epistemically virtuous in the strong sense =df-metaphysical their tending to have attitudes that respect truth as a manifestation of character. These claims are compatible with a wide range of substantive views, since it is an open question what it takes to respect truth. Still, they are highly non-trivial. Indeed, these claims severely restrict the scope of epistemic virtue in a way that is unusual for responsibilism. Responsibilists often sing praises of intellectual courage, open-mindedness, and the like. While one might try to argue that intellectually courageous and open-minded beliefs are necessarily respectful of truth, these claims are hardly obvious. As far as I’m concerned, this result is a good one: while there is something morally virtuous about open-mindedness and virtuous simpliciter about intellectual courage, it is unclear that these properties necessarily contribute to properly epistemic virtue. While some responsibilists won’t share my taste for an austere conception of the properly epistemic, it is a taste that defenders of AR should, I believe, acquire. For it is false that intellectual courage, open-mindedness, and the like are required for justification, rationality, or knowledge. ... KR’s Accounts of Justification and Rationality This remark brings us to KR’s account of the epistemic statuses central to traditional epistemology. We can start with justification. Like many epistemologists, I think there are several properties often lumped under the heading of ‘justification.’ I reserve the term ‘rationality’ for a more subjective status, and ‘justification’ for a more objective but still perspective-dependent status. This distinction aligns with my distinction between the two main forms of respect for truth, recommending: KR-R: A doxastic attitude’s being epistemically rational =df-metaphysical its being conscientious relative to the standard of truth. KR-J: A doxastic attitude’s being epistemically justified =df-metaphysical its being heedful of the standard of truth. Of course, we want beliefs to be both heedful and conscientious. Because I seek to preserve an analogy with ethics, I don’t want to capture this point by saying that justification or rationality requires both: an act can be right, and justified, even if the agent believes it is wrong (cf. Huck Finn), and an act can be rational without being justified because it is wrong but the agent believes otherwise, in accordance with how things appear to them. I would



 

hence prefer to stipulate a term to refer to the union of the two properties, which I’d call full epistemic worth. Now, these analyses lack substantive implications independently of a first-order theory of heedfulness and conscientiousness. There are different options I don’t want to prejudge. My earlier glosses might seem to suggest that KR is neo-evidentialist: I suggested that one is heedful of the standard of truth by being heedful of the evidence. But as we’ll see when turning to my full first-order story, KR is compatible with reliabilist themes and avoids stock problems for evidentialism. ... KR’s Account of High-Grade Knowledge A disclaimer is in order before I describe KR’s take on knowledge. Earlier I emphasized that KR is only an account of epistemic normativity, and that not all objects of epistemological investigation are clearly normative. Indeed, I consider the study of what Dretske () called ‘epistemic seeing’ (seeing that p) as well as the study of remembering that p, intuiting that p, and other factive mental states to be core epistemology. I don’t consider any of these states to be normative. If knowledge is just a generic factive mental state, I don’t consider it normative either, and favor divorcing the theory of knowledge from the theory of epistemic normativity. KR doesn’t, however, require this claim. Defenders of KR have several options. First, they could follow Sosa in distinguishing animal knowledge and a fancier sort of knowledge, offering an account only of the latter. They could then insist that the former isn’t normative. Second, they could instead deny that the determinate factive mental states are forms of knowledge, and say that purported cases in which children or dogs have knowledge without respect for truth aren’t cases of knowledge but just cases involving determinate factive mental states. Finally, they could insist that even rudimentary knowers can manifest respect for truth, by adopting a first-order theory that only requires respectful subjects to respond to reasons as a manifestation of reasons-sensitive dispositions.

 

For my defense of this picture, see Sylvan . For antecedents, see Foley  and Kornblith . I use the term ‘high-grade’ rather than ‘reflective’ for greater generality, since not all fans of bi-level epistemology must hold that high-grade knowledge requires reflective ascent. Indeed, in and after his  and  works, Sosa distanced himself from the old label, reserving ‘reflective knowledge’ for mere meta-apt belief and introducing the new ideas of full aptness and judgmental knowledge to capture the fancier achievement.

Responsibilism within Reason



Each option is respectable. For this chapter’s purposes, I adopt the first but understand high-grade knowledge broadly, so that most human knowledge and some non-human animal knowledge counts. Having issued these disclaimers, I can now state KR’s account of high-grade knowledge: KR-K: S’s belief that p is high-grade knowledge =df-metaphysical S’s belief that p is accurate, heedful of the standard of truth, and its accuracy manifests this heedfulness. This account partly resembles some reliabilist VEs in invoking manifestation (e.g. Sosa [] and Turri []). Gettier cases are cases where the belief is accurate and heedful but not accurate as a manifestation of heedfulness. As a result, the view has similar Gettierological payoffs. There are, however, big differences with reliabilist VE’s story, especially if Sosa is the comparison. Recall how in his classic () discussion, Sosa used the animal/reflective knowledge distinction to address fake barn country, claiming that we have animal knowledge but lack reflective knowledge there; while Sosa since his  and  works has distanced himself from using the ‘reflective’ label for the highest epistemic achievement, the approach to fake barn country remains similar. My account of high-grade knowledge is structurally more like his account of animal knowledge, since heedfulness needn’t be second order. If so, I cannot make Sosa’s move. My response is that there are other ways to diagnose the intuition. It is open to me to offer Littlejohn ()’s type of diagnosis: the accuracy of the belief doesn’t manifest its heedfulness. It is easy to offer this diagnosis if manifestations of heedfulness coincide with manifestations of a sort of ability (though I don’t think Littlejohn’s broader point requires his distinction between senses of ‘able’). Another difference noted in passing is that KR’s account of high-grade knowledge doesn’t emphasize the second order. It would if conscientiousness were required for knowledge, or if heedfulness required conscientiousness. But I think neither is true. The core contrast for KR is evinced by the contrast between the knowledge of blindsighters and the knowledge of normal perceivers: it is a first-order difference concerning whether the subject heeded a sufficient reason. Sosa’s older 

I am using ‘accuracy’ here in Sosa’s sense, to refer to the property that a belief has when its content is true. I am not using it in the new sense introduced by epistemic utility theorists like Pettigrew .



 

‘animal’/‘reflective’ lingo is hence misleading: the core distinction is between knowledge understood as registration of facts and knowledge understood as Reason-produced accuracy. .. The Second Part of KR: A Substantive Account of What Respect Requires The first tier of KR doesn’t give enough information to derive clear predictions about cases. I’ll now provide this information. But it is worth stressing that this feature of the first tier is no bug, given its point. The point was to limn the internal structure of a normative domain. This kind of project is standard in meta-ethics. Compare the fitting-attitudes account of value, according to which being valuable consists in being a fitting object of valuing. It is no objection to this view that it makes no clear predictions about what is valuable, which is not its point. It is also no objection that it appeals to normative notions. Still, besides wanting a real definition of various epistemically normative properties in fundamental normative terms, one wants a story about what it takes for the definiens to be instantiated. My story reveals the deliberate ambiguity of my title. This story is ‘within reason’ not only in the sense of being reasonable to accept, but also in linking responsibilist virtues to the abilities that constitute the faculty of theoretical reason. Some preliminary clarifications are in order about the nature of the story. This account will appeal to the notion of a reason, which in metaethics is standardly taken to be the paradigm of normativity. But this feature doesn’t contravene the normative fundamentality of responsibilist virtue, for two reasons. First, because the account is first order rather than meta-normative, it doesn’t follow that the account ultimately takes reasons rather than virtues to be analytically prior. Second, reasons can be analyzed in terms of responsibilist virtues: roughly, reasons are considerations that figure in the virtuous operation of theoretical reason. This account combines the Kantian aspiration of explaining reasons in terms of Reason familiar from Korsgaard (, , ), with virtue-ethical doubts about whether reason-relations can be captured by simple principles familiar from McDowell () and Dancy (). While I am not as skeptical in the epistemic case as Dancy and McDowell are in the ethical case that simple principles can be found, I don’t see a helpful way of stating them without implicitly adverting to responsibilist virtue. A second clarification is in order. One might worry that my appeal to reason conflicts with the spirit of VE, which was meant to supplant

Responsibilism within Reason



‘intellectualist’ models of epistemic normativity. The response to this concern is that it rests on implicit overintellectualization of the workings of reason. As Dancy (: –) notes in the non-epistemic case, reasons-responsiveness needn’t involve inference (when inference is understood as involving rule-following): [m]oral thinking . . . can perfectly well be thought of as reasoning just in the sense that it is deliberation, i.e., the discerning of interrelations of reasons, [which] involves two basic stages. First, one recognizes what reasons are thrown up in the situation one is in; this is not inference . . . Secondly, one works out what those reasons require of one. This sort of working out need not be thought of as inference either.

Responding to epistemic reasons can work in this way. After pondering the evidence, one might treat it as favoring the best-supported hypothesis via a reasons-sensitive competence. This process needn’t involve inference, at least when understood in the natural sense that involves rule-following. Still, it is reasoning – a working out of what the evidence recommends via Reason. ... Responsibilist Virtues, Rational Capacities, and Reasons The idea of linking responsibility and Reason is standard in the philosophy of action. Over there, there has been a shift away from the ‘true self’ model of responsibility – a model that perhaps implicitly guided earlier responsibilists to fixate on character – and toward the reasons-responsive models of Wolf (), Fischer and Ravizza (), and Nelkin (). The core of this conception of responsibility is that responsible creatures are creatures with the capacity to respond to reasons. The exercise of responsibility consists in the ‘rational control’ of actions and attitudes by rational capacities, which are dispositions to respond to reasons. This model has been extended to the epistemic case by Hieronymi () and McHugh (). Of course, such models are merely intended to capture responsibility in the weak sense associated with accountability, which one can manifest while being blameworthy. But the capacities invoked by these theorists can operate more or less well. The well functioning of these capacities will consist in responsiveness to good reasons. Given this point, we can exploit certain structural distinctions familiar from performance epistemology to

 

Consider Code : , who takes inspiration from Sosa  in rejecting what both call the ‘Intellectualist Model of Justification.’ For further discussion of inference and its relation to reasons-responsiveness, see Sylvan b.



 

illuminate the distinctions among exercises of rational capacities that suffice for various epistemic statuses. My substantive account of virtue invokes reasons-sensitive dispositions. These dispositions proximally function to respond to reasons understood as they are standardly understood in ethics – viz. as facts that count in favor of acts and attitudes, where these facts are not necessarily ones to which we have access. Of course, for the view to remain true to the spirit of responsibilism, the favoring relation and the relation of correctly responding to reasons (‘proper basing’) might need analyzing in responsibilist terms. But even without such an analysis, appealing to reasons at this tier doesn’t involve any commitment to normative fundamentality: this tier isn’t an account of the constitution of epistemic normativity but rather a first-order account. Now, one might separately worry about the appeal to reasons understood as facts, insisting that only reasons we possess can epistemically justify. I agree with the latter claim, but don’t think it threatens the helpfulness of such reasons for understanding epistemic normativity. Possession needn’t be understood normatively. If we understand possessing a reason as accessing the fact that provides it, we can then understand access in terms of factive mental states like seeing that p, remembering that p, intuiting that p, and so on. One might also insist that when it comes to rationality, the reasons to which one responds in exemplifying this virtue need only be apparent facts. But again, we needn’t understand the relevant notion of appearance normatively: an apparent fact is just one that non-doxastically seems to obtain. A final preliminary note: I assume that the paradigm examples of epistemic reasons are pieces of evidence, understood as truth-indicating facts or apparent facts. I don’t assume that only these can provide epistemic reasons. Epistemic reasons for agnosticism will include considerations about the evidence (e.g. whether it is sufficient). Perhaps these are just higher-order evidence. But I’ll remain neutral, assuming only evidence provides the paradigm example of an epistemic reason for (dis)belief, and 



One could also take this relation to be non-normative: perhaps the evidence-for relation isn’t normatively constituted, and the capacities constitutive of theoretical reason are capacities to respond to evidence. Note that if accessing a fact is understood as a determinable factive mental episode that takes as determinates factive mental episodes like seeing that p and remembering that p, there is no need to impose any normative qualification on such access (e.g. to qualify it as proper access). I take this way of understanding access to be more intuitive than an understanding that would count mere believing as a form of access. It is counterintuitive to claim that mere belief that p gives one access to the fact that p (though pre-Williamsonian epistemologists often talked otherwise).

Responsibilism within Reason



that the yardstick of sufficiency tracks whether heeding the reasons would be respectful of truth. ... Conditions for Respectful Belief-Formation We now have the necessary ingredients for a substantive account of our responsibilist virtues and the epistemic statuses they ground. We can begin with the thought that the responsibilist virtues coincide substantively with certain rational capacities. Epistemic rational capacities in turn are dispositions to form doxastic attitudes in certain conditions, and to take certain kinds of considerations into account in doxastic deliberation. Here there are two sets of crucial distinctions: distinctions concerning the extent to which (i) these dispositions qualify as rational capacities and (ii) these dispositions qualify as manifestations of rational capacities. Corresponding to these distinctions will be substantive conditions for various forms of respect for truth, and for respectful attitudeformation. A disposition to form doxastic attitudes of type D in conditions of type C qualifies as a minimal rational capacity if conditions of type C constitute sufficient apparent epistemic reasons for attitudes of type D. A disposition to form doxastic attitudes of type D in conditions of type C qualifies as a full rational capacity if conditions of type C constitute sufficient possessed epistemic reasons for doxastic attitudes of type D, where these reasons aren’t merely apparent. When the ‘apparent’ is understood as ‘apparent relative to one’s doxastic attitudes,’ having a disposition of the first sort coincides as a first-order matter with local conscientiousness relative to the standard of truth. When the ‘apparent’ is understood relative to the non-doxastic appearances, having a disposition of the first sort coincides as a first-order matter with weak local respectfulness of the standard of truth. Having a disposition of the second sort coincides as a substantive normative matter strong local respectfulness of the standard of truth. These dispositions can in turn be manifested in different ways. Suppose one seeks to predict the behavior of some object, relying on an inductive disposition to form predictions of type T given evidence of 

This substantive account remains high level, since there are further questions about when (apparent) facts provide (apparent) evidence. But substantive normative theorizing has levels of its own: as Kagan  stresses, normative ethics operates at two levels of generality, with factoral normative theorizing seeking just to articulate all factors relevant to the exemplification of normative properties, and foundational normative theorizing seeking an account of which factors are most fundamental. See Sylvan  for the foundational story.



 

type E. This disposition is a minimal rational capacity. A belief might manifest this capacity despite the evidence being misleading or merely apparent. But it might manifest this capacity when the evidence is not misleading. It might be • • • •

correct as a manifestation of a full rational capacity, manifest a full rational capacity but be correct just by luck, incorrect but manifest a full rational capacity, incorrect but manifest a minimal rational capacity.

Each case corresponds to a substantive condition for some form of respect: • • • •

Fully respectful belief: A belief is fully respectful iff correct as a manifestation of its responsiveness to possessed epistemic reasons (where this is mediated by a full rational capacity). Strongly heedful belief: A belief is strongly heedful iff it conforms to possessed epistemic reasons as a manifestation of a full rational capacity. Weakly heedful belief: A belief is weakly heedful iff it conforms to apparent epistemic reasons in the non-doxastic sense of ‘apparent’ as a manifestation of a minimal rational capacity. Conscientious belief: A belief is conscientious iff it conforms to apparent epistemic reasons in the doxastic sense of ‘apparent’ as a manifestation of a minimal rational capacity.

There are intermediate cases between the first and second statuses. These are Gettier cases: the belief will be correct and reasons-responsive, but not correct as a manifestation of reasons-responsiveness. ... Conditions for Justification, Rationality, and High-Grade Knowledge We can now turn to substantive conditions for the various epistemic statuses and see the case for a further distinction among sorts of rationality. From KR-R and the substantive conditions for conscientiousness, we can derive the following conditions for rationality: KR-RS: A doxastic attitude is rational iff it conforms to apparent epistemic reasons in a doxastic sense of ‘apparent’ as a manifestation of a minimal rational capacity. This thesis illuminates rationality in the structural sense standard in ethics (e.g. in Scanlon [], Kolodny [], and Broome []), which Worsnip () encourages epistemologists to take into account, and which has an antecedent in Foley’s () notion of rationality. This

Responsibilism within Reason



notion of rationality doesn’t correspond neatly to the sort that interests internalists in epistemology. But note that we can divide KR-R into two claims, with the distinction between conscientious and weakly heedful belief providing the foundation: KR-R-: A doxastic attitude’s being structurally rational =df-metaphysical its being conscientious relative to the standard of truth. KR-R+: A doxastic attitude’s being substantively rational =df-metaphysical its being weakly heedful relative to the standard of truth. We can then also divide KR-RS into two claims: KR-RS-: A doxastic attitude is structurally rational iff it conforms to apparent epistemic reasons in a doxastic sense of ‘apparent’ as a manifestation of a minimal rational capacity. KR-RS+: A doxastic attitude is substantively rational iff it conforms to apparent epistemic reasons in the non-doxastic sense of ‘apparent’ as a manifestation of a minimal rational capacity. Both sorts of rationality then turn out distinct from justification if the latter is understood in terms of strong heedfulness, via the following precisification of KR-J: KR-J+: A doxastic attitude’s being justified =df-metaphysical its being strongly heedful with respect to the standard of truth. Given the substantive conditions for strong heedfulness, we can then derive: KR-JS+: A doxastic attitude is justified iff it conforms to possessed epistemic reasons as a manifestation of a full rational capacity. Finally, we get substantive conditions for high-grade knowledge by understanding high-grade knowledge as fully respectful belief and invoking the substantive conditions for the latter: KR-KS: A belief is high-grade knowledge iff it is correct as a manifestation of its responsiveness to possessed epistemic reasons (where this manifestation is mediated by a full rational capacity). The result is a complete first-order normative epistemology for the most central epistemic statuses. Extensionally, this view will fall somewhere between evidentialism and Sosa’s () performance epistemology for judgment. It is unlike evidentialism in three respects. First, it fundamentally provides substantive conditions for doxastic justification. Second, unlike Conee and Feldman’s



 

() evidentialism, the notion of a reason to which this view appeals is non-mentalist. Of course, only possessed reasons do much epistemic work. But possession is a relation that holds in virtue of factive mental states, not the non-factive ones of Conee and Feldman’s mentalism. Although reasons that appear to exist relative to non-factive mental states play some role, cognition doesn’t aspire merely to respond to these apparent reasons. A third difference is that KR invokes dispositions to conform to reasons, inducing a threefold pattern reminiscent of Sosa’s AAA pattern. A belief can • • •

merely conform to some reasons (by being favored by them); manifest a reasons-sensitive disposition; conform as a manifestation of a reasons-sensitive disposition.

The view is, however, not extensionally equivalent to Sosa’s, since Sosa doesn’t require reasons-responsiveness even at the judgmental level. I’ve argued elsewhere that this feature of Sosa’s view is problematic, so I take the difference to represent an advantage for KR. .. A Consistent Reduction Plan To remain true to responsibilism, we need an analysis of epistemic reasons either in responsibilist terms or in terms that render reasons nonnormative. My preference is for a version of the first option that makes KR Kantian in a further way, but I think there are respectable versions of the second. Before discussing either option, I’ll stress again that it is compatible with these metaphysical stories that we have independent conceptual grasp on epistemic reasons and the conditions under which they make beliefformation respectful. If we have this grasp, we don’t need the reductions below to derive clear predictions from my two-tiered account. And we do, I believe, have sufficient pre-theoretical grasp of the concept of a reason for the account to be predictive. The point of this section, then, is merely to show that appealing to reasons in a first-order theory doesn’t contravene the metaphysical claim that responsibilist virtues come first. ... A Consistently Kantian Reduction Plan Following Way (), Gregory (), and Silverstein (), defenders of the first reduction plan will first invoke a high-level analysis of reasons as 

See Sylvan forthcoming.

Responsibilism within Reason



inputs to good reasoning, where ‘reasoning’ is understood broadly to include non-inferential transitions like the transition from seeming to see that p to believing that p. Good theoretical reasoning is then analyzed as reasoning that is respectful of truth, where there is no commitment to such reasoning being economically codifiable without invoking responsibilist normativity (otherwise the view would be guaranteed to be purely Kantian at the fundamental level, not virtue-theoretic). We needn’t deny that good reasoning follows patterns, however: it is just that nothing holds them together apart from their contributions to respect for truth. There is a further Kantian step we can take without sacrificing this virtue-theoretic theme. We can give a constitutivist reduction of theoretical normativity, viewing good patterns of reasoning as patterns constitutive in the sense that thought must often enough conform to them to count as reasoning at all. This step also shows how we might ground epistemic normativity in the non-normative: the study of epistemic normativity is the study of the statuses associated with patterns of thought constitutive of theoretical reasoning, full sensitivity to which constitutes respect for truth. ... Non-normativism about (Epistemic) Reasons? A different option is to deny that invoking reasons is invoking some nonresponsibilist normative property, by denying that reasons are essentially normative. This option flouts the recent meta-ethical chant that has it that reasons are the paradigmatic normative items. To those entranced by this chant, this option will seem bewildering. But it has been pursued by Broome () and Finlay (), who take the fundamental notion of a reason to be explanatory, and take normative reasons to be a special case of reasons so understood. On this view, normative reasons are normative only in virtue of being explanations of normative facts. Now, Broome and Finlay appeal to non-virtue-theoretic normative notions: for Broome, reasons to ϕ are explanations of why one ought to ϕ, while for Finlay, reasons are explanations of things are good in some way. But one could imagine a virtue-theoretic variant on which reasons provide explanations of why certain attitudes would be virtuous in a responsibilist sense. A different option would treat epistemic reasons specifically as nonnormative. The most obvious version of this view would combine the view that all epistemic reasons are evidence with the view that the evidence-for relation is non-normative, consisting perhaps in some truthindication or probabilification relation, with probability understood nonnormatively.



 

. Answers to the Standard Objections Whichever form it takes, the view I’ve outlined dispatches with the standard objections to AR. The first horn of the YK dilemma is easily addressed. To be sure, KR imposes requirements that one might call ‘motivational.’ But as a substantive matter, these requirements will be met by subjects who attain the relevant epistemic statuses. For given our first-order story, it will be fine to ascribe such respect to justified believers and high-grade knowers. While it is not uncontroversial that rationality, justification, and high-grade knowledge require responding to reasons, these claims are defensible and cannot be rejected out of hand. Indeed, these claims have been defended by some reliabilist virtue epistemologists when reasons are understood as I’ve understood them. The further requirement that this responsiveness manifest a form of reasons-sensitivity enables KR to secure reliabilist VE’s good predictions about cases of accidental conformity to reasons. In Greco’s (: ) example of the poor math student who accidentally lands on a correct algorithm for solving a problem, such sensitivity is not manifest. It would be a mistake to insist that reliability of character is what’s really needed. Poor math students can manifest understanding of some algorithms, and when they solve problems as a manifestation of that understanding, they attain justified belief and perhaps high-grade knowledge. Character-based views undergenerate epistemic status in such cases. The second horn of the YK dilemma is straightforwardly addressed if it takes Kornblith’s form. Kornblith was worried about the explanatory helpfulness of Zagzebski’s merely counterfactual appeal to virtue properties. Our view doesn’t make a counterfactual appeal: it requires that the justified believer actually manifest respect. So, it cannot face Kornblith’s worry. Greco’s version of the worry is more interesting. He insists that it is unclear what explanatory role Zagzebski-type virtues as such play. Here he invokes a sub-dilemma. He observes that the success component of Zagzebski-style virtues can be understood either to yield truth-reliability 



Matt Stitcher emphasized in p.c. that it is unclear that respect for truth is a kind of motivation rather than a kind of deliberative constraint. I agree that respect isn’t any sort of desire. As an anti-Humean about motivation, I don’t think it follows that respect isn’t a kind of motivation. Perhaps more revealing is the point that the person who manifests respect for truth needn’t have any thoughts about the value of truth among their reasons. See Sylvan and Sosa , where Sosa agrees that rationality and justification entail reasonsresponsiveness.

Responsibilism within Reason



or to not do so. If the latter holds, Greco thinks the account makes bad predictions: math students who follow unreliable algorithms accidentally but are generally responsible in Zagzebski’s sense won’t thereby have justified beliefs. If the former holds, Greco (: ) thinks ‘it is the agent-reliability that is doing the work, and nothing about Zagzebski-type virtues as such.’ But this last claim doesn’t extend to KR for two reasons. The first is a general reason that also helps Zagzebski. The fact that reliability is necessary for manifesting responsibilist virtues obviously doesn’t imply that it is sufficient. Clairvoyance cases suggest that a responsibilist virtue property is also necessary, as Greco (: ) agrees. Perhaps it isn’t necessary for animal knowledge, and perhaps clairvoyants have such knowledge. But one can use this point to argue that animal knowledge is not normative, and hence not among KR’s analysanda. Now, one might instead complain that it is ad hoc for Zagzebski to stipulate that responsibilist virtues are truth-reliable, and that our intuitive notions of, e.g., open-mindedness and intellectual courage don’t entail truth-reliability. One might hence worry that there isn’t a principled responsibilist rationale for this requirement. But KR avoids this objection. Sensitivity to reasons understood as facts is constitutive of what I earlier called strong heedfulness, which does entail at least indexical truthreliability (and one might take the new evil demon problem to suggest this is all we can demand). There is a third way to make Greco’s point, but it requires a normative analogue of the exclusion principle that threatens any appeal to virtue. One might argue that the properties in my substantive account of respect do the explanatory work, and that the supervening respect isn’t doing any work. But this objection overgeneralizes. The reasoning is only compelling given the more general principle that the properties subvening normative statuses do the real work. This principle suggests that the non-normative properties that subvene any virtue properties do the work, not the virtue properties as such. Such, at any rate, follows barring a naturalist identity theory. Yet such a theory is open to the responsibilist too, particularly if they adopt the Kantian strategy of Section .... To avoid overgeneralization, the objector must provide a more minimal supervenience base that can do the explanatory work without subvening respect for truth. But we need sensitivity to reasons to get the right predictions about clairvoyance cases, and it subvenes respect for truth. 

See Sylvan  for such an argument.



 

Perhaps there is some other story, but one hasn’t been given; again, Greco himself agrees that clairvoyance cases necessitate a responsibilist requirement for justification. So much for the YK dilemma. KR also addresses Sosa’s objections. Now, Sosa (: ) does mention respect for truth in passing and he cites my dissertation as an example of a view which invokes this idea. He dismisses the idea, claiming that his hedge fund managers and waste disposal engineers needn’t have such respect to attain justification or knowledge. But his dismissiveness rests on a failure to distinguish the relevant notion of respect from the irrelevant notion Darwall () calls ‘appraisal respect.’ Appraisal respect amounts to esteem, and requires having a ‘high opinion’ of the object of respect. By contrast, recognition respect in Darwall’s () sense just requires giving the object of respect ‘appropriate weight in deliberation.’ Yet if hedge fund managers and waste disposal engineers exhibit sensitivity to evidence, they do then exhibit recognition respect for truth, by giving appropriate weight in their doxastic deliberation to facts about what’s probably true given their evidence. Hence, Sosa’s alleged examples of justification without respect for truth are only examples of justification without appraisal respect for truth, not without recognition respect for truth. If we stipulate that the subjects lack recognition respect for truth – i.e. that they don’t appropriately take truth-related reasons into account – my intuitions change. The cases then don’t seem to be cases of justified belief. Recognition respect for truth doesn’t require love of truth or the norms of truth: one might wish that one could avoid playing by these rules, despite playing by them on pain of epistemic irrationality. All it requires, in Darwall’s idiom, is appropriately taking truth-related considerations into account. That is another way of saying that it only requires being heedful of the epistemic reasons. That is intuitively required for justification. KR also avoids Sosa’s unwanted blurring of the moral and the epistemic. Admittedly, talk of respect sounds more natural in the moral case. But the general notion of recognition respect isn’t distinctively moral, nor is the moral case the only case in which recognition respect matters. One can act with or without recognition respect for any rule, and exhibiting recognition disrespect is sufficient for blameworthiness relative to the rule as a general matter. The epistemic case is just a special case: failures of respect for truth make epistemic blame fitting, manifestations of weak respect make epistemic excuse fitting, and strong respect entails justification. KR only 

Darwall () now uses recognition respect in an essentially second-personal way, but the original notion was more general, and could take mere objects and norms as objects.

Responsibilism within Reason



commits us to a structural analogy between epistemology and morality, and doesn’t automatically yield moral encroachment. I conclude, then, that none of the strongest philosophical objections to earlier versions of AR extend to KR.

.

Concluding Remarks

I think I’ve done enough to show that the stock objections to AR rest on mistaking its periphery for its core. Of course, KR’s mere avoidance of objections doesn’t give us sufficient reason to accept it, let alone prefer it to reliabilist VE or non-virtue-theoretic views. But while a wholesale defense of KR would take another paper, we can conclude with some reasons to prefer it. The main point to make is that KR shares the attractions but lacks the vices of its competitors. Let’s take the non-virtue-theoretic competitors first. By resting on a substantive appeal to reasons-sensitivity, KR absorbs the attractions of evidentialist approaches. But it avoids their flaws for the same reason. Perhaps most importantly, as I’ve argued elsewhere, the requirement that justified beliefs conform to reasons as a manifestation of sensitivity to reason-relations addresses the improper basing objections reliabilists have pressed. The same feature enables KR to circumvent the speckled hen problem: while the belief that the hen has  speckles might conform to the evidence, it couldn’t do as a manifestation of a disposition to conform. Similar claims go for the comparison with reliabilism. Sensitivity to objective reason-relations entails reliability, and so KR secures the payoffs yielded by making reliability necessary for justification. But because KR does not take reliability to be sufficient for justification, it promises to avoid the clairvoyance problem, especially if reasons-sensitivity is understood as sensitivity to facts to which one has conscious access (as I did). Moreover, by drawing a distinction between rationality and justification that rests on a principled distinction among reasons drawn outside epistemology, the account also avoids the new evil demon problem. That problem collapses 

 

KR is not incompatible with moral encroachment. One could hold a substantive view of respect for truth that requires testimonial justice, where this is morally encroached. Indeed, Fricker’s () responsibilist epistemology of testimony resembles my general epistemology, similarly resting on a framework of reasons-sensitive dispositions. By avoiding the ideology of character, KR also avoids the empirical objections of situationists, as I argue in Sylvan . See Sylvan a and Lord and Sylvan forthcoming.



 

into the problem of giving a rationale for denying justification but according an inferior status to demon-worlders, which KR offers. What about KR’s virtue-theoretic competitors? Some reasons to prefer KR to these repeat reasons for preferring KR to pure reliabilism. Unadulterated versions of virtue reliabilism don’t explain the epistemic status lacking in clairvoyants. KR is also preferable to hybrid versions of VE. These can seem ad hoc, giving no unified reason for thinking that both reliabilist and responsibilist properties are constitutive of virtue. KR, by contrast, derives reliabilist and evidentialist themes from a unified core that invokes responsiveness to reasons understood as they are in ethics, as facts that objectively favor acts and attitudes. Furthermore, KR does so without the excesses of previous responsibilisms. I conclude, then, that we have strong reason to take KR seriously as a competitor to virtue reliabilism and non-virtue-theoretic approaches. R E F E R EN C E S Alfano, M. . ‘Expanding the Situationist Challenge to Responsibilist Virtue Epistemology,’ Philosophical Quarterly : –. Alston, W. . Epistemic Justification. Ithaca, NY: Cornell University Press. . ‘Virtue and Knowledge.’ Philosophy and Phenomenological Research : –. Arpaly, N. . Unprincipled Virtue. Oxford: Oxford University Press. Baehr, J. . The Inquiring Mind. Oxford: Oxford University Press. Broome, J. . ‘Reasons,’ in Wallace, R. Jay et al. (eds), Reason and Value: Themes from the Moral Philosophy of Joseph Raz. Oxford: Oxford University Press. . Rationality through Reasoning. Oxford: Blackwell. Code, L. . ‘Toward a ‘Responsibilist’ Epistemology,’ Philosophy and Phenomenological Research : –. Conee, E. and Feldman, R. . Evidentialism: Essays in Epistemology. Oxford: Oxford University Press. Dancy, J. . ‘Critical Study of Paul Grice, Aspects of Reason,’ Philosophical Quarterly : –. . Ethics without Principles. Oxford: Oxford University Press. Darwall, S. . ‘Two Kinds of Respect,’ Ethics : –. . The Second-Person Standpoint. Cambridge, MA: Harvard University Press. Dougherty, T. . ‘Knowledge Happens: Why Zagzebski Has Not Solved the Meno Problem,’ Southern Journal of Philosophy : –. Dretske, F. . Seeing and Knowing. Chicago, IL: University of Chicago Press. . Knowledge and the Flow of Information. Stanford, CA: CSLI Publications. 

See Sylvan forthcoming, for a full defense of this claim.

Responsibilism within Reason



Finlay, S. . A Confusion of Tongues. Oxford: Oxford University Press. Fischer, J. M. and Ravizza, M. . Responsibility and Control. Cambridge, UK: Cambridge University Press. Foley, R. . The Theory of Epistemic Rationality. Cambridge, MA: Harvard University Press. . ‘A Trial Separation between the Theory of Knowledge and the Theory of Justified Belief,’ in Greco, J. (ed.), Ernest Sosa and His Critics. Oxford: Blackwell. Fricker, M. . Epistemic Injustice. Oxford: Oxford University Press. Goldman, A. I. . ‘A Causal Theory of Knowing,’ Journal of Philosophy : –. . ‘What Is Justified Belief?’ in Pappas, G. (ed.), Justification and Knowledge. Dordrecht: D. Reidel. Greco, J. . ‘Two Kinds of Intellectual Virtue,’ Philosophy and Phenomenological Research (): –. . Achieving Knowledge. Cambridge, UK: Cambridge University Press. Gregory, A. . ‘Reasons as Good Bases,’ Philosophical Studies : –. Harman, G. . ‘Virtue Ethics without Character Traits,’ in Byrne, A., Stalnaker, R., and Wedgwood, R. (eds), Fact and Value. Cambridge, MA: The MIT Press. Hieronymi, P. . ‘Responsibility for Believing,’ Synthese : –. Hurka, T. . ‘Virtuous Acts, Virtuous Dispositions,’ Analysis : –. Kagan, S. . Normative Ethics. Boulder, CO: Westview Press. Kolodny, N. . ‘Why Be Rational?’ Mind : –. Kornblith, H. . ‘Linda Zagzebski’s Virtues of the Mind,’ Philosophy and Phenomenological Research : –. . ‘Knowledge Needs No Justification,’ in Smith, Q. (ed.), Epistemology: New Essays. Oxford: Oxford University Press. Korsgaard, C. . The Constitution of Agency. Oxford: Oxford University Press. . Self-Constitution. Oxford: Oxford University Press. . ‘The Activity of Reason,’ in Wallace, R. J., Kumar, R., and Freeman, S. (eds), Reasons and Recognition. Oxford: Oxford University Press. Kvanvig, J. . The Intellectual Virtues and the Life of the Mind. Lanham, MD: Rowman & Littlefield. Littlejohn, C. . ‘Fake Barns and False Dilemmas,’ Episteme : –. Lord, E. and Sylvan, K. Forthcoming. ‘Believing for Normative Reasons: Prime, Not Composite,’ in Bondy, P. and Carter, J. A. (eds), Well-Founded Belief: New Essays on the Basing Relation. London: Routledge. McDowell, J. . ‘Virtue and Reason.’ The Monist : –. McHugh, C. . ‘Exercising Doxastic Freedom,’ Philosophy and Phenomenological Research : –. Nelkin, D. . Making Sense of Freedom and Responsibility. Oxford: Oxford University Press. Olin, L. and Doris, J. . ‘Vicious Minds,’ Philosophical Studies : –. Pettigrew, R. . Accuracy and the Laws of Credence. Oxford: Oxford University Press.



 

Parfit, D. . On What Matters (Vol. ). Oxford: Oxford University Press. Roberts, R. C. and Wood, J. C. . Intellectual Virtues. Oxford: Oxford University Press. Scanlon, T. M. . What We Owe to Each Other. Cambridge, MA: Harvard University Press. . ‘The Unity of the Normative,’ Philosophical Studies : –. . Being Realistic about Reasons. Oxford: Oxford University Press. Schroeder, M. . Slaves of the Passions. Oxford: Oxford University Press. Silverstein, M. . ‘Reducing Reasons,’ Journal of Ethics and Social Philosophy (): –. Sosa, E. . ‘The Raft and the Pyramid,’ Midwest Studies in Philosophy : –. . Knowledge in Perspective. Cambridge, UK: Cambridge University Press. . Judgment and Agency. Oxford: Oxford University Press. Sylvan, K. a. ‘Epistemic Reasons I: Normativity,’ Philosophy Compass : –. b. ‘Epistemic Reasons II: Basing.’ Philosophy Compass (): –. . ‘Responsibilism out of Character,’ in Fairweather, A. and Alfano, M. (eds), Epistemic Situationism. Oxford: Oxford University Press. . ‘Knowledge as a Non-normative Relation,’ Philosophy and Phenomenological Research : –. . ‘An Epistemic Nonconsequentialism,’ Philosophical Review (): –. Forthcoming. ‘Can Performance Epistemology Explain Higher Epistemic Value?’ Synthese. Sylvan, K. and Sosa, E. . ‘The Place of Reasons in Epistemology,’ in Star, D. (ed.), The Oxford Handbook of Reasons and Normativity. Oxford: Oxford University Press. Thomson, J. J. . ‘The Right and the Good,’ The Journal of Philosophy : –. Way, J. . ‘Reasons as Premises of Good Reasoning,’ Pacific Philosophical Quarterly. Online first. DOI: doi:./papq. Wolf, S. . Freedom within Reason. Oxford: Oxford University Press. Worsnip, A. . ‘The Conflict of Evidence and Coherence,’ Philosophy and Phenomenological Research : –. Zagzebski, L. . Virtues of the Mind. Cambridge, UK: Cambridge University Press. . ‘Intellectual Motivation and the Good of Truth’ in DePaul, M. and Zagzebski, L. (eds), Intellectual Virtue. Oxford: Oxford University Press. Zimmerman, M. . ‘Taking Luck Seriously,’ The Journal of Philosophy : –.

Index

AAA,  abilism, –, – ability cognitive, , –, , –, , –, –, , –, –, , , –,  extended, ,  innate,  intellectual, –, See also ability, cognitive introspective,  nature of,  recognitional, –, , , ,  reflective,  accessibility, , , , –, , , ,  achievement, , , , –, , , , , –, , –, , , , , –, –, , – cognitive, , , –, –, , , , , –,  intellectual, , ,  practical,  Adams, Robert, , – aesthetics,  agency, , , , , – cognitive, – agent epistemic, ,  moral,  ampilism, –, – anti-reductionism (about testimony),  appropriate condition, –, , , –, ,  apt belief. See belief, apt aptness, , , , –, , , See also belief, apt attentiveness,  autonomy, 

Baehr, Jason, , , ,  barn facade cases, , , –, , –,  Battaly, Heather, , ,  belief ample,  apt, , , , –, , ,  heedful, , – norm of,  perceptual, , , –, , –, , , ,  respectful, – safe, See safety sensitive, See sensitivity Boghossian, Paul,  Bradford, Gwen,  Burge, Tylar, – character cognitive,  intellectual, , ,  character trait, –, , , –, , –, , , , –, , , , –, – character-based virtue epistemology, , , See also virtue responsibilism charity,  Chrisman, Matthew, –,  clairvoyance cases, , , ,  closed-mindedness, , –, –, – cognitive ability. See ability, cognitive cognitive character. See character, cognitive coherence,  coherentism,  competence, , –, –, , , , , –, –, –, , –, , , , , , ,  cognitive, ,  epistemic, , 





Index

competence (cont.) perceptual,  reliable, , , –,  visual, ,  competence-based virtue epistemology, , See also reliabilism, virtue conscientiousness, , , –,  courage, –, , , , , , –, –, –, ,  creativity, ,  credit, , ,  curiosity, –, , , , –,  Dancy, Jonathan,  defeater,  deontology, –, –,  difficulty, –, –, –,  disposition, , , , , –, , –, , –, , –, –, , , –, , , , , , , –, , , , –, ,  cognitive, –, ,  reliable, –, ,  dogmatism, –, –, , –,  education, , , ,  effects vice, , , –, , –, , – empiricism, , – entitlement, , –, –, –, –, – epistemic community,  competence, See competence, epistemic content, , , –, –,  corruption,  good, , , , , –, ,  malevolence, –,  motivation, , ,  motive, , –, , , ,  norm, –, –,  normativity, –, , , , , , , , , –,  obfuscation,  pollution,  responsibility, , ,  worth,  epistemology anti-luck, –, –, , – anti-risk, , , –, –, – contextualist,  knowledge-first, , 

modest virtue, , –, , See epistemology, anti-luck; epistemology, anti-risk performance, ,  regulative,  reliabilist, , See also reliabilism robust virtue, –, , –, , –, –, – traditional, , , ,  vice,  Erik, Jaentsch,  ethics, –, , , , , , , , –,  normative, ,  virtue, –, , ,  eudaimonia, –, ,  evidentialism, –, , , , ,  externalism, , , , , –, ,  faculty, , , , , ,  cognitive,  visual,  final value. See value, final flourishing, , , ,  foundationalism, –, ,  Fricker, Miranda, –,  generosity,  Gerken, Mikkel,  Gettier problem, , , , , , , , , , –, ,  Graham, Peter,  Greco, John, , –, , , , ,  happiness, ,  heedfulness, , –, ,  honesty,  humility, –, , ,  Hyman, John,  inappropriate condition, , –, , ,  inquiry, , , , , , , , – aim of, ,  norm of,  virtue of, – intellectual arrogance, ,  perseverance, – vice, , , , , , , , – virtue. See virtue, intellectual intelligibility problem, , –, –, –,  interlocution, , , , See testimony

Index internalism, , , , , , , –, , –, –, –,  introspection, , –,  intuition the ability,  rational,  invariantism,  justice, , ,  justification, , –, –, –, –, –, , , , –, –, , , , –, –, , –, , – doxastic,  meta-, – objective,  subjective,  Kallestrup, Jesper, –,  Kantian responsibilism, , ,  Kelp, Christoph,  kindness, ,  knowledge action and,  analysis of, , ,  animal, , , –,  attribution of, ,  concept of, , ,  condition on, , , , , , , , , ,  definition of,  high-grade, –, –,  nature of, , –, , , , ,  perceptual, , , , –, , , ,  a priori,  propositional, , ,  reflective, , – scientific,  source of,  structure of, ,  testimonial, , See also testimony; interlocution theory of, , –, –, , –,  understanding and, , ,  value of, , , , , –, , , –, , , , See also value, epistemic Kornblith, Hilary,  Kriegstein, Hasko von, ,  Lackey, Jennifer,  lottery problem, , –, 



luck, , , –, , , –, –, , –, –, , –,  environmental, –,  epistemic, , –, , –, – Maitra, Ishani,  manifestation, , , , –, , , –, , , –, –, , ,  mentalism,  Miscevic, Nenad,  Montmarquet, James, ,  moral responsibility, ,  moral virtue. See virtue, moral Neta, Ram,  Nguyen, Christopher Thi,  normal condition, , –,  normal function, – normality,  normativity, , –, , , , –, –, , ,  epistemic, See epistemic normativity moral, ,  practical,  open-mindedness, –, –, , , , , , –, , –, ,  perception, , , , , –, , , , , , , ,  perceptual prejudice,  performance, –, , , , , –, , –, , –, –, –, , –, , ,  ample,  apt, ,  cognitive, ,  personalist vice, , , –,  Pollock, John,  practical reasoning, , ,  pragmatic encroachment, ,  Pritchard, Duncan, , –, , –,  proper function, , , , , – propositionalism, , , – rationality, , , , , , –, ,  reasoning deductive, , , –, –,  inductive, , , –, ,  reductionism (about testimony), 



Index

regress problem, , , ,  reliabilism, –, , –, , –, –, , , , –, –, , , , , , , , , , –, ,  simple, ,  virtue, –, , , , , –, –, –, , –, , ,  responsibilist vice, , , –, –, – Riggs, Wayne,  risk, , , –, –, –,  epistemic, , – Roberts, Robert Campbell, , , – safety, , , , , , –, –, –, , , , –, , ,  Sellars, Wilfrid, –,  sensitivity, –, , ,  situationism, , – skepticism,  Sosa, Ernest, , , –, , , , , ,  SSS, , –, –,  statism, –, , –, ,  Steward, Helen,  Sylvan, Kurt, ,  teleology, –, , –, –, , , , ,  temperance, , , – testimonial injustice, –, –,  testimony, , , , , , 

thoroughness,  Turri, John, – value of achievement, , ,  of apt belief,  epistemic, , –, –, ,  final, , , , ,  instrumental,  intrinsic,  of knowledge. See knowledge, value of moral,  practical, , ,  of truth,  value problem,  virtue epistemic, See virtue, intellectual intellectual, –, , –, , , –, , , , –, –, –, –, –, , , , –, , –, –,  moral, –, , , , , –, –, , , , , ,  responsibilist, , –, , , –, , –, , ,  virtue responsibilism, –, , , –, –, –, , ,  warrant, , , –, –, –, –, ,  well-being, ,  Wood, W. Jay, , , – Wright, Crispin,  Zagzebski, Linda, , , –, –