A Plea for Epistemic Excuses

A Plea for Epistemic Excuses Clayton Littlejohn King's College London 1. The Demoniacs A typical epistemology course begins with a discussion of the distinction between justification and knowledge and ends with no discussion of the distinction between justification and excuse.1 This is unfortunate. While the topic of excuses might not be particularly sexy, a better understanding of the justification-excuse distinction should shed light on important debates in epistemology. In particular, it should help us understand the significance of some intuitions that shape the internalism-externalism debate. I have two aims in this paper. The first is to fill a lacuna in the literature by providing an account of how epistemic excuses work. The second is to use this account to shed light on Cohen's (1984) new evil demon argument. Let's start with the second thing. The demoniacs are non-skeptical internalists about justification who think that intuitions about Cohen's thought experiment show us that internalism is true.2 His thought experiment is simple. We imagine a subject in the 1 I want to thank Maria Alvarez, Matt Benton, Cameron Boult, Jessica Brown, Charles Cote-Bouchard, Stewart Cohen, Juan Comesaña, Trent Dougherty, Julien Dutant, Jeremy Fantl, Claire Field, John Gibbons, John Hawthorne, Frank Hoffman, Chris Kelp, Jon Kvanvig, Maria Lasonen-Aarnio, Dustin Locke, Arturs Logins, Matt McGrath, Rachel McKinnon, David Owens, Tom Pink, Mark van Roojen, Nico Silins, Mona Simion, Chris Tucker, John Turri, Brian Weatherson, Jake Wojtowicz, and Tim Williamson for discussing these issues with me. 2 Strictly speaking a 'demoniac' is someone supposedly possessed by a demon. I don't assume that demoniacs in philosophy are demoniacs in this sense, but I do assume that they have good senses of humor. On the present usage demoniacs are internalists about justification who believe that the new evil demon argument shows that internalism is true. An internalist wouldn't be a demoniac if they didn't think that we could use Cohen's thought experiment to establish the truth of internalism. The arguments from Unger (1975) might be taken to support internalism without supporting the demoniac view. (If it's not possible to rationally believe anything about the external world, internal duplicates couldn't differ in terms of what they could justifiably believe about the external world.) Huemer's (2006, this volume) argument for phenomenal conservatism/internalism might be thought of as an argument for internalism and it good case, one who knows things about her environment. We then introduce the idea of an epistemic counterpart, a subject who is perfectly alike in terms of the 'inner' aspects of their mental lives (i.e., someone with just the same beliefs, the same apparent memories and perceptions, the same intuitions, the same good epistemic character, and so on who reason just as the subjects in a good case do). On (some) externalist views, it's possible for epistemic counterparts to differ in terms of what they know and in what they justifiably believe. Although counterparts invariably believe the same things, these externalist views imply that their beliefs can differ in justificatory status. A bedeviled subject, deceived by a Cartesian demon, would believe everything that her counterparts in good cases know to be true but wouldn't be justified in holding these beliefs.3 However similar things are in the good and bad cases, the similarities don't ensure that the deceived subject's beliefs are justified. Cohen thinks that this shows that the externalist views are mistaken: [T]he evil demon hypothesis ... uncovers a defect in the Reliabilist position ... on a Reliabilist view ... the beliefs generated by those processes [operative in the bad case] are never justified. Is this a tenable result? I maintain that it is doesn't rest on the intuitions that Cohen hopes to elicit. Smithies' (2012) argument for internalism similarly doesn't appeal to these intuitions. Naylor's (1988) arguments would seem to support internalism (albeit about personal justification, not doxastic justification) by appeal to assumptions about the relationship between justification and responsibility. Finally, Kant defended a kind of internalism about justified action and his defense doesn't seem to appeal to premises about epistemic constraints on norms and normative conditions. For a discussion of Kant's case for internalism, see Herman (1993). 3 On one view, the beliefs aren't justified because the processes that produce them aren't reliable. On another, the beliefs aren't justified because the beliefs aren't apt. On another, the beliefs aren't justified because they don't constitute knowledge. Some philosophers defend views that they describe as externalist that are designed to vindicate the judgment that a bedeviled subject's beliefs aren't justified. In my view, these views are unmotivated (because we shouldn't try to accommodate the relevant intuition) and the resulting accounts of justification often seem unnatural. It's possible to have a view on which some demons but not all demons rob you of justification (a possibility that allows someone to say that their view isn't an internalist view) but the resulting account of justification seems gerrymandered since our intuitions about justification tend not to track the subtle distinctions between kinds of demonic deception. not ... [P]art of what the hypothesis entails is that our experience is just as it would be if our cognitive processes were reliable. Thus, on the demon hypothesis, we would have every reason for holding our beliefs that we have in the actual world. It strikes me as clearly false to deny that under these circumstances our beliefs could be justified. If we have every reason to believe e.g., perception is a reliable process, the mere fact that unbeknown to us it is not reliable should not affect its justification-conferring status (1984: 281). We can extract from this an argument against externalism: The New Evil Demon Argument P1. In the bad case, the subject's beliefs don't have the properties that figure in externalist accounts of justification. P2. The beliefs are equally justified in the good and bad case [Equality].4 P3. The beliefs are justified in the bad case [Sufficiency]. C. So, the internal properties are sufficient for justification in the good case and bad. The external properties aren't necessary for justification. This argument rests on two verdicts, Equality and Sufficiency. Some people find these verdicts intuitive.5 Taken together they tell us that justification does not depend upon 4 Some authors (e.g., Bach (1985), Engel (1992)) think that the believers are equally justified even though their beliefs are not. For a discussion of the distinction between personal and doxastic justification, see Lowy (1978). The distinction is inspired by the more familiar distinction between the assessment of agents and their actions. See Bennett (1995) and Moore (1997). 5 I don't know of any study of philosophers' intuitions about these cases, but empirical research into folk intuitions about theories of justification in the context of the law suggest that the folk think that justification depends upon the objective nature of an agent's deeds, not (just) the agent's mental states. See Darley and Robinson (1998) for discussion. Huemer's (this volume) suggestions about how 'justified' works in English does not fit well with their findings. how things are external to the subject (nothing external to the subject could constitute a harm or benefit) and that the conditions common to the good case and the bad are all we could ever need for justification. I've argued elsewhere that Equality undermines Sufficiency on the grounds that Equality leads to skepticism. In this essay, I'll bracket the skeptical concerns and focus on the intuitive motivation for these verdicts. Quite apart from the skeptical worries, I have some general concerns about the demoniac approach to justification and related notions. In the framework that I like to use when thinking or talking about justification, justified belief is understood as the belief that's permitted or proper. Permission or propriety is understood in relation to norms. Norms will identify conditions under which some subject should or shouldn't respond in certain ways (e.g., if you acquire evidence for p, you shouldn't decrease your confidence in it; if you aren't in a position to know p, you shouldn't believe it; if p is incompatible with your evidence you shouldn't believe it; if you discover that a peer disagrees with you about p, you should decrease your confidence, etc.). A belief will be justified iff a thinker could believe p without violating any of the norms or would only violate a norm because there is some further overriding norm that requires it. On this way of thinking, if you think, as I do, that there is a norm that says that we shouldn't believe what we don't know and shouldn't come to believe what we're not in a position to know, there couldn't be a justified belief that didn't constitute knowledge, not if there's no overriding norm that requires us to believe what we wouldn't know. It is just this sort of view about justification that Cohen and his demoniacs would have us reject. His argument, if successful, would force us to revise our views about epistemic norms and deny that there could be norms that had application conditions like this one: KNB: One should not believe p when one is not in a position to know p.6 6 For discussions of intuitions about error cases and their implications for norms of assertion and action, see Douven (2006), Gerken (2011), Kelp and Simion (forthcoming), Locke (2015), and McKinnon (2013, 2015). Two epistemic counterparts could come to believe p when only one such subject conformed to KNB. Since we'd respond sympathetically to both subjects, this is supposed to be an indication that KNB doesn't really tell us when some subject shouldn't believe something and thus doesn't identify any necessary condition on justification. We'll only find the real epistemic norms when we find some that can be demon-proofed. Similar arguments can be run to try to rule out certain environmentalist views about reasons, too. Some of us think that the things that a subject knows about her environment might be among her reasons for believing things, feeling things, and doing things and that such reasons might play an important role in explaining why her beliefs, feelings, and actions are justified: KPR: The facts that one knows (including facts about the environment) are potential reasons that could be one's reasons for believing things, feeling things, or doing things.7 Using Cohen's cases, some would argue that it couldn't really be that a subject's reasons for believing things, feeling things, or doing things are facts about her environment on the grounds that this implies that subjects in the good case might be able to believe, feel, or do things for reasons that couldn't be her counterparts' reasons for believing, feeling, or doing these things. A subject's reasons in any case are just the reasons her counterparts have in the bad case, so they'd either be facts she knows about herself, the mental states they all share in common, or propositions (true or false) that these subjects have in mind.8 7 For criticism of such views, see Comesaña (2016), Fantl and McGrath (2009), Gibbons (2010), Rizzieri (2011), and Turri (2009, 2013). 8 Some authors want say (with Cohen) that propositional justification is common to the good cases and the bad while insisting (against Cohen) that counterparts might possess different bodies of evidence or reasons. (See Alvarez (2010), Hornsby (2007, 2008), Hyman (1999, 2006), Littlejohn (2012), Lord (this volume), Mitova (forthcoming), Unger (1975), and Williamson (2000) for defenses of this view.) They might also think (against Cohen again) that these subjects' reasons for believing things, feeling things, or doing things might differ. (See Dutant (this volume), Hornsby (2007, 2008), and Lord (this volume) for discussion.) Like Cohen, I think that this last claim is dubious, but I think the first two claims are true. While there might be differences in the reasons that explain the actions and attitudes of subjects in good and bad cases (e.g., in both cases a reason why a subject believes she has hands is that things look a certain way but it's only in one case that a reason why a subject believes she has hands is that she's deceived by a I have a few reservations about the demoniacs' argument. First, I think that KNB and KPR are correct, so naturally, I'm inclined to think that arguments to the contrary are unsound. Because it seems that there are considerations that support these claims, it seems perfectly sensible to me for people to try to work towards reflective equilibrium. I don't think the best way to reach that state is to accept the demoniac's argument without considering the case for alternatives, particularly if, like me, you just don't find Sufficiency or Equality all that intuitive. Second, I don't think that the strategy is a good one to pursue. Some writers (e.g., Kant) defend internalist views on substantive grounds. They start by identifying the things that matter and then try to work out an account of the right or the justified keeping their eyes fixed on substantive considerations. In the end, they might offer internalist views of justification or rightness according to which the stuff that rightness is made of is all 'in the head', but they might not have reached those conclusions by first trying to show that the only conditions that can figure in the application conditions of norms are conditions that satisfy epistemic constraints (e.g., that we cannot fail to see that they obtain when they do or that we cannot be fooled into believing that they obtain when they don't). I don't think that genuine norms have to be demon-proof. If it turns out that the demons are very good at their jobs and there are no demon-proof norms, it shouldn't turn out that there are no norms. demon), this difference in explanatory reasons gives us no reason to think that there can be differences in motivating reasons (i.e., a subject's reasons for φ-ing). One worry I have about so-called 'disjunctive' approach to the subject's reasons for actions is that the view implies that while subjects in good and bad cases will both have their reasons for φ-ing only one subject will be right about what their reasons were for φ-ing. If my reason for voting for Hillary is, say, that she has more sensible policies and I know that what I've just said is true, it would be weird to think that I'd have a counterpart who believes all and only the same things, feels all and only the same things, does all and only the same things but then votes for Hillary where their reason for voting for Hillary wasn't my reason for casting that vote even though they'd insist that their reason for voting for Hillary was precisely that she had more sensible policies. While I don't think we have any special authority over whether there was anything that was our reason for φ-ing (just like I don't think we have any special authority over whether there was some individual we were thinking about or some existing that we saw or were aware of), I think we have some authority over what, if anything, our reasons would be. When we consider Cohen's bedeviled subjects, it's hard not to feel sympathetic. The demon arranges things so that these subjects are completely ignorant of what's happening in their surroundings. In spite of these differences, the 'inner' aspects of their mental lives are just like our mental lives. They reason just as we do. We know that we'd believe just what they'd believe, that we'd be fooled just as they've been fooled, and that there aren't better ways of reasoning that would lead us not into deception without making unreasonable sacrifices. The interesting question isn't whether we should feel sympathy for the bedeviled. We should. This is the demoniacs' starting point. As they see things, our sympathetic response to the bedeviled subjects is just the recognition that these subjects' attitudes are justified. In explaining why they accept Equality and Sufficiency, they seem to operate with two assumptions: Sympathy for the Bedeviled: We should respond sympathetically whenever some subject φ's and is the epistemic counterpart of a subject who φ's in a good case. Recognition Thesis: This sympathetic response (if fitting) is the recognition that the relevant subject's φ-ing was justified. These assumptions aren't premises in the new evil demon argument, but they capture the demoniacs' take on the relevant cases. If we reject Sympathy for the Bedeviled, we'd have to reject the kind of internalism that demoniacs defend. If the demoniacs rejected the Recognition Thesis, they'd have to accept that the kind of sympathetic response we have to bedeviled subject is not the recognition that the relevant subjects' actions or attitudes are justified. At best, this kind of response would only sometimes (if fitting) indicate that the relevant actions or attitudes are justified. If only this weak connection holds between our sympathetic responses and our recognition that the subjects' attitudes are justified, we might naturally want to know whether there's any good reason to think that our sympathetic response to the original case is any indication that Sufficiency and Equality are correct. The sympathetic response wouldn't show that the externalist view was mistaken. The dialectical situation will be complicated. Let me offer a brief outline of the paper's argument. In §2, I shall offer a general account of how excuses work. In doing so, I'll explain how excuses differ from other kinds of defense (i.e., justifications and exemptions or denials of responsibility). The key idea is that an excuse removes responsibility without (a) identifying a sufficient reason for the agent to violate a norm and (b) without showing that the agent cannot be held accountable for her attitudes or actions. If there can be excusatory defenses that occupy this space, they have to show that the relevant subject's actions or attitudes did not manifest de re unresponsiveness.9 They have to show that the agent was disposed to respond appropriately to normatively significant conditions and that this disposition is responsible for the relevant action or attitude. To do this, we have to show that the agent's actions or attitudes result from the excellent use of her rational capacities. Thus, there must be a gap between (a) φ-ing as a result of using our rational capacities excellently and (b) meeting the standards that we should meet in φ-ing. On this account of excuses, the kind of sympathetic response we feel towards bedeviled subjects is a sign that the subject's actions or attitudes should be excused because they result from the excellent use of the subject's rational capacities but (crucially) this response does not indicate that the subject's actions or attitudes are justified as that requires that the subject conformed to the relevant standards. Whatever standard we use to assess actions or attitudes, it should be possible for someone to fall short even if she uses her rational capacities well (i.e., in just the way that a subject could use them in conforming to a norm and believing or acting with justification). In §3, I shall argue for two points. The first is that a certain kind of rational excellence exculpates. The second is that this kind of exculpation doesn't involve justification, only rationality. I address objections from those who want to identify rationality with justification by trying to show that the excellent use of rational capacities can always fall short and that it should be possible for two agents to use their capacities equally well without thereby doing equally well in conforming to a norm or meeting a standard. In this way, I try to address the demoniacs who insist on identifying rationality with justification and encourage us to 9 Following Arpaly (2002), I take considerations about de re responsiveness and unresponsiveness to be central to responsibility. think that justification is a purely internal matter. Even if justification were a purely internal matter (say, because all genuine norms had internal application conditions), we'd still need to draw a distinction between rationality and justification. In §4, I briefly discuss the newest evil demon problem, a problem that arises for internalism about justification. 2. Pleas I'll approach the Recognition Thesis indirectly. We first need to have a general understanding of the difference between kinds of pleas. Let's start with Austin. In this passage, Austin distinguishes between two ways of removing responsibility: In general, the situation is one where someone is accused of having done something, or (if that will keep it any cleaner) where someone is said to have done something which is bad, wrong, inept, unwelcome, or in some other of the numerous possible ways untoward. Thereupon he, or someone on his behalf, will try to defend his conduct or to get him out of it. One way of going about this is to admit flatly that he, X, did do that very thing, A, but to argue that it was a good thing, or the right or sensible thing, or a permissible thing to do, either in general or at least in the special circumstances of the occasion. To take this line is to justify the action, to give reasons for doing it: not to say, to brazen it out, to glory in it, or the like. A different way of going about it is to admit that it wasn't a good thing to have done, but to argue that it is not quite fair or correct to say baldly ‘X did A’. We may say it isn't fair just to say X did it; perhaps he was under somebody's influence, or was nudged. Or, it isn't fair to say baldly he did A; it may have been partly accidental, or an unintentional slip. Or, it isn't fair to say he did simply A—he was really doing something quite different and A was only incidental, or he was looking at the whole thing quite differently. Naturally these arguments can be combined or overlap or run into each other. In the one defence, briefly, we accept responsibility but deny that it was bad: in the other, we admit that it was bad but don't accept full, or even any, responsibility. Note two things about this passage. First, Austin's examples of excuses are a motley bunch. Among the things that Austin thinks can excuse are accidents or slips, nudges, and epistemic defects or shortcomings. Second, Austin works with a dichotomous scheme: • A justification shows that it's not appropriate to criticize the subject for φ-ing because there was sufficient reason to φ. • An excuse shows that it's not appropriate to criticize the subject for φ-ing even though there wasn't sufficient reason to φ. There are only two kinds of defenses on this scheme. A justification shows that criticism isn't appropriate because there was sufficient reason for the agent to φ. Even if the subject committed some wrong in φ-ing, the justificatory defense is supposed to show that there was sufficient reason for φ-ing in spite of the fact that it was wrongful.10 In offering an excuse, we concede that the subject didn't have sufficient reason for φ-ing but maintain that it wouldn't be appropriate to blame the subject in spite of this. As Austin understands things, excuses cover every unjustified but blameless wrong. Applied to belief, the idea would be that you justifiably believe p iff you believe for sufficient reason.11 You should be excused for believing p iff you lack justification but are 10 Bear in mind that wrongs needn't be wrong all things considered. I fear that most epistemologists will read 'sufficient reason' as something like 'sufficiently strong evidence'. That's not how I intend it. In the case of action we cannot say whether your reason for φ-ing was sufficient until we know what it was and what reasons there were that spoke against φ-ing. Thus two subjects could φ for the very same reasons and only one of them φ for a sufficient reason. In the case of belief, I want to allow for the same sort of possibility. It's possible that two subjects believe p for the very same reasons where only one subject's reasons are sufficient. The first subject, but not the 11 somehow blameless in spite of this. To embed this within a larger framework, we could start with some norm(s) that govern belief with this form: you shouldn't believe p unless C obtains. These norms alert us to the kinds of conditions that would prevent a belief from being justified and a belief counts as justified iff it doesn't violate any such norm without sufficient reason (where the reasons that might justify such violations would be specified by further norms). A belief is excusably held iff it's held in violation of such a norm but the believer isn't to be blamed for this. You might have noticed that this doesn't tell us much about what justifies or excuses belief. Austin gives us no positive account of what excuses are or how they get their work done. As you can see above, all there is to Austin's characterization of an excuse is that they are defenses that show that it's not appropriate to hold someone responsible that differ from justificatory defenses. There's no positive account of the mechanisms by which excuses exculpate. Maybe Austin thought that this could not be done, perhaps because excusing conditions are too motley a bunch. If we work with Austin's dichotomous scheme and his open-ended conception of excuse we should be skeptical of contrast arguments: 1. This defense, D, differs from some paradigmatic case of an excusatory defense, D', in some salient way (i.e., the underlying mechanisms by which they remove responsibility differ). 2. Thus, D is not an excusatory defense. 3. All successful defenses are either justifications or excuses. 4. D, if successful, must be a justification. Anyone who accepts Austin's dichotomous scheme will accept (3), but they shouldn't think that there's any good inference from (1) to (2). You wouldn't reason in that way if second, might have had defeating evidence available to her. It's an interesting question whether the reasons that apply to a subject and prevent her reasons for believing from being sufficient to justify belief are all available to the subject or part of her evidence. At this stage, we should leave the issue open. In the paper's final section, I shall point to cases that I think show that the reasons that bear on whether to believe p needn't be accessible to the subject or part of her evidence. you thought that we couldn't isolate some single mechanism that's operative in every excusatory defense. Consider an example. Any sensible person working with this scheme would recognize infancy as an excusing condition. What about cases in which women kill or attempt to kill an abusive partner? Should these women say that their actions were justified or excused? It seems that a justification should only be available in those cases where they had sufficient reason to harm or kill in self-defense. If so, what about the cases where the grounds for self-defense were absent? Should we say that the killing or attempt was excused only insofar as the woman could plausibly claim that she was like an infant in that she lacked the rational capacities necessary for assuming responsibility for her deeds? That would be degrading. The woman who judges on the available evidence that she's in mortal danger and uses force to defend herself shouldn't have to argue that she was somehow like an infant or a mentally ill person to defend herself from criticism if it turns out that the threat was merely apparent or that she could have defended herself with less force. For this sort of reason, it's generally recognized that contrast arguments are just bad arguments. We should be able to offer an exculpatory defense that doesn't amount to a justification that doesn't purport to show that the person being defended lacks the rational capacities needed to assume responsibility for her deeds, but contrast arguments close off this possibility.12 It's not hard to find contrast arguments in the literature on epistemic justification. 13 To combat this suggestion that Cohen's deceived subject is merely excused, people often argue as follows: 12 See Ferzan (2011) for a helpful discussion of the history of the battered woman defense. 13 See, for example, Bird (2007: 97), BonJour (2002: 248), Cohen and Comesaña (2013: 18), Fantl and McGrath (2009: 125, fn. 23), Ichikawa (2014), Lackey (2007: 606), Locke (2015: 83), Madison (2014, forthcoming), Pryor (2001: 117), Rizzieri (2013: 103), Russell (2001: 39), and Wedgwood (2002: 349). Other writers (e.g., Douven (2006)) contrast what's excusable with what's reasonable without explaining why such a contrast is important. It should be noted that while Wedgwood (2002) thinks there's an important contrast to be drawn between cases of, say, insanity and cases of mistaken belief based on hallucination, he focuses on rationality rather than justification. My critical remarks do not apply to his discussion. We can also find contrast arguments in 1. The subject's beliefs aren't just blamelessly held. These beliefs differ in significant and salient ways from the merely blameless beliefs (e.g., beliefs held by people who are brainwashed). 2. Thus, it would be a mistake to say that the reason the subject shouldn't be blamed for believing what she does is that she should be excused. 3. If they aren't just excusably held, they must be justifiably held. 4. Thus, if we can defend her from criticism, our defense must be justificatory. In offering examples of excusable epistemic failures, standard examples involve drugs that interfere with your rational faculties and capacities (Steup), incapacitation (Steup), brainwashing (Madison, Smithies), and delusion or insanity (Gerken, Wedgwood). It's obvious that there's some difference between erring because you've been drugged or brainwashed and erring because you've reasoned impeccably from some perceptual beliefs formed in response to illusion or hallucination, so we're supposed to conclude from this that the difference here is the difference between justification and excuse. As Madison succinctly puts it: Mere blamelessness is insufficient for the positive epistemic status of the beliefs in cases of subjects and their evil genius-deceived counterparts. It is therefore implausible to suggest that the traditional internalist’s judgments of sameness of justification in these cases could arise merely by confusing these two notions (2014: 69.) The main problem with this kind of argument is that it seems to assume that all successful non-justificatory defenses are somehow alike (e.g., someone who has been brainwashed, say, should be excused for believing what she does, so other excuses should other literatures. See Rosen's (2014) arguments concerning duress, justification, and excuse. be like that). We're invited to think that Cohen's subject isn't anything like a subject who has been brainwashed or drugged the subject's beliefs couldn't merely be excused and hence must be justified. It would, of course, be silly to deny that there's some difference between the cases in which someone follows the evidence and arguments where they lead and forms a false belief and the case in which someone is brainwashed, but it's still not clear why we'd account for this difference by saying that brainwashing excuses and systematic deception justifies. Why shouldn't we say that this is just another case of excuse, albeit one in which the grounds for excusing differ from those that are present in cases of brainwashing? If the category of excuse is sufficiently heterogenous, we could say that insanity and ignorance both function as excuses, albeit excuses that do their work in different ways. Because of an unfortunate tendency to think that contrast arguments help us identify defenses as justificatory, it might be helpful to consider an alternative scheme to Austin's, Strawson's (1962) trichotomous scheme: • A justification shows that it's not appropriate to criticize the subject for φ-ing because there was sufficient reason to φ. • An excuse shows that it's not appropriate to criticize the subject for φ-ing even though there wasn't sufficient reason to φ and even though she has the intact rational capacities necessary for being held accountable for having φ'd. • An exemption shows that it's not appropriate to criticize the subject for φ-ing on the grounds that she either lacked the rational capacities needed for assuming responsibility for her actions and attitudes or because these rational capacities were compromised. On this view, there three kinds of defenses: justifications, excuses, and exemptions.14 To show that something is justified, we have to show that no wrong was committed or that it was done for adequate reason. To show that a subject should be exempted, we'd identify global exculpating conditions that would preclude us from holding the subject accountable generally. To show that something should be excused (rather than exempted) we'd have to show that there's some exculpatory factor that removes responsibility even though the subject has the rational capacities that would allow us to hold her responsible for her actions and attitudes. If there really are three kinds of exculpatory defenses, it's clear that contrast arguments are bound to fail. A salient difference between the way in which D and D' remove responsibility might show that D and D' belong to different categories, but classifying D as an exemption, say, wouldn't give us any reason to classify D' as a justification rather than an excuse. This more fine-grained scheme has something going for it. People often say that forgetfulness or accidental slips have some power to exculpate, but there's no temptation to say that these must be justifications simply because they differ in salient ways from temporary insanity or infancy.15 While nothing of substance seems to turn on whether we lump all the excuses together or distinguish excuses from exemptions, there is a substantive question as to whether a defense that cites the subject's ignorance or mistaken beliefs justifies or merely excuses. I'll work with the trichotomous scheme and explain why the demoniacs haven't given us any reason to think that the beliefs formed in the bad case are justified rather than excused.16 14 Some authors refer to exemptions as 'denials of responsibility'. For helpful discussions of the trichotomous scheme, see Duff (2007), Ferzan (2011), Gardner (2007), Horder (2004), Tadros (2005) and Wallace (1998). 15 For a critical discussion of the suggestion that children should be exempted, see Boult (forthcoming). 16 Although ignorance and mistake can excuse, they don't invariably excuse. One reason that ignorance or mistake might not excuse is that the ignorance or mistake is not nonculpable. It should also be noted that normative mistakes might not have the same significance as factual mistakes. It's one thing to say that someone shouldn't be blamed for hitting someone with their car because they mistakenly believed that there wasn't a cyclist ahead and another to say that someone shouldn't be blamed because they mistakenly believed that cyclists don't matter. I'll have more to say on this below. 3. Rational Excellence, Rationality, and Justification In the trichotomous scheme excuses differ from exemptions in that excuses can do their work even when the relevant agent's rational capacities are intact so that she can assume responsibility for her actions or attitudes. They differ from justifications in that they can do their work even when an agent that can be held responsible shouldn't be held responsible even though she violated a norm that governs her actions or attitudes without sufficient reason for doing so. How is this possible? If this agent can be held to account, why shouldn't we hold her accountable for her wrongs? How could an excuse do its job if it isn't an exemption and isn't a justification? To understand how excuses in the relevant sense are possible, it's helpful to think about what a successful excuse succeeds at doing. When an agent violates a norm and we know that she can be held accountable for her actions or attitudes, an excuse is appropriate only if blame is not. Blame would be appropriate if the agent's actions or attitudes manifested a kind of unresponsiveness to the normative reasons that she acted against or believed against. Thus, the excuse works only if it can show that the subject's actions or attitudes didn't manifest this kind of unresponsiveness without assuming that the subject had sufficient reason to commit the wrong and without assuming that the subject's ability to respond to reasons had been compromised. This is possible only when the subject's actions or attitudes manifested a kind of rational excellence.17 On the assumption that the subject's rational capacities were intact and the subject could be held accountable for her actions or attitudes it would be totally obscure why we shouldn't hold the agent accountable for violating a norm unless we could see the agent's actions or attitudes as manifesting the kind of responsiveness to the relevant normative reasons we'd expect from an agent. If the agent didn't use her rational capacities excellently, we wouldn't see her actions or attitudes as manifesting this kind of responsiveness. Thus, we should see an excuse as something that removes responsibility 17 Lasonen-Aarnio (this volume) and Miracchi (this volume) have interesting discussions of how much this might require (e.g., whether it requires only that in the circumstance the agent responds as the virtuous subject would or whether it requires the manifestation of some virtue). and shows an agent to be blameless by affirming that the subject handled the situation competently, using her rational capacities as excellently as anyone could expect a responsible agent to do. Some authors insist that this kind of rational excellence constitutes a justification. Fantl and McGrath, for example, say in their discussion of Williams' petrol and tonic case that the agent who accidentally poisons the guests by serving them Bernies (i.e., petrol with tonic and lime) acted with justification because it was 'rational' and an action that 'made the most sense'. One reason to be skeptical of this suggestion is that it seems that this would have to be true in Austin's original examples: • The agent was nudged. • It was partly an accident or an unintentional slip. • It wasn't fair to simply say that he was φ-ing because he was doing something else and his φ-ing was merely incidental. • He was looking at the matter quite differently and didn't realize that he was φ-ing. If Austin sorted his cases into the right categories, it's hard to see how an excuse could be in order if the agent's actions weren't ones that 'made the most sense' would be 'rational'. Suppose that someone in the agent's situation could have expected that it was likely that they'd be nudged if they didn't try to do things differently. The nudge wouldn't excuse if they proceeded in spite of this. If instead they could have avoided the nudge but didn't take reasonable steps to avoid it, the nudge wouldn't excuse. Suppose that someone in the agent's shoes should have expected that there was a chance of a slip. Again, if the things they did didn't make sense given that they've taken this possibility into account, the slip wouldn't excuse. The slip excuses only if the subject shows all the care and caution required by the possibility of a slip. These points carry over to Austin's third and fourth cases, cases that look like excuses that work only if the agent was ignorant of some fact or non-culpably mistaken about something. My argument against their treatment of Williams' case might be stated like this. Because we're offering an excuse, not an exemption, we have to show that the agent cannot be blamed even though she has the rational capacities needed to be answerable for her actions or attitudes. If such an agent did not use her rational capacities excellently (e.g., by doing the things that could reasonably be taken to be right, by doing the things it would make sense for a conscientious agent to do), she wouldn't deserve an excuse and would be properly held to account for any norm violations. Thus, the agent can be excused only if she shows excellence in the exercise of her rational capacities. How should we cash out this talk of the excellent use of the subject's rational capacities? I would do so by thinking about the way the agent used her rational capacities and the way that an agent who manages to conform to the relevant norms and escape blame would have used their rational capacities. If, say, she uses her rational capacities in just the way that a subject in a good case does and this subject isn't culpable for her actions or attitudes, she shouldn't be culpable for her actions or attitudes either. Neither could be blamed because neither agent's actions or attitudes would manifest de re unresponsiveness. Since it's hard to imagine how we could see some agent's actions or attitudes as manifesting de re unresponsiveness if they're the same on the inside as some agent who conforms to all the relevant norms, this seems like a decent test for determining whether a subject deserves an excuse. 3.1. On the Distinction between Rationality and Justification In the trichotomous scheme excuses remove responsibility by affirming that the agent handled her situation competently and displayed a kind of rational excellence. While she didn't conform to the norms, this wasn't down to the way she exercised her rational capacities. (The difference between her case and the good case wasn't a difference that could be traced back to the way she handled herself and so couldn't be any indication of de re unresponsiveness.) Assuming that a rational belief or rational action is the result of the excellent use of a subject's rational capacities, excuses (in the strict sense) require rational belief and action. If this is right and excuses aren't justifications, we have to distinguish rationality from justification. Cohen objects to this maneuver. We both agree that the subject in the bad case has rational beliefs. (As I'd put it, these beliefs are the products of the excellent use of the subject's rational capacities, something that, had the subject been in less inhospitable circumstances, would have resulted in knowledge.) I think that our sympathetic response to the bedeviled subject's handling of her situation is just the recognition that her attitudes are rational and an indication that she should be excused for violating any norms that she violated. The trouble with this response, Cohen says, is not so much that it doesn't go far enough; rather, it goes too far for my purposes. He thinks that it doesn't make any sense to distinguish justification from rationality. In saying that the bedeviled subjects' attitudes are rational, I'm committed to the Sufficiency. I can't now challenge the Recognition Thesis because rational belief just is justified belief. Cohen's response assumes this thesis: Identification Thesis: S's φ-ing was justified iff S's conscientious duplicates' φ-ing was rational.18 His initial defense can be found in this passage: ... distinguishing between rational beliefs and justified beliefs does not make any sense. It makes no more sense than distinguishing between moral acts and justified acts. Justification is domain relative. A justification for φ-ing can be moral, rational, legal, etc. It is moral to φ just in case φing is morally justified. In the same way, it is rational to F just in case one’s φ-ing is rationally justified. So it is rational to believe just in case believing is rationally justified. When one distinguishes between rational belief and justified belief, one can only be distinguishing between rationally justified belief, and some other kind of justified belief. What is the other kind of justification (MS: 5)? 18 This has to be qualified so that it's clear what kind of justification is at issue and restricted to agents that are conscientious to handle cases of rational but amoral agents. For criticism of the thesis, see also Littlejohn (2012) and Sylvan (ms). I don't see how his points about the domain relativity of justification could support the Identification Thesis. I've suggested that there are different heuristics that we might use to determine whether some subject's actions or attitudes are justified and whether they are rational. They'd be justified iff they conformed to the relevant norms (e.g., epistemic, moral). They'd be rational iff the subject used her rational capacities in the way that some subject in a good case would have. There's no reason to assume from the outset that the actions or attitudes that satisfy this test for rationality will thereby conform to the relevant norms. We can respect his points about domain relativity by saying (e.g.) that it wasn't reasonable or rational in some moral sense for a conscientious agent to φ if her φ-ing as she did couldn't have conformed to moral norms and yet it just doesn't seem to follow that the rational action is automatically justified. It might be thought, however, that the demoniacs could come at this from a different angle. They might try to show that it isn't possible for some subject who satisfies our tests for rationality to fail to act or believe with justification. They might try to show that the real norms that determine whether our actions or attitudes are justified are the kinds of things we couldn't fail to conform to if our actions or attitudes are rational. In other words, they might insist that every agent who uses her rational capacities excellently is guaranteed to conform to the norms that govern action and belief. In some ways, this response is more promising than the first. One cannot simply say that the distinction makes no sense if we have, as I've suggested, different ways of getting a grip on the relevant notions (e.g., justification as a matter of conforming to norms and rationality as a matter of using one's rational capacities excellently so that the use 'matches' the use of some subject in a good case). I shall now argue that there are structural reasons to be dubious of the Identification Thesis. We cannot hope to show that we'll always conform to the norms when we use our rational capacities excellently. If the excellent use of rational faculties is sufficient for rationality, it is not sufficient for justification, not if justification requires conforming to norms.19 19 Of course, if you think that justification doesn't require to conforming to norms as conceived of here or you think that a belief might not be rational if it is the result of the excellent use of one's rational capacities, all bets are off. To see this, consider the following argument: Against the Identification Thesis 1. A knows that there are two norms she must conform to, N1 and N2. N1 says that A must φ if C obtains. N2 says that A must not φ if C doesn't obtain. [Suppose]. 2. C obtains [Suppose]. 3. A rationally believes C doesn’t obtain [Suppose A has strong evidence that C doesn't obtain even though it does]. 4. It is rational for A to believe she must not φ. [(3)]. 5. If it is rational for A to believe she must not φ, A is justified in refraining from φ-ing [Identification Thesis, Enkratic Requirement].20 6. A is justified in refraining from φ-ing [From (4), (5)]. 7. A must φ [From (1), (2)]. 8. A is not justified in refraining from φ-ing [(7)].21 It certainly seems that (1)-(3) could all obtain. There could be pairs of norms (e.g., believe p when there's sufficient evidence for it and don't believe p when there is not; adjust your credence if you get new evidence that bears on p and don't adjust it otherwise) that would require some individual to φ if C and refrain if ~C. It also certainly seems that a subject might have strong evidence that C doesn't obtain in C-worlds. If we assume, as 20 If the Identification Thesis is correct, (5) is equivalent to (5*), 'If it is rational for A to believe she must not φ, A's refraining from φ-ing is rational'. If (5*) is false, it must be possible for A to be rationally required to φ even when A rationally believes that φ-ing is rationally required. Titelbaum (2015a) shows that this violates the Enkratic Requirement (i.e., the wide-scope requirement that says that rationality requires you refrain from both believing that φ-ing is rationally required and failing to φ). For defenses of that requirement, see Broome (2013), Littlejohn (forthcoming), and Titelbaum (2015b). 21 Cohen and Comesaña (this volume) say that there are plenty of things to say about the premises but do not identify which premise(s) they would reject. Huemer (this volume) offers a parody argument, but the parody argument has been discussed elsewhere. See Littlejohn (forthcoming) and Titelbaum (2015b) for responses to it. many demoniacs do, that strong evidence for p entails that it's rational for a thinker with this evidence to believe p even when p is mistaken, we're in business.22 Since (6) and (8) are incompatible, we have to reject one of the first five premises. It isn't hard to undercut the argument if you reject the Identification Thesis. If you reject that thesis on the grounds that you think that there may be more to justification or conforming to a norm than rationality, you would allow that it's possible for situations to arise in which you might use your rational capacities excellently and fail to conform to some norms. This easy response isn't available to the demoniacs who accept the thesis. Nobody could sensibly question the combination of (1) and (2). The demoniacs will try to block the reasoning that leads to (6) and (8) by focusing on (3) and (5). Having granted (1) and (2), they might insist that when we choose our application conditions carefully we'll see that (3) is mistaken. Alternatively, they might take issue with (5). I'll discuss these responses in turn. On (3) The demoniac might challenge (3).23 According to this line of response, either there cannot be strong but misleading evidence about whether C obtains or it cannot be rational to believe on the basis of such evidence when the target proposition is false. Since it makes little sense for the demoniac to say that two thinkers with the same evidence might believe on the basis of the same evidence and end up with beliefs that differ in rational status, we'll just focus on the first disjunct. On this line of response, it isn't possible to have strong but misleading evidence about whether C or ~C obtains. As a general proposal about norms, there cannot be misleading evidence about their application conditions that makes it rational to believe that these conditions obtain when they do not. For this reason, genuine norms can be demon-proofed. 22 See Cohen and Comesaña (this volume) where they say, "Our view is that a subject S is rational in believing a proposition p just in case S’s evidence sufficiently supports p. We also think that a body of evidence can sufficiently support a proposition without entailing it". 23 In personal communication, Cohen suggested that this might be the premise he would question. It's clear that the demoniacs reject the kind of infallibilist view that says that it's only possible to rationally believe a condition obtains when your evidence entails that this condition obtains and this condition does indeed obtain: Restricted Infallibility Thesis: If C is an application condition in a norm, it is not possible to rationally believe C to obtain when it does not obtain.24 It would seem, though, that the demoniac would agree to at least this much of the evidentialist view: Sufficient Evidential Support Thesis: It is rational for S to believe p if S's evidence for p provides sufficiently strong support. We should understand the Sufficient Evidential Support Thesis as requiring a level of evidential support that falls short of entailing support; otherwise, the demoniac would face problems with things like inductive inference. This puts the demoniac in an impossible situation. To see why, let's suppose you think that φ-ing is the kind of thing that might be required in some circumstances and prohibited in others. If so, you should think that there's some norm that states at least one condition under which φ-ing is prohibited and some second norm that states at least one condition under which φ-ing is mandatory. We shouldn't need anything more than C's obtaining and C's failing to obtain to play these theoretical roles. Suppose that C's obtaining requires Agnes to φ and C's failing to obtain requires her to refrain from φ-ing and let's suppose that C obtains. Now, suppose that C's obtaining or failing to obtain doesn't supervene upon Agnes' mental states. It would seem, then, that we can use the possibility of demonic deception to generate this pair of worlds: in w1, Agnes knows that C obtains, knows the normative significance of C's obtaining and failing to obtain, and φs accordingly; in w2, however, Agnes is in the very same mental states as w1, reasons in just 24 If you see section 6 of their paper, it is clear that the entirety of Cohen and Comesaña's attack on Williamson turns on just this point. Let's note one wrinkle. Comesaña (forthcoming) notes (rightly) that if a subject's evidence can be false, a subject can have entailing evidence for falsehoods. the same way, φs accordingly, but C failed to obtain. Our norms would say, of course, that Agnes would be required to φ in w1 and would be required not to φ in w2, but this goes against the demoniac's Demonic Doxastic Demandingness Thesis. The demoniac would thus say that C couldn't have the normative significance we initially thought. Of course, you might be thinking, this just goes to show that the application conditions cannot be constituted by conditions that don't supervene upon the subject's mental states. This isn't surprising because you'd expect that the demoniac would be an internalist about normatively significant conditions. That's right, but notice that the reasoning above suggests that there couldn't be any norms that could explain why φ-ing would sometimes be required and sometimes prohibited unless we found conditions that satisfied the right epistemic constraints. And what are those constraints? The Restricted Infallibility Thesis tells us that a norm's application condition would have to be the kind of condition that you couldn't possibly mistakenly believe to obtain while being rational in that belief. In turn, the Sufficient Evidential Support Thesis tells us that these would have to be conditions that you couldn't have strong but misleading evidence about. If the evidence was misleading but provided strong enough support to make it rational to believe that the condition was met, we'd have our case in which (1)-(3) held. Here's my bold conjecture: there is virtually nothing of recognizable normative significance like this. It isn't hard to imagine cases in which you have strong but misleading evidence about which external conditions obtain. It might seem more difficult to imagine cases where you have strong but misleading evidence about which internal conditions obtain, but it isn't that hard to do it. Ask yourself whether you think that mental states that you were previously conscious of could be the kinds of things that presently have normative significance (e.g., states that obtained a second ago that you seem to recall obtaining). It's not hard to imagine cases in which a demon duplicates your present mental states including the apparent awareness of past mental states. It certainly seems possible that someone makes rational mistakes about very recent mental states, but then the demoniac has to say that such conditions for this very reason couldn't have normative significance for anyone. Maybe time-slice internalism doesn't seem all that bad if you have internalist sympathies, but you aren't giving the demon its due if you think that the demon couldn't make it reasonable for you to form false beliefs about your present mental states. It's hard to believe that the demon couldn't give someone strong but misleading evidence about the mental states that they are in so that they form false beliefs about their own mental lives in the specious present.25 If this kind of evidential situation is possible, even most of the seemingly normatively significant facts about our present mental states would be ruled out as ineligible when it came to stating the application conditions of norms. If we combine the Sufficient Evidential Support Thesis with the pessimistic claim (or optimistic, if you like to root for the bad guys) that the demon could provide strong but misleading evidence for propositions about your present mental life, it starts to look as if there couldn't be any norms at all to explain why our attitudes or actions would be prohibited in some circumstances and required in others. If my conjecture is correct and there is nothing of normative significance that satisfies the conditions imposed by the Restricted Infallibility Thesis (i.e., having sufficient evidence for believing C obtains/doesn't obtain is possible only in Cworlds/~C-worlds respectively), the denial of (3) leads directly to the denial of norms. The worst explanation of Equality and Sufficiency is the impossibility of norm violation. Some things are justified, but not everything. On (5) My opponent might try to block the argument by challenging (5), the premise that says that if it is rational for A to believe she must not φ, A is justified in refraining from φ-ing. There is a simple argument for (5), though, that they should consider. Because my opponent accepts the Identification Thesis, they see (5) as equivalent to the claim that that if it is rational for A to believe she must not φ, A's refraining from φ-ing is rationally permitted. If they reject this, then, they think that it can be rational to both (a) believe that she must not φ and (b) be rationally permitted to φ against this judgment. This, 25 I will present one such example momentarily. however, violates the Enkratic Requirement, a wide-scope requirement that forbids φ-ing while believing that φ-ing is prohibited (or refraining from φ-ing when believing that φing is required). Thus, we have an argument for (5), one that many of my opponents would find compelling. Many writers (especially those with internalist sympathies) think the Enkratic Requirement captures a genuine requirement of rationality. As such, I wouldn't expect that many of them would reject (5) to block the reasoning, but some of my opponents might be willing to incur the cost to save their view or might count costs differently than I do. Because there is some controversy about the requirement, it might be unwise to rest too much weight on it without some further defense of it. While I don't have space to say much in support of the Enkratic Requirement here, I won't offer a defense of it but will explain why the demoniacs shouldn't reject (5). Suppose that Agnes reasons as follows: Agnes' reasoning: P1. I must φ if C obtains but must not φ if C doesn't obtain. P2. C doesn't obtain. C. So, I must not φ. The response we're considering focuses on (5), not (1)-(3). We should be able to assume, then, (1)-(3) and see whether the demoniac can reject (5). Suppose (1)-(3) hold in w1. In w1, Agnes knows (P1), reasonably but mistakenly believes (P2), and thus can come to reasonably believe (C) by means of competent deduction. If there's a counterexample to the Enkratic Requirement, it might seem that w1 gives us just what we need since it seems to be a case in which Agnes both reasonably believes she must not φ and yet must φ nevertheless. The demoniac couldn't see the case just described as a counterexample to the Enkratic Requirement. Suppose that Agnes is in all and only the same non-factive mental states in w2 as w1. Suppose that in w2, however, C does obtain. It would follow, then, that in w2 Agnes would be obligated not to φ (just as she concludes). The demoniac is committed to the following thesis: Demonic Doxastic Demandingness Thesis: It couldn't be that S is obligated to φ in w1 and refrain from φ-ing in w2 if S is in the same mental states in both worlds. The demoniac's Demonic Doxastic Demandingness Thesis rules out the possibility that Agnes' obligations would differ in the way that they'd have to for there to be a counterexample to the Enkratic Requirement. It thus rules out a reading of the case in which a world in which (1)-(3) is true is a world in which Agnes both reasonably believes she must not φ and is nevertheless required to φ. Once (1)-(3) are granted, the demoniac cannot then hope to challenge (5). Their only avenue of response is to reject (3), but we've already seen that that's not a promising line for them to take. On the Significance of our Sympathetic Response It seems to me that internalists and externalists should agree that there can be cases of non-culpable factual mistake that cause trouble for the Identification Thesis. (They might disagree about which cases are relevant because the cases will involve strong but misleading evidence about the application conditions of some norms and they disagree about what those conditions might be, but they shouldn't disagree about the Restricted Infallibility Thesis. They should agree that that thesis is mistaken.) If my conjecture is correct and it is possible to have strong but misleading evidence about whether just about any condition (internal or external) obtains, situations can arise in which a subject who knows the relevant norms will mistakenly come to conclude on the basis of wellsupported beliefs that they are under an obligation to do something that is in fact prohibited (or conclude that they are prohibited from doing something that is in fact required). To deny this, one must either insist that there is more to rational belief than strong evidential support or insist that it's not possible to have misleading evidence about the facts that constitute the application conditions of our norms. Neither claim is prima facie plausible. In my view, when a subject is moved by a mistaken belief to do things that would have been right if the belief hadn't been mistaken about the facts or ignorant of them, we should respond sympathetically and should excuse the subject's failure. Thus, we shouldn't see our sympathetic response to our subjects' handling of error cases as having the significance the demoniacs take it to have. We can put further pressure on the demoniac view by coming at these issues from a slightly different angle. I'll try to show that the sympathetic response the demoniacs try to tap into is misleading. It is not the recognition that a subject's φ-ing was justified. The demoniac tries pressure the externalist to revise her views by considering cases in which it might seem to the subject that some norm's application condition is met when in fact it isn't. When the story is told in the right way, we respond sympathetically to the way that the subject handles her situation. This response, according to the Recognition Thesis, just is the recognition that the subject did succeed in conforming to the relevant norms. Assuming that the sympathetic response is fitting, we have to say that the putative norm wasn't a genuine norm. It couldn't because conforming to it wouldn't have any bearing on justification. We can show that this methodology for testing norms and competing accounts of justification is problematic. The relevant kind of sympathetic response isn't the recognition that the relevant subject's actions or attitudes are justified. Above, I wanted to suggest that if someone had sufficient evidence for falsely believing that a norm's application condition is met, we would think of their response as rational (because the failure to meet the norm was the result of the excellent use of the subject's rational capacities) but not justified (because the actions or attitudes didn't conform to the relevant norm). I'll press a similar point in this section but will focus on the Recognition Thesis and less on the relationship between justification and rationality. Consider an example inspired by Prior (1958).26 The example's function isn't to threaten the Restricted Infallibility Thesis, but to demonstrate the unreliability of the kinds of intuitions that demoniacs use to motivate it and attack proposals about norms that don't satisfy it. Let's suppose that Agnes has just finished her first book. It's a work of non-fiction. I can't recall the topic because it wasn't very interesting, but that's not what matters. What matters is that Agnes was as careful as anyone could be in writing an 26 I first came across Prior's discussion of 'blocks' in Hawthorne (ms). Thanks to John for discussing them and the purposes that they might be put to at length. ambitious piece of non-fiction. She didn't add a sentence to the book without carefully checking it against the evidence. When she sent it off to the fact-checker, the factchecker reported that she found precisely one mistake in the book. As often happens in cases like this, the fact-checker dropped dead before she could tell Agnes where the mistake was found. Following the evidence one more time, she added a sentence affirming that her book contained an error. Let's suppose that Agnes promised three things: (a) She'd wiggle her index finger iff her book contained an error. (b) She'd wiggle her ears iff the propositions expressed by the sincere utterance of the sentences contained in her book formed an inconsistent set of propositions. (c) She'd twiddle her thumbs iff it seemed to her that her book contained an error. Setting the preface aside for the time being, consider (a) and the claims in the body of Agnes' book. I take it that the demoniacs who accept the Demonic Doxastic Demandingness Thesis would be cool to the suggestion that there's really any norm that requires people like Agnes to do things on the condition that her book contains an error. Keeping Agnes' evidence and her mental states fixed, it seems we can imagine two ways of filling out the details of Agnes' case: Flawed: The fact checker was right. There was an error on p. 237. Flawless: The fact checker was wrong. There was no error in the body of the book. The demoniac would (presumably) reason as follows. Look, the difference between Flawed and Flawless has no bearing on the kinds of demands Agnes is under. Since she isn't required to wiggle her finger in Flawless, she couldn't be required to wiggle her finger in Flawed. By contrast, the demoniac who believes that some things are justified, some things aren't justified, and that the difference between them has to do with conforming to norms might say that real norms focus on things like consistency or require subjects to respond to how things seem to them, not to external facts that could be obscure to them. Our example shows that this is a difficult position to defend, particularly in light of our intuitions. Consider (b) and consider Flawed. Let's suppose that Agnes wiggles her ears and twiddles her thumbs. Because of the fact-checker's testimony, it seems quite natural to think that Agnes could rationally believe both that the sentences in her book expressed an inconsistent set of propositions and that it seemed to her that it contained an error. Our intuitive reaction to her wiggling, I take it, is sympathetic in just the way our reaction to Cohen's bedeviled subjects is.27 Now consider Flawless. Suppose that in Flawless Agnes wiggles her ears and twiddles her thumbs. We would, I think, respond sympathetically to this performance, too. According to the Recognition Thesis, this sympathetic response is really just the recognition that Agnes met the demands imposed upon her by the relevant norms. I'll go out on a limb here and conjecture that you do not recognize that she conformed to the relevant norms. Consider the preface sentence in Flawless. This sentence says, 'There is an error contained in this book'. If you thought that she conformed to the requirement represented by (b), you would have to recognize that the sentence expressed a proposition that helped constitute an inconsistent set of propositions I don't think you recognize that at all. I don't think you know that it expressed such a proposition and thus required her to wiggle her ears. For one thing, if you knew it, it would be true. If, however, propositions are things that must either be true or false (and things that can never be both), it's not true that the sentence expresses a proposition. For another, even if this view of propositions was incorrect, I very much doubt that you know what to say about the liar and if you don't know what to say about the liar, you really cannot say whether or 27 Some people object to the fact that these examples involve actions, not beliefs. We can easily modify the case so that the focus is not on wiggles and twiddling and instead on Agnes' beliefs about whether she should wiggle or twiddle. The same points would apply. In Flawed and Flawless, Agnes would believe she should wiggle and twiddle but we cannot say that there were seemings and beliefs that would justify such wiggles and twiddles but we must respond sympathetically if Agnes believes she ought to wiggle and twiddle. Thus, the sympathetic response is not the recognition that the relevant attitudes are justified. not Agnes met the demand imposed by (b). You should be sympathetic when you think about Agnes' wiggling in Flawless, but you shouldn't think that this sympathetic response is the recognition that she met the demand imposed by (b). That is an open question I very much doubt you'll settle in your lifetime. Consider (c) and consider Flawed and Flawless. You should be equally sympathetic when you think about Agnes twiddling her thumbs in these cases. If the sympathetic response is the recognition that she meets the demand imposed by (c) in both cases, you have to think that it seems to Agnes in both cases that there is an error in her book. You do not think this. If seemings (sic) are propositional attitudes that can be assessed for accuracy or veridicality in the way that beliefs can, Agnes does not have the same seemings in Flawed and Flawless. It cannot be that it seems to Agnes in Flawless that her book contains an error. If it did, that seeming would either be accurate or inaccurate. If it were accurate, however, it would be inaccurate. If it were inaccurate, however, it would be accurate. You should be sympathetic when you think about Agnes twiddling her thumbs in these cases, but you could not possibly think of this sympathetic response as the recognition that she met the demands that she was under unless you think that you can identify what those demands are. You cannot hope to do that if you cannot solve the liar paradox and I very much doubt that you'll ever know how to solve that paradox. I also very much doubt that you'd think that we have to get to grips with that paradox to know whether to be sympathetic to Agnes. Her response to the situation is perfectly rational. We know that. Justified? We'll never know because we'll never know if she conformed to the demand imposed by (c). Let me wrap this up with two remarks about the Restricted Infallibility Thesis. First, the intuitions that seem to motivate the Restricted Infallibility Thesis are demonstrably unreliable. They force us to try to find conditions to plug into our norms that satisfy the Demonic Doxastic Demandingness Thesis. Unfortunately, if we rely on our sympathetic responses to properly chosen pairs of cases we'll see that we'll respond sympathetically to subjects that do violate norms or respond sympathetically when we know that we cannot know whether they conformed to the norm or not. Thus, we cannot see our sympathetic responses as the recognition that these subjects conformed to the relevant norms. Our sympathetic responses to subjects who reason and respond in ways similar to subjects in good cases are an unreliable guide for determining whether some subject did in fact conform to a norm or act or believe with justification. Second, the Restricted Infallibility Thesis places an implausible constraint on the formulation of norms. There are few, if any, possible conditions that satisfy the thesis because it's possible to have strong but misleading evidence for believing just about anything. 28 There might well be good reasons to reject externalist accounts of epistemic norms, but the demoniacs haven't identified those reasons. 4. The Newest Evil Demon Problem I’ve tried to show that none of the considerations Cohen offers give us any good reason to reject externalist accounts of justification. My aim has been to show that externalists can draw distinctions that we have good independent reason to draw and that once we draw these distinctions externalists can accommodate some of the intuitions that he appeals to and can rightly reject some of the important assumptions that feed into his attack on externalism. In this concluding section, I shall gesture at a problem for internalist views motivated by Cohen’s argument. It's important to remember that while there are intuitions that seem to cause trouble for externalists, there are also considerations that point in a different direction: The Newest Evil Demon Argument P1. In the bad case, the subject's beliefs don't have the properties that figure in externalist accounts of justification [Assume]. P2. The beliefs are equally justified in the good case and the bad case [Equality]. 28 It is interesting that Cohen resisted this point in conversation. It seems that he's insisting that for some conditions it's not possible to rationally but falsely believe these conditions to obtain on the basis of strong but misleading evidence. It also seems that his case against Williamson rests entirely on the idea that it's always rational to believe things on the basis of strong evidence even when the evidence is misleading. P3. For some beliefs, your beliefs wouldn't be justified unless you were in a better epistemic position than a mental duplicate of yours in a bad case [External Constraint].29 C. The relevant class of beliefs couldn't be justified however strong your epistemic position might appear. This objection is directed against every non-skeptical internalist view supported by the New Evil Demon Argument. To avoid the skeptical conclusion, we'll either have to reject Equality or External Constraint. Since Equality is the crucial assumption in the new evil demon argument, the demoniacs have to reject External Constraint. I've argued for External Constraint elsewhere on intuitive and theoretical grounds.30 If External Constraint is correct, externalism is the only coherent alternative to skepticism. If External Constraint is correct, there's no sound argument from Sufficiency and Equality to some non-skeptical internalist view because Sufficiency and Equality lead to skepticism. To get a feel for how the Newest Evil Demon Argument is supposed to work, consider the role that beliefs play in rationalizing feelings, beliefs, and actions. Consider this quartet of claims: 1. Enkrasia: Rationality requires you to be enkratic (i.e., to avoid failing to φ when believing that you're required to φ). 2. The Identification Thesis: S's φ-ing was justified iff S's conscientious duplicates' φ-ing was rational. 3. Conformity: Justification requires norm conformity.31 29 For a similar argument against internalism, see Srinivasan (2015). See Littlejohn (2012) for a transcendental argument from the possibility of categorical norms to externalism about the justification of belief. 31 Huemer (this volume) challenges this, in a way. He thinks that 'ought' and 'should' can be read objectively or subjectively and that it's only on the subjective reading that justification connects to norm conformity. In turn, he suggests that the internalist and externalist might be engaged in a shallow semantic dispute. This is possible, but I think it really depends upon the details of the internalist view in question. In my (2011, 2012), I argued that the kind of internalist view that Huemer's (2006, this volume) arguments support is a kind of internalist view that doesn't leave room for categorical reasons and 30 4. Sufficient Evidential Support Thesis: It is rational for S to believe p if S's evidence for p provides sufficiently strong support. Once you have these assumptions in place, consider cases in which you have strong evidence for the following propositions: (a) I ought to be angry with Bob. (b) Bob threatened someone I love. (c) I ought to punch Bob in the nose. (d) I ought to believe that Bob threatened someone I love. In having strong evidence for such propositions, (4) tells us that you can rationally believe them. From this and (2), it follows that you can justifiably believe them. If you justifiably believe that you ought to angry with Bob but aren’t angry with him or you justifiably believe that Bob threatened someone you love, it wouldn’t be rational for you to maintain this belief and fail to be angry with him. If so, it seems that it would be rational for you to be angry with him. If it is, this anger is justified. If it is, there’s no norm that governs your emotional responses that you’d violate by being angry with him. And yet, there could be, couldn’t there? Isn’t this a plausible norm for anger? Fittingness: You shouldn’t be angry with someone unless that anger is fitting. The anger wouldn’t be fitting if, say, your evidence is misleading. Maybe it wasn’t Bob who threatened someone you loved. If it wasn’t Bob, the anger wouldn’t be fitting. Since Fittingness seems like a pretty good norm, I’d think that there’s a perfectly good sense in which you shouldn’t be angry with Bob. If so, (3) tells us that the anger isn’t justified. At best, it’s excused. If so, (2) tells us that it’s not rational. Now, this is where I’d get off, but Cohen assures us that the distinction between the anger it would be rational to feel and couldn't vindicate certain moral intuitions (e.g., about reparative duties we have in the wake of wrongdoing). Since I don't think debates about whether we owe reparative duties to, say, the people who have been wrongly incarcerated for crimes they didn't commit aren't shallow semantic disputes, it's hard for me to see how disputes about Huemer's internalist view (a view that I've argued undermines a variety of things that we'd want to say about norms generally and reparative duties in particular) could amount to nothing but shallow semantic disputes. justified anger doesn’t make sense. If it’s not rational, what is it then? Are you supposed to be Enkratic? Violating Enkrasia is the clearest possible case of irrationality! If the Identification Thesis is right and you’re rationally required to be Enkratic, you cannot justifiably have the attitudes that make you akratic. So, are we supposed to say that the arbitrarily strong evidence that supports the belief that Bob threatened someone you love doesn’t make your belief rational? I don’t see a good way forward here apart from (a) admitting that rationality and justification fracture in just the way I’ve suggested and (b) admitting that the justificatory status of your beliefs depends upon whether they’d justify the affective responses they rationalize. I don’t see any good way to make sense of (b) without embracing a factive account of justification, so that’s how I’d sort this mess out. We get similar problems if we think about (c) and (d). The evidence, (1) tells us, makes it rational to believe (c) and (d). This, in turn, tells us that these beliefs are justified. This, in turn, tells us that no norm is violated in forming such beliefs. This, in turn, tells us that we ought to believe that Bob threatened someone I love. This in turn tells us that you ought to punch Bob. Again, the problem here is that the evidence can be misleading. Because it’s misleading, we don’t want to detach the conclusion that you ought to hit someone and we don’t want to detach the conclusion that you ought to believe that Bob threatened someone you love. To my mind, the obvious thing to do is distinguish between rationality and justification and let justification depend upon the justificatory status of the things that the belief rationalizes.32 32 Interestingly, Cohen’s response is to insist that the Identification Thesis is correct and simply say that it’s not rational to do things like punch someone you rationally believe threatened someone you love. (Personal communication. I don't know if he still endorses this response.) His example is slightly different, but he thinks that the rationality of action depends upon facts that you couldn’t know. (He thinks that it is irrational not to shoot a terrorist disguised to look like an electrician no matter what your evidence is. From personal communication.) Since he's an internalist about the rationality/justification of belief and externalist about the rationality/justification of action, he rejects the Enkratic Requirement. I see that as a significant cost. Feldman (1988) once embraced this combination of views, but he has since decided that it is best to be an internalist about both kinds of justification. If he thinks that we are justified in acting on beliefs about obligation that are supported by the evidence, this resulting view leads to troubling 6. Conclusion Externalists should agree with the demoniacs that it's appropriate to feel sympathy for the bedeviled, but we should now see why it's a mistake to take that sympathetic response to be some sort of indication that the bedeviled subjects' attitudes are justified. They are excused. They are excused because their beliefs result from the excellent use of the subjects' rational capacities but fall short of being justified because they don't (always) meet the standards or norms we use to assess belief. The arguments just offered show that it's a mistake to try to identify norms that can be demon-proofed or to think of justification as something that's common to the best cases and the cases of systematic deception. References Alvarez, Maria. 2010. Kinds of Reasons. Oxford University Press. Arpaly, Nomy. 2002. Unprincipled Virtue. Oxford University Press. Audi, Robert. 2001. An Internalist Theory of Normative Grounds. Philosophical Topics 29: 31–45. Austin, J. L. 1956. A Plea for Excuses. Proceedings of the Aristotelian Society 57: 1–30. Bach, Kent. 1985. A Rationale for Reliabilism. The Monist 68: 246–63. Bennett, Jonathan. 1995. The Act Itself. Oxford University Press. Bird, Alexander. 2007. Justified Judging. Philosophy and Phenomenological Research 74: 81– 110. BonJour, Laurence. 2002. Internalism and Externalism. In P. Moser (ed.), The Oxford Handbook of Epistemology. Oxford University Press, pp. 234–64. Boult, Cameron. Forthcoming. Excusing Prospective Agents. Logos & Episteme. Broome, John. 2013. Rationality Through Reasoning. Wiley. Cohen, Stewart. 1984. Justification and Truth. Philosophical Studies 46: 279-95. consequences about the justification of action and moral culpability. For a discussion, see Littlejohn (2014). Cohen, Stewart. MS. Theorizing about the Epistemic. Cohen, Stewart and Juan Comesana. 2013. Williamson on Gettier Cases in Logic. Inquiry 56: 15-29. Cohen, Stewart and Juan Comesana. Forthcoming. Being Rational and Being Right. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Comesana, Juan. 2016. Falsehood and Entailment. Philosophical Perspectives 29: 82-94. Darley, John and Paul Robinson. 1998. Testing Competing Theories of Justification. North Carolina Law Review 76: 1095-1143. Douven, Igor. 2006. Assertion, Knowledge, and Rational Credibility. Philosophical Review 115: 449–85. Duff, R.A. 2007. Excuses Moral and Legal: A comment on Marcia Baron's 'Excuses, Excuses'. Criminal Law and Philosophy 1: 49-55. Dutant, Julien. Forthcoming. Knowledge-based Decision Theory and the New Evil Demon Problem. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Engel, Mylan. 1992. Personal and Doxastic Justification. Philosophical Studies 67: 133–51. Fantl, Jeremy and Matt McGrath. 2009. Knowledge in an Uncertain World. Oxford University Press. Feldman, Richard. 1988. Subjective and Objective Justification in Ethics and Epistemology. The Monist 71: 405-19. Ferzan, Kimberly. 2011. Justifications and Excuses. In J. Deigh and D. Dolinko (ed.), The Oxford Handbook of the Philosophy of Criminal Law. Oxford University Press. Gardner, John. 2007. Offenses and Defenses. Oxford University Press. Gerken, Mikkel. 2011. Warrant and Action. Synthese 178: 529-47. Gibbons, John. 2010. Things That Make Things Reasonable. Philosophy and Phenomenological Research 81: 335–61. Hawthorne, John. MS. The Surprise Examination Paradox. Herman, Barbara. 1993. The Practice of Moral Judgment. Harvard University Press. Horder, Jeremy. 2004. Excusing Crime. Oxford University Press. Hornsby, Jennifer. 2007. Knowledge in Action. In A. Leist (ed.), Action in Context. De Gruyter, pp. 285–302. Hornsby, Jennifer. 2008. A Disjunctive Theory of Acting for Reasons. In F. Macpherson and A. Haddock (ed.), Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Huemer, Michael. 2006. Phenomenal Conservatism and the Internalist Intuition. American Philosophical Quarterly 43: 147–58. Huemer, Michael. Forthcoming. Defending a Brain in a Vat. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Hyman, John. 1999. How Knowledge Works. Philosophical Quarterly 49: 433-51. Hyman, John. 2006. Knowledge and Evidence. Mind 115: 891-916. Ichikawa, Jonathan Jenkins. 2014. Justification is Potential Knowledge. Canadian Journal of Philosophy 44: 184-206. Kelp, Christopher and Mona Simion. Forthcoming. Criticism and Blame in Action and Assertion. Journal of Philosophy. Lackey, Jennifer. 2007. Norms of Assertion. Nous 41: 594-626. Lasonen-Aarnio, Maria. Victims of Deceit and the Virtue of Reasonableness. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Littlejohn, Clayton. 2011. Defeating Phenomenal Conservatism. Analytic Philosophy 52: 35-48. Littlejohn, Clayton. 2012. Justification and the Truth-Connection. Cambridge University Press. Littlejohn, Clayton. 2014. The Unity of Reason. In C. Littlejohn and J. Turri (ed.), Epistemic Norms. Oxford University Press. Littlejohn, Clayton. Forthcoming. Stop Making Sense? On a Puzzle about Rationality. Philosophy and Phenomenological Research. Locke, Dustin. 2015. Practical Certainty. Philosophy and Phenomenological Research 90: 72-95. Lord, Errol. Forthcoming. Defeating the Externalist's Demons. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Lowy, Catherine. 1978. Gettier’s Notion of Justification. Mind 87: 105-8. McKinnon, Rachel. 2013. The Supportive Reasons Norm of Assertion. American Philosophical Quarterly 50: 121-35. McKinnon, Rachel. 2015. The Norms of Assertion: Truth, Lies, and Warrant. Palgrave. Madison, Brent. 2014. Epistemological Disjunctivism and the New Evil Demon. Acta Analytica 29: 61-70. Madison, Brent. Forthcoming. Epistemic Value and the New Evil Demon. Pacific Philosophical Quarterly. Miracchi, Lisa. Forthcoming. Competent Perspectives and the New Evil Demon Problem. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press. Mitova, Veli. Forthcoming. Truthy Psychologism about Evidence. Philosophical Studies. Moore, Michael. 1997. Placing Blame. Oxford University Press. Naylor, Margery Bedford. 1988. Epistemic Justification. American Philosophical Quarterly 25: 49-58. Prior, Arthur. 1958. Epimenides and the Cretan. Journal of Symbolic Logic 23: 261-66. Pryor, James. 2001. Highlights of Recent Epistemology. British Journal of Philosophy of Science 52: 95-124. Rizzieri, Aaron. 2011. Evidence Does Not Equal Knowledge. Philosophical Studies 153: 235–42. Rosen, Gideon. 2014. Culpability and Duress: A Case Study. Proceedings of the Aristotelian Society, Supplementary Volume 88: 69-90. Russell, Bruce. 2001. Epistemic and Moral Duty. In M. Steup (ed.), Knowledge, Truth, and Duty: Essays on Epistemic Justification. Oxford University Press, pp. 34-48. Smithies, Declan. 2012. Moore's Paradox and the Accessibility of Justification. Philosophy and Phenomenological Research 85: 273-300. Srinivasan, Amia. 2015. Normativity without Cartesian Privilege. Philosophical Issues 25: 273-99. Steup, Matthias. 1988. The Deontic Conception of Epistemic Justification. Philosophical Studies 53: 65-84. Strawson, P. F. 1962. Freedom and Resentment. Proceedings of the British Academy 48: 1– 25. Sylvan, Kurt. MS. Rationality and Justification: Reasons to Divorce? Tadros, Victor. 2005. Criminal Responsibility. Oxford University Press. Titelbaum, Michael. 2015a. How to Derive a Narrow-Scope Requirement from WideScope Requirements. Philosophical Studies 172: 535–542. Titelbaum, Michael. 2015b. Rationality’s Fixed Point (Or: In Defense of Right Reason). In T. Szabó Gendler and J. Hawthorne (ed.), Oxford Studies in Epistemology 5. Oxford University Press, pp. 253–294. Turri, John. 2009. The Ontology of Epistemic Reasons. Noûs 43: 490–512. Turri, John. 2013. Bi-Level Virtue Epistemology. In J. Turri (ed.), Virtuous Thoughts: The Philosophy of Ernest Sosa. Springer. Unger, Peter. 1975. Ignorance. Oxford University Press. Wallace, R.J. 1998. Responsibility and the Moral Sentiments. Harvard University Press. Wedgwood, Ralph. 2002. Internalism Explained. Philosophy and Phenomenological Research 65: 349-69. Williamson, Timothy. 2000. Knowledge and its Limits. Oxford University Press. Williamson, Timothy. Forthcoming. Justifications, Excuses, and Skeptical Scenarios. In F. Dorsch and J. Dutant (ed.), The New Evil Demon. Oxford University Press.