How is wishful seeing like wishful thinking? Susanna Siegel * draft for Rutgers Epistemology Conference

REC readers: If this paper is too long for you, you can shorten the journey in any of these ways: - skip sections 3-5 on responsive dispositions (pp. 9-15) - skip Part II (pp. 15-29) - skip all Part III (pp. 29-end)

A Simple Argument It is a commonplace that beliefs can be epistemically appropriate or epistemically inappropriate. Suppose Frank wishfully believes that today, all his best features are evident even to the most casual observation. Passers-by who greet him with neutral expressions are holding back their appreciation of him, because they’re embarrassed to show it. Anyone who seems displeased with him is feeling inadequate in the face of his many assets. These interpretations make sense to Frank given his original wishful belief, but they are unreasonable, and that belief is the source of their unreasonableness.

Can a perceptual experience ever be a source of unreasonableness, because of its dependence on a desire? Could wishful seeing compromise the rational support offered by experience, just as wishful thinking compromises the rational support offered by Frank’s wishful belief? Here is a simple argument that it could.

Premise: Beliefs can be ill-founded by wishful thinking. Premise: Wishful seeing is possible. Conclusion. Experiences can be ill-founded by wishful seeing.

What kinds of considerations could possibly support such a conclusion? The argument’s conceit is that wishful seeing and wishful thinking are sufficiently analogous that they both have ill-founding effects. There are many places to probe the argument, starting with whether the idea of wishful seeing makes sense to begin with. Even if it does make sense, one might strongly suspect that equivocation on “ill-foundedness” between the premise and conclusion is unavoidable because beliefs are ill-founded if they are formed or maintained epistemically inappropriately, but experiences are not governed by epistemic norms. And even if equivocation on “ill-founded” can be avoided, nothing in the argument explains why the fact that wishful thinking can ill-found beliefs should tell us anything about what wishful seeing can do to the epistemic status of experiences. This is a major gap in the argument. Perhaps wishful seeing and wishful thinking are psychologically similar phenomena, but their epistemic effects differ: wishful seeing does not compromise the rational powers of experience at all.

But if wishful seeing does compromise the rational powers of experience, then a major consequence follows. Experiences are not uniformly receptive. They are not a

1 landing pad for information (or misinformation) that tumbles in along a-rational channels, naively open to the objects, properties, and events that are there for us to perceive. Experiences can fail to be receptive in these ways, not because they are off the grid of rational assessment, as an undirected ‘raw feel’ would be, but because the grid has a place for experiences that arise in some of the same ways as irrational beliefs do. And that opens the possibility that the notion of ill-foundedness that we use to convict some beliefs might convict some experiences as well.

Let me lay some cards on the table, though they are probably showing already. The Simple Argument leaves much to be explained, but it helps us see what additional factors would be needed to clarify the premises and help them support the conclusion. I think those factors can be provided. My aim here is to provide them, by clarifying and defending the premises and giving them the help they need to produce the conclusion.

Wishful seeing is clearly coherent. Just imagine that Jill wants Jack to be happy, and as a result, she is subject to an illusion when she looks at him: he looks happy when she sees him – even though you would see his neutral expression for what it is, as would most anyone else who wasn’t subject to any similar distortion. Or imagine that as a result of wanting him to be happy, Jill fails to see his furrowed brow and slumped shoulders, focusing instead on his bouncy step and unfrowning mouth. For all she sees, it seems reasonable to conclude that he is happy. In both cases, Jill ends up with a visual experience of Jack as happy, with her desire for him to be happy playing a key role in producing the experience.

What about ill-founding: is the Simple Argument bound to equivocate in applying this notion to both beliefs and experiences? What would it be for an experience to be formed or maintained epistemically inappropriately? At a minimum, factors that explain why the subject has the experience would prevent it from providing any rational support for believing (at least some of) its contents, or from providing as much rational support as an experience could otherwise provide. Ill-founded beliefs have this feature as well: the rational support (propositional justification) they can offer to subsequent beliefs is reduced or eliminated. Beyond this shared core feature, the only leftover thing missing from ill-foundedness of beliefs is their being epistemically inappropriate or doxastically unjustified.1 It seems a potentially merely verbal issue whether we extend this language to experience or not. It’s not in dispute that experiences can have the core feature of ill-foundedness.

1 I’ll use “doxastically unjustified” to mean, formed or maintained epistemically inappropriately. On this usage it is a substantive question whether a belief’s being formed and maintained in proper response to one’s evidence suffices to make it doxastically justified.

2 On behalf of the Simple Argument, the charge of equivocation could be answered by insisting at the outset that there’s no bar to considering experiences epistemically inappropriate or doxastically unjustified. That answer is a bit jarring. Two less jarring answers are available. First, the equivocation on “ill-foundedness” is avoidable, by scaling back to the common status that ill-founded beliefs can share with experiences: an experience or belief provides less rational support for subsequent beliefs, as a result of ill-founding factors, than it could without those factors. Second, the equivocation is harmless, because it doesn’t matter whether wishful seeing produces in experiences the “extra” aspect of ill-foundedness that uncontroversially applies to beliefs – namely, epistemic inappropriateness.2

In any case, any residual discomfort produced by applying even the scaled-back notion of ill-foundedness to experiences could be avoided, by reformulating the conclusion of the Simple Argument in terms of belief. A belief that P is formed by endorsing an experience, if the belief is formed on the basis of an experience that includes P in its content. The conclusion could then be reformulated like this:

Conclusion*. Beliefs formed by endorsing experiences are ill-founded when those experiences are ill-founded by wishful seeing.

The two conclusions are not equivalent, because the reformulated one limits the beliefs that can be ill-founded in the relevant way to beliefs formed by endorsing experiences. In contrast, the original conclusion allows that any beliefs formed primarily on the basis of experience to be ill-founded, if the experience is itself ill-founded by wishful seeing.3 Although the major gap in the Simple Argument is a defense of the idea that experiences and beliefs are symmetrically vulnerable to ill-founding by desires, trying to close the gap might seem premature, so long as a different pre-emptive worry is unaddressed. What if in wishful seeing, wishful experiences depend on desires in ways that are completely different from the ways in which beliefs depend on desires, when desires ill-found beliefs? If desires ill-found beliefs by controlling them in ways that can have no analog for experiences, then the basic reasoning behind the Simple Argument can’t get off the ground. In contrast, if beliefs and experiences can depend on desires in analogous ways, then we can go on to ask whether those analogous modes of dependence have similar epistemic effects on beliefs and experiences. If desires can in

2 It would be useful to have a word that isolates the core feature of ill-foundedness. I call experiences with this feature “epistemically downgraded” in Siegel (2013). 3 Both the original conclusion and the reformulated ones fail to specify that the ill- foundedness of an experience (its downgrade) may be specific only to some contents, such as those specifying shape but not those specifying color. We’ll see examples later. Self-ascriptions of ill-founded experiences may not themselves be ill-founded, even if beliefs formed by endorsing the experiences are, when the contents of the belief are the contents relative to which the experience is downgraded.

3 principle influence experience in the same way as it influences and ill-founds beliefs, then the Analogy thesis is true:

Analogy thesis: It is possible in principle for an experience to depend on a desire, in ways that are structurally analogous to modes in which a belief that P depends on a desire, where that mode of dependence makes the belief ill-founded.

Is the Analogy thesis true? If ill-founded wishful thinking were simply a matter of a belief causally depending on desire, then the question would reduce to whether experiences can causally depend on desires. But many beliefs that are not ill-founded causally depend on a desire (even a desire with the same content as the belief). You might hope that a hypothesis P is true, and for that reason be motivated to investigate whether it is true, and then end up believing P on the basis of the strong evidence that you find for it once you look into the matter. Nor does ill-founding come from mere counterfactual dependence. You might start out believing P on the basis of a hope, then lose the hope, but come to base your belief properly on evidence for P. At the later time, your belief counterfactually depends on your desire, but it is not then ill-founded by the desire.4

Analogous points hold for experiences. You might want to see Jack as he crosses the finish line, and so you position yourself accordingly. When you see him finish the race, your visual experience causally depends on your desire, but if wishful seeing is ill- founded, your visual experience is a poor candidate for wishful seeing. Similarly, an experience of Jack approaching the finish line could start out as a desire-induced hallucination, and morph into a veridical experience that is sensitive to what Jack actually does; or it could start out sensitive to what he does, but morph into a total hallucination that is indifferent to Jack’s actual movements.

So experiences, like beliefs, can depend on desires, without the desire ill- founding them. But how do beliefs depend on desires, when the desires do ill-found them? Prima facie it is somewhat mysterious how desires (or preferences, needs, goals) could have any impact at all on the rational status of beliefs. The paradigmatic ill- founding relation is between beliefs and bodies of evidence. Responsiveness to evidence has no analog in the case of desire. A desire that P does not generally provide evidence bearing on P, and it does not seem psychologically possible in explicit reasoning to treat a desire as rationally supporting a belief, in the same way that we can clearly explicitly treat evidence as rationally supporting a belief, even if we are wrong

4 Conversely, you could start out believing P on the basis of evidence that supports P, but subsequently your desire that P might make you disposed to keep believing P, even if: you learn that all the evidence on which you based your belief in P is discredited, or you gain evidence against P. Here, wishful thinking is a mode of maintaining a belief in P, rather than a mode of initially forming it.

4 about the extent of that rational support. 5 So when beliefs are based on desires, there must be some other route from the desire to the belief.

I’m going to attempt to describe two ways in which beliefs can and often do depend on desires, and argue that experiences can in principle depend on desires in those same two ways. If those arguments succeed, then the analogy part of the Analogy Thesis is true twice over. I’ll call the first shared mode of dependence responsive or evidence-related, and the second one selective or inquiry-related. In the responsive mode of dependence, desires control which beliefs you form in response to evidence, and which experiences you have, in response to the distal stimuli you perceive. In the selective mode of dependence, desires select which evidence you have in the first place, and which distal stimuli you experience in the first place. Ultimately I’ll argue that the Analogy Thesis is true because:

• responsive wishful seeing is analogous to responsive wishful thinking, • selective wishful seeing is analogous to selective wishful thinking, and • both responsive and selective wishful thinking can be ill-founded.

But what are responsive and selective wishful seeing? Wishful seeing in the responsive mode is a form of cognitive penetration of experience by a desire. Here the desire influences how an object looks when you see (and attend) it. If experiences are cognitively penetrable, then the processes that generate perceptual experiences are not all modular. I’ve emphasized “experience” because usually the thesis that perceptual processes are modular is not formulated (either by proponents or opponents) in a way that applies specifically to perceptual experiences. Roughly, a process is modular, if it is triggered by a distinctive range of stimuli and unfolds in the same way, regardless of what the subject wants, fears, or believes about the environment.6 Although cognitive penetrability of experiences is incompatible with modularism about the processes that generate perceptual experience, it is compatible with modularism about perceptual processes that generate pre-conscious perceptual states.

Wishful seeing in the selective mode is a selection effect on experience, in which a desire selects objects and properties for you to experience, without necessarily influencing how the objects or properties look when you see them. Selection processes can also inhibit experiences, by anti-selecting objects or properties for experience. Unlike wishful seeing by cognitive penetration, wishful seeing by selection is compatible

5 But suppose P = I want to fly. A desire that P is then a (second-order) desire to want to fly. Perhaps the fact that I have that second-order desire provides evidence that I want to fly – in which case my desire that P would provide evidence that P. But when values for P are external-world propositions rather than propositions about one’s own desires, this won’t happen. 6 Fodor (1983).

5 with modularism about perceptual experience – and with standard modularism. Modularists about perceptual experience will deny that our experiences are ever cognitively penetrated, but this empirical controversy is irrelevant here. Our interest is in the coherence of the wishful seeing in either form – and in what its epistemic upshot actually is, if modularism about experience is wrong.

Identifying modes of dependence on desire won’t help shore up the Simple Argument, if those modes of dependence cannot be shared by both beliefs and experiences, or if they never ill-found beliefs. And one might have the impression that responsive and selective modes of dependence each fall short of providing what the Simple Argument needs. Responsive modes of dependence are good candidates for ill- founding beliefs, but it might seem obscure how experiences could depend on desires in anything like the same way. They seem apt for the ill-founding part of the Analogy Thesis, but not for the analogy part. Selective modes of dependence seem to have the opposite shortcoming. It is clear that desires can select which things we experience and what evidence we have, but the fact that a desire guides how evidence is selected may seem quite a poor candidate for ill-founding beliefs. They seem apt for the analogy part of the Analogy Thesis, but not for the ill-founding part. If these impressions survive, then the prospects for the Analogy Thesis are dim. To dispel them, my discussion of responsive modes focuses mainly on their application to experiences, and my discussion of selective modes focuses mainly on their potential to ill-found beliefs.

Although selection effects and cognitive penetration are psychologically different, my defense of the Simple Argument entails that epistemically, they are similar. Cognitive penetrability per se has no special epistemological significance: an important category in the psychology of perception is less important in the epistemology of perception. That outlook leaves me with two main tasks: defend the Analogy Thesis for both modes of dependence (Parts I and II), and defend the epistemological symmetries between wishful seeing and wishful thinking (Part III).

Part I. Responsive modes of dependence on desire

We can tighten our grasp of the Analogy Thesis by identifying the type of wishful thinking that’s operative in the analogies between wishful seeing and wishful thinking.

1. The core phenomenon of wishful thinking Wishful thinking appears to be a diverse phenomenon, given the range of cases uncovered by experimental psychology in which needs, goals, and preferences influence beliefs, and given the various mechanisms by which the influence occurs. (I’ll lump together needs, goals, preferences, and hopes as “desires”). What a person believes about the causes of illness can be influenced by desire for circumstances to be under our control, rather than left to fate; how one answers an important question can be influenced by their desire for a simple or concise answer – regardless of whether the

6 answer is well-supported by what else they know; beliefs already formed can be maintained in the face of counterevidence by a desire not to be wrong.7 The category of wishful thinking (it is sometimes called “motivated cognition” or “motivated reasoning”) sprawls even farther if we factor in the ways that desires could in principle influence belief, even if no experimental results speak to whether it does. Perhaps the ways are too obvious to warrant experiments, or too hard to operationalize. But through the sprawl, we can identify an overarching phenomenon: a subject forms, tries to form, or maintains some of her beliefs by manipulating her own evidential situation, or responding to the evidence she has in ways that facilitate her ending up believing P, when: she wants P to be true, or when P’s being true would satisfy some other desire the subject has, or when believing P would satisfy some other desire she has.

Given the ultimate goal of filling in the Simple Argument, we can focus on cases of wishful thinking in which a belief depends on a desire in a way that makes the belief ill-founded. I’ll call this the core phenomenon of wishful thinking. You might think this point of focus barely narrows the category of wishful thinking at all, because any case of belief belonging to the overarching phenomenon is always ill-founded. Don’t the examples of motivated cognition smell that way? But the overarching phenomenon underspecifies the mode of dependence of belief on desire. It allows that the desire could be mediated by epistemically well-founding processes (where those are processes that result in epistemically inappropriate beliefs). And sometimes it is unobvious whether the mediating processes are well-founded or not. Suppose you unknowingly manipulate your evidential situation so as to screen out any evidence against P and against your grounds for P, but your belief remains sensitive to the evidence you have. Is your belief ill-founded, or not? Kelly (2008) argues that beliefs maintained in this way could be well-founded, whereas Kornblith (1983) suggests the opposite. A version of this controversy will arise in defending the Analogy Thesis for selective modes of dependence. By holding fixed that the core phenomenon of wishful thinking ill-founds the resulting beliefs, we can then ask which cases of the overarching phenomenon plausibly have this result.

The core phenomenon of wishful thinking differs from a neighboring phenomenon that I’m going to exclude from discussion here, even though it is discussed under the heading of motivated cognition.8 In the neighboring phenomenon, which I’ll call wishful entertaining, a desire that P obscures your evidential situation, interfering with memory and introspective access to the evidence you once had bearing on P, and

7 On beliefs about causes of illness, see Hirschberger 2006, Mandelbaum and Ripley 2012. On desire for concise answers, see Kruglanski and Webster 1996. Wanting not to be wrong will be discussed shortly. For an overview of experimental work on wishful thinking, see Molden 2012. 8 Moskowitz, Golwitzer, et al 1999, also Sinclair-Kunda 1999, Lieberman, Ochsner, Gilbert et al, 2001.

7 to your previous attitudes toward P. Suppose you have wanted your distant acquaintance X to win a prize. Someone asks you: Did X win the prize? --That X won sounds plausible, but you can’t recall whether it really happened, or you just wanted it to be true. You recall that Y nominated X for the prize, and that the influential Z said he thought X should win the prize, but not whether you heard subsequently that X really did win it. If you exit from this haze by judging that X won the prize, that would be a core case of wishful thinking. But the haze-inducing interference with memory and introspection per se isn’t the core phenomenon as defined here, though it may be a precursor to the core phenomenon.

In some cases of wishful thinking, the desire and the belief may have the same content. If you desire and believe P, then you want it to be the case that P and you think it is the case that P. Alternatively, the belief may concern a specific situation, while the desire is much less specific, such as the desire to control one’s circumstances, or a desire to reach a simple and concise conclusion. In other cases, a preference is implicit in the structure of processing incoming information that pertains to a prior belief. Suppose that negative affect attaches to taking in information that you recognize as counting against P, or as potentially counting against it, and that the negative affect motivates you to avoid processing that information further (you forget it immediately, or fail to recognize that it counts against P), or to discount it (you suspect it isn’t true, or doesn’t count very much against P).9 These responses to the incoming information could encode a preference to maintain the belief that P, or a preference not to be wrong about P. Those are ways of being invested in believing that P.

A desire to keep believing that P differs from desire that P. You might be invested in believing that winters where you live are especially severe, while wanting badly to move to Hawaii. You don’t want to live where the winters are severe. Desiring (e.g., preferring) to believe that P also differs both from desiring to: believe that P only if P, and from desiring to: believe that P only if P is well-supported by evidence, or a desire to: believe that P only if P is epistemically appropriate to believe. People who believed the world would end generated highly implausible explanations for its continuation, even though they presumably did not wish for the end of the world. They also, apparently, did not desire to: believe the world will end, only if that belief was evidentially well-supported, true, or epistemically appropriate.10

2. Desire-dependent dispositions and ill-founding How can we understand the mode of dependence of belief on desire in wishful thinking?

9 On discounting, see Simon et al 1995. 10 Festinger 1956, When Prophesy Fails.

8 Let’s say that relative to a belief that P, a desire is rationalizing, if it is either: a desire that P, a desire to believe P, a desire that Q, or a desire to believe Q, where the subject takes P to support Q, or to explain away apparent counterevidence to Q. Consider a desire to be in control of your health. If someone else gets sick, that desire might make you come to believe, or want to believe, that they got sick by making unwise lifestyle choices. You might review what you know about what they eat, how much they exercise, their exposure to toxins, etc, in hopes of finding something to explain their illness. Explanations you previously dismissed as superstitious might come to seem worthy of consideration. If P = the ill are partly to blame for their illness, rather than being unlucky, then P would explain away counterevidence to something you want to be true, or want to believe: that you can control your health by making wise lifestyle choices (= Q), and so are not vulnerable to bad luck in health. If the ill were just unlucky, then you could be unlucky too – and that is at odds with Q.

The notion of a rationalizing desire helps us characterize modes of dependence on belief in wishful thinking. Such modes of dependence can be seen as psychological dispositions that explain why the subject believes that P, where she has these dispositions, because she has a desire that rationalizes P. We can say that a subject’s disposition to believe that P is desire-dependent, if she has the disposition because she has a rationalizing desire.

This way of thinking about modes of dependence has its limits. Not every belief explained by desire-dependent disposition is thereby ill-founded. At least not obviously so. Suppose you believe P on good grounds (it doesn’t matter what P is). Let Q = you are not misled by misleading evidence against P. You don’t want to be misled, so you avoid encountering evidence bearing on P. Perhaps some versions of this holding pattern for your belief that P don’t ill-found it.

But if a desire-dependent disposition explains why a subject believes that P is ill- founding, it could be ill-founding in two ways. First, it could ill-found the belief, regardless of whether the disposition is desire-dependent. For instance, suppose a rationalizing desire for P makes you jump to the conclusion that P, when your evidence doesn’t support P. You have a desire-dependent disposition to believe P that ill-founds it, but the belief would be ill-founded by the disposition to jump to a conclusion that’s unsupported by your evidence, even if your rationalizing desire didn’t explain why you were disposed in that way. You could be disposed in that way because of a fear, or because you tend to reason badly. Alternatively, a disposition could ill-found a belief, at least partly in virtue of the fact that the subject has that disposition, because she has a rationalizing desire. We’ll see examples of both kinds along the way.

3. Evidentially bad responsive dispositions Let’s say that a subject has an evidentially bad responsive disposition, if she is disposed to believe P in response to evidence that doesn’t support P. Evidentially bad

9 responsive dispositions can operate through more specific dispositions, such as being disposed to:

o disregard defeaters for P, or for your evidence for P. o believe without justification that one has good reason to believe that P. o misinterpret evidence against P, or evidence that is neutral as regards P, as evidence for P. o interpret sources of evidence against P as illegitimate.

An evidentially bad responsive disposition to believe P is desire-dependent, if having a rationalizing desire explains why the subject is so disposed, or why she has specific dispositions that mediate them. For example, in the last two cases, absent the rationalizing desire (and modulo overdetermination), a subject might interpret evidence against P properly, instead of misinterpreting it as supporting P, or recognize strong evidence against P, or not question the legitimacy of a sources of evidence against P.11

Having a desire-dependent evidentially bad responsive disposition is a way of being resistant to evidence in the following sense:

Resistance to evidence Across a range of (actual or counterfactual) evidential situations in which a subject S considers bodies of evidence bearing on P that differ from S’s actual total evidence, where some of those bodies of total evidence do not include any evidence that supports P, S is disposed to believe P, because of her rationalizing desire.

The disposition described here allows that S has evidence for P, but disallows that the evidence controls her belief. It disallows that if she lost that evidence without gaining any additional evidence, she would lose the belief. When a desire-dependent disposition to resist evidence explains why S believes P, does it ill-found that belief? If it explains why S believes P, and S’s belief is not supported by the evidence she has, then (I’m assuming) the belief is ill-founded, independently of the fact that her disposition is desire-dependent. But suppose S believes P, and has excellent evidence for P (that she recognizes as such), and her desire-dependent disposition to believe P would be triggered, only if she had much weaker evidence. Could the desire-dependent disposition still explain why she believes P, if the disposition is not triggered?

11 On misinterpreting evidence that doesn’t favor P as evidence that does, see Stillwell and Baumeister 1997. On taking sources of counterevidence to be illegitimate, see Lord et al 1979, Allyn and Festinger 1961.

10 It seems useful to distinguish two senses in which S’s belief could be said to be based on the excellent evidence. On the thin, recognition-based notion, S bases her belief on the excellent evidence, by virtue of the fact that she recognizes that the evidence supports P. How she would respond to changes in that evidence doesn’t matter. On the robust, control-based notion, S bases her belief on the excellent evidence, to the extent that she would adjust her belief in response to changes in that evidence. Both notions of basing may be useful. What matters is that the epistemic stauts of S’s belief seems affected by the extent to which it is controlled by the excellent evidence that she thinly-bases her belief on.

To see how, we need to refine the notion of ill-foundedness. Just as justification can come in increments, ill-founding of beliefs can be incremental as well. There may also be a useful on-off notion of ill-founding, and for all I’ve said, there could be increments of ill-founding that don’t reach the threshold of outright ill-founding. Since the relationship between incremental and outright ill-founding is obscure, in asking whether the desire-dependent dispositions ill-found beliefs, let’s ask simply whether those dispositions make any incremental or pro tanto contribution to ill-founding.

Focusing on increments of ill-foundedness also gives us a way to express the intuitive idea that two people who believe P on the basis of the same excellent evidence for P could differ in how well-founded their beliefs are, because they differ in how sensitive to that evidence they are. Consider S and S*, who differ in how they would respond to weakenings of their currently excellent evidence that P. In response to any weakening of the evidence, short of overwhelming counterevidence, S’s rationalizing desire would kick in and make S maintain her belief that P at the same level that was appropriate for excellent evidence. In contrast, S*’s belief is controlled by the excellent evidence in a more robust way: if the evidence for P weakened a bit, S* would back off from her belief accordingly, but if it weakened a lot, S*’s desire-dependent disposition would kick in, and S*’s belief would stay where it was, short of overwhelming evidence against P, which would make S* simply give up her belief. S and S* both have a desire- dependent disposition to resist evidence, but their dispositions have different boundary conditions: only overwhelming counterevidence can put the breaks on S’s disposition to resist evidence, and so her disposition is stronger than S*’s, which operates across a smaller range of triggering conditions. Since S is more resistant to evidence than S*, intuitively that makes S’s beliefs more ill-founded than S*’s. In a third subject who believed P on the basis of the same excellent evidence, the belief would not be ill- founded at all, if it was controlled by evidence perfectly – in response to any change in the evidence, the subject would adjust her belief in an appropriate way. Even if one denies these differences in ill-founding (e.g., on the grounds that only the thin notion of basing can determine ill-foundedness), the Analogy Thesis could still hold between bad thin basing, and – as we’re about to see - cases of cognitive penetration where Jill’s desire actually distorts Jack’s expression.

11 So having a desire-dependent disposition to resist evidence can ill-found beliefs. The ill-founding part of the Analogy Thesis is true, for responsive modes of dependence. That brings us to the analogy part of the Analogy Thesis. Is there any mode of dependence on desire that’s analogous to resistance to evidence? Experiences are not formed in response to evidence in the way that beliefs are. But they are formed in response to distal perceptual stimuli – roughly, in the case of visual experience, the things we see. Given that we can see ordinary objects, events, places, the sky, etc., we can ask: how do things look when we see them? It seems that experiences can depend on rationalizing desires in a way that is structurally analogous to resistance to evidence that stems from rationalizing desires. This is what happens in wishful seeing by cognitive penetration. Wishful seeing by cognitive penetration lets us see why the analogy part of the Analogy Thesis is true.

4. Wishful seeing by cognitive penetration In wishful seeing by cognitive penetration, rationalizing desires effect how things look when you see them.

Resistance to distal stimuli Across a range of distal stimuli, where some of these stimuli are not F: if you perceive those stimuli, you are disposed you end up with an experience attributing F-ness to them, because of a desire that attributes F.

For example, because Jill wants Jack to be happy, across a range of ways for Jack to be when she sees him, (walking, sitting, grinning, wincing, etc), Jill is disposed to end up with an experience attributing happiness to him.

Feel free to skip to section 5 if the idea of resistance to distal stimuli is clear enough already. But more exactly, we can say that a desire that p attributes a property F, if p can be expressed by sentences of any of these forms: o is F, or something is F, or All F’s are G’s, or All G’s are F’s, G’s are F’s, F’s are G’s.12 For instance, if F = redness, then redness is attributed by a desire that object o is red, something is red, all red things are square, all squares are red, squares are red, red things are square. Similarly, we can say that an experience attributes F to something if its contents are true, only if a sentence of

12 In “black is the color my friend wears most often”, black is what a property is attributed to, not a property that is attributed to something else. But there could be cases of cognitive penetration involving such desires. Your friend only wears black and dark blue, but you experience every color she wears as black, because you want black to be the color she wears most often.

12 one of those forms is true.13 For example, suppose you hope that the shapes in front of you are square. This desire attributes squareness. (It also attributes the property of being a shape, but let’s focus on squareness). If your square-hope makes your square- experience resistant to stimuli, then we can say the desire distorts those stimuli. (More exactly, the desire would distort their shape, if you perceived them – modulo overdetermination, etc). Of course some stimuli in the range might really be square, making it sound odd to say that the desire distorts them. But the pattern of experiences that define the mode of dependence is meant to make explicit a way in which the F- experiences of any stimulus in range is controlled by the F-desire. Even a veridical F- experience could be controlled by an F-desire, and in that sense, even a veridical F- experiences could be resistant to F-stimuli, and even F-stimuli could be distorted by the desire.

5. Strength of resistance in experience and belief The analogy part of the Analogy Thesis is bolstered by the observation that resistance to distal stimuli varies in strength along three dimensions, all of which are also found in resistance to evidence. Specifically, these observations bolster the analogy part, as opposed to the part about ill-founding.

Let’s start with the dimensions of strength in the case of experience. We can ask: how wide is the range of distal stimuli, to which your F-experience is made resistant by the F-attributing desire? There are different respects in which the range might be more or less wide.

First, there is the sheer quantity of different types of stimuli. For instance, suppose the desire controls a subject’s F-experience across a small range of stimuli. Jack sees Jill wearing a neutral expression, wants her approval, and as a result, has an experience attributing happiness to Jill’s face. If her face were nearly-happy, he would still experience it as happy, but if it were clearly sad (or clearly angry), Jack would experience it as it is. In a facial morphing task, where a range of faces that differ slightly in their expression are lined up, with sad (or angry) Jill-faces at one end that change gradually and end up with happy Jill-faces at the other end. Approval-craving Jack places the line between sad and happy faces closer to the sad end, whereas on a different day, when he’s feeling more autonomous, he puts the line farther away from the sad end.14

Second, there is the relationship between the stimuli that get experienced as F (= faces experienced as happy), and F-ness (=happiness). This relationship captures the

13 What if experiences don’t have contents, but are instead relations to properties or complexes of properties? Then an experience attributes a property if the property is a relatum of the experience. Johnston 2004, 2006, Bengson (forthcoming). 14 The facial morphing task illustrates this dimension of strength, on the assumption that it operationalizes how the faces look to Jack, not just how he interprets them.

13 degree of exaggeration imposed by the desire. For instance, in the case of redness, there is an intuitive sense in which a desire making an orangish surface look red constitutes less of a distortion of the stimuli, compared to a desire making a greenish surface look red. In the case of squareness, a desire that x is square that led your experience it as square would distort a nearly-square rectangle that you perceived less than it would distort a circle. As we saw in the case of happiness, a desire that x is happy that led to your experiencing a face as happy would distort the stimuli less, if the face would normally be experienced as neutral, compared to if the face would normally be experienced as furious.15 Perhaps not every property that can be represented in experience will be associated with a similarity structure that lets us define this sub- dimension, but some properties clearly will.

A third dimension along which wishful seeing by cognitive penetration can be more or less extreme concerns what happens when you look longer or more carefully at a distorted stimulus. Suppose your desire to be facing a square makes a nearly-square rectangle look square (when normally it would not look square). What happens when you examine it more closely, trying to discern what shape it is? Will the effect of the desire persist, given subsequent sensory inquiry, or will subsequent inquiry make the experience more dependent and less resilient to the distal stimuli? The more resistant to further sensory inquiry it is, the longer the effect of the desire will persist.

If we think of distal stimuli that are examined longer or more carefully, beyond the point when they are initially experienced as F, then we can understand resistance to this type of sensory inquiry as a special case of the resistance to distal stimuli. Call long- gazed stimuli those stimuli that a subject continues to look at or examine more carefully, after she initially has an F-experience of those stimuli. And consider a pair of cases in which your desire to be facing squares makes you experience the shapes you face as square, across a range of circumstances, in some of which the shapes aren’t square. We can say that one of the experiences in the pair resists further sensory inquiry than the other, if the desire makes the first experience resist some long-gazed stimuli, but doesn’t make the second experience resistant to any long-gazed stimuli.16

15 The face example illustrates the independence of the exaggeration-dimension from the quantity-dimension. A desire could distort only small range of stimuli, yet still yield exaggerated misperceptions. For instance, in principle, a desire to see a face as happy might distort an oddball pattern of stimuli, such as producing happy-face experiences of very angry faces and neutral faces, but not of nearly-angry or nearly-sad faces. 16 I’ve focused so far on cases in which a desire that P penetrates an experience with content P, and so the contents of both states attribute the same properties. But the same type of resistance can be found in cases where the contents differ. For instance, suppose you want to be near a gun, and you believe that there are lots of shiny guns around. If you see something vaguely gun-shaped, your desire to be near a gun will make you experience the thing as shiny, and from there you’ll conclude that it is a gun.

14

How do these observations bolster the Analogy Thesis? All three dimensions of strength of resistance to distal stimuli have analogs for resistance to evidence. Just as resistant to non-F stimuli is more extreme, when the subject is disposed to have F- experiences in response to more non-F stimuli, so too resistance to evidence is more extreme, when the subject is disposed to believe that P, in response to more bodies of evidence she considers that don’t support P. Just as resistance to non-F stimuli is more extreme, when the subject disposed to have F-experiences in response to non-F stimuli that are not only not-F, but are in an intuitive sense far from being F, so too resistant to evidence is more extreme, when the subject is disposed to believe p, in response to bodies of evidence that an intuitive sense do not come close to supporting p.

Finally, just as resistance to non-F stimuli is more extreme, when the subject is disposed to keep having F-experiences after further sensory inquiry into non-F stimuli that she already experienced as F, prior to looking longer or more carefully at it, so too resistance to evidence is more extreme, when the subject is disposed to keep believing p, even after examining evidence she has on which she bases her belief, when it does not in fact support P, or doesn’t support P as much as she takes it to.

So it seems that in the responsive mode, wishful seeing by cognitive penetration and resistance to evidence are structurally analogous. The experience and beliefs cases depend on desire in similar ways. And they can be resistant in the same variety of ways, which mediate the influence of desires on beliefs and experiences. Together with the ill- founding potential of being resistant to evidence, we have the first case for the Analogy Thesis.

Part II. Selective modes of dependence on desire

The Analogy Thesis may be true of responsive dispositions, but most actual cases of the overarching phenomenon of wishful thinking operate through selective modes of dependence – with or without responsive modes. We can say that a subject has a selective (inquiry-related) disposition, if she is disposed to maintain a belief that P, by being disposed in any of these ways:

• to avoid encountering evidence against P.

Let’s say that your desire does not control the conclusions you draw from your experience. Those conclusions are sensitive to the properties presented in experience, together with your background beliefs. But your desire to be near a gun, together with your belief that the guns you are likely to see are shiny, jointly make your experience resistant to a range of illumination properties other than shininess. If you lost either the belief or the desire, then (ceteris paribus, and assuming no overdetermination, etc), your experience would be less resistant with respect to illumination.

15 • avoid encountering any evidence bearing on P. • to quickly forget prima-facie evidence against P, or against propositions that the subject takes to support P, while remembering well evidence for P. • to scrutinize new putative evidence for not-p, or putative undercutters for evidence for p, while taking new putative evidence for p at face value.17

I’m going to focus exclusively on a slightly more general selective disposition that includes the disposition to avoid evidence against P (for P that you believe). This is the disposition to avoid encountering information that is uncongenial to your belief that P. Relative to a belief or a desire that P, uncongenial information is evidence against P, evidence against your grounds for P, or evidence against what you take your grounds for P to be. In its desire-dependent form, you are disposed in this way, because believing P would satisfy a rationalizing desire that you have.

I’ve defined this disposition in terms of a belief that the subject already has. But the same disposition to avoid uncongenial information could also explain why someone initially forms a belief that P. Suppose you are looking at a crowded display of colored shapes. You hope (for some reason) that all the squares in the display are red, and that hope makes you disposed to overlook squares that aren’t red. You end up believing on the basis of perception that all the squares in the display are red. You’re disposed to overlook non-red squares, because you’re disposed to avoid encountering uncongenial information – and by experiencing non-red squares, you’d get information that’s uncongenial to your hope that P. Of course, it needs to be explained how this could happen. The next three sections address that question.

Selective modes of dependence in general are excellent candidates for the analogy part of the Analogy thesis, because both selection of evidence and selection of distal stimuli are inevitable. We usually can’t consider all possible evidence for what we believe, and of course never visually experience everything there is to see. In fact, defending this part of the Analogy thesis for this disposition seems simple. So far I’ve said exactly nothing about what evidence is. But on the plausible assumption that avoiding seeing distal stimuli can be a way of avoiding uncongenial information, it is trivially true that experiences and beliefs can depend on desire in the selective mode. The assumption does not imply that a bloody knife is evidence (in the relevant sense), or that you can have evidence that a raw steak was just sliced by picking up the bloody knife that sliced it. What’s important is that if you saw the bloody knife, you could have

17 On not scrutinizing whether new evidence supports P, see Lord et al 1979. On quickly forgetting putative evidence against P while remembering well evidence for P: Santiago, Kunda and Fond 2009. On not gathering further evidence bearing on P: Hart 2009. On scrutinizing new putative evidence against P while taking new putative evidence for P at face value: Hart 2009, Ditto et al 1998, Giner-Sorolla 1999, Sherman et al 2005. On avoiding likely sources of evidence against P: Hart 2009, Tagliacozzo 1981.

16 evidence that a raw steak was just sliced, by seeing that the knife is bloody.18 On this way of understanding distal stimuli as potential evidence, no special case for the analogy part of the Analogy thesis needs to be made, for the selective mode of dependence. Instead, what’s needed is a defense of the ill-founding part of the Analogy thesis. And for that, we need a better understanding of how selective modes of dependence can operate, together with some reasons to think that when they operate that way, they can ill-found beliefs.

6. Two ways to avoid encountering evidence against your beliefs A selective disposition to avoid uncongenial information could operate in several ways. To zero in on the one that’s relevant for defending the Analogy Thesis, we first need to characterize it more carefully.

Suppose you see someone in a café wearing a funny hat with too many colors. You don’t know her but at first sight, she annoys you. You don’t like her hat, and that makes you feel an urge to avoid her, and especially to avoid talking to her. As it happens, the hat-wearer is on the opposite side in a heated local dispute about whether to build a wind turbine next to City Hall. She is a well-informed proponent of building it, and disposed to make the case to anyone who will listen. You’re opposed on broadly aesthetic grounds, and haven’t heard any detailed case for why City Hall is the best location for a wind turbine. But you’re not completely closed-minded. The facts and considerations that the hat-wearer is disposed to discuss would move you if you heard them – even coming from her.

In avoiding the hat-wearer, you are in fact avoiding a source of evidence against something you believe (viz, that the City Hall wind turbine should not be built). And if articulate, outgoing, informed proponents of the City Hall wind turbine tend to wear funny hats, then in being disposed to avoid talking to people in funny hats, you are in one sense disposed to avoid uncongenial information (relative to your opposition to the wind turbine). But for all that, your aversion to the hat-wearer need not be psychologically linked to any aversion to encountering uncongenial information. You might just dislike talking to people when they’re wearing funny hats, and so avoid it. Analogously, you might be averse to the Fox News channel, because you don’t like their logo, without being averse to assimilating the information they broadcast, even though those broadcasts are incongruent with what you believe. In these cases, you’d have a desire-dependent selective disposition, for all I’ve said about them so far. But it wouldn’t be the kind of selective disposition that I want to focus on. To have the relevant kind of selective disposition, you need to be averse to encountering

18 Or by having an experience that represents that the knife is bloody (or that a metallic oblong object is covered in red liquid). I’m using “S sees that P” in the way that entails that S has a visual experience with the content P, not in a way that entails that S knows that P (or even believes that P).

17 uncongenial information, not just averse to a source that happens to be a source of uncongenial information.

Those two aversions could be psychologically linked in various ways. And they would be linked in a more realistic version of the hat case, such as one in which your aversion to talking to the hat-wearer mediates an aversion to assimilating and believing things she says. (Or you don’t like the Fox News logo because you dislike the contents of their broadcasts). At one extreme of explicitness, you would judge (or express your suspicion) that people who wear funny hats like hers tend to favor the wind turbine, or that the fact that the woman in the café is wearing that hat is a piece of evidence that she favors it. At another extreme, your aversions to the hat, to talking to the hat-wearer, and to believing things she says could be psychologically correlated by association, without your having any mediating disposition to judge that the hat-wearer probably favors the wind turbine.19

If being disposed to avoid uncongenial information explains why you have a belief, can the belief be ill-founded by that disposition? In the kind of case we’re after, the disposition itself does the ill-founding – rather than any awareness you have of being disposed to avoid information, or awareness that there is uncongenial information that you are avoiding. It is not especially surprising that selective modes of dependence can ill-found beliefs, when the ill-founding is mediated by awareness of the selection. Similarly, we wouldn’t need a special argument for the conclusion that a subject’s perceptual experience that all the squares in the crowded display of colored shapes are red can ill-found the corresponding belief (that all the squares in the display are red), by virtue of the subject’s endorsing her experience while knowing (or having reason to believe) that there are many shapes in the display she hasn’t seen, and lacking independent reason to think they’re not non-red squares. Here, too, the ill-founding isn’t coming directly from the desire-dependent disposition that explains why you have the belief or the experience in the way wishful Frank’s belief that his virtues are evident to all explains how he interprets all of his social interactions. Instead it is coming from the fact that you ignore a defeater – a defeater that may even be generated, in some cases, by exercising the selective disposition to avoid encountering potential evidence against what you believe.

19 If the association is symmetrical, then part of your current psychological makeup is that your aversion to talking to her goes with your aversion to assimilating and believing what she says, in the sense that if you lost either aversion, you’d lose the other – barring whatever psychological transformations would be needed to alter the association. In an asymmetrical case, one aversion is mediating the other, but not vice-versa. For instance if instead the hat-aversion is asymmetrically mediating the aversion to believing her, then coming to like the hat would make you more disposed to assimilate and believe what she says, whereas being won over to the pro-wind-turbine side by hearing its virtues explained by someone hatless would leave your distaste for the hat intact.

18

The trouble is that some ways of being averse to uncongenial information bring with them such forms of awareness, and then it’s arguably the awareness that does the ill- founding – not the disposition itself. So we need to see if there is a way for the selective disposition to operate, without generating such awareness, and then we can ask whether it ill-founds beliefs when it operates that way.

But could there be such cases? How do you know what to avoid, without having enough information about it to have evidence that you’re avoiding evidence, or are likely to be avoiding evidence, or even just avoiding potential evidence against P, when you believe P? This problem might seem insuperable. On the one hand, the kind of selective disposition that matters requires some sensitivity to the status of what is avoided as a source of opposition to your belief. But on the other hand, it might seem that any such sensitivity would generate awareness of the selection that rationally pressures you to back off from your belief that P.

The solution is to examine more closely some ways of being sensitive to uncongenial information, sources of it, or potential sources of it. It’s useful to switch examples. The hat case isn’t ideal, because no matter how covertly your aversion to the hat-wearer is linked in your mind to aversion to believe her, it’s a burden on the imagination to picture being unaware of the aversion to believing her. There are more natural and familiar examples.

7. Outgroup hiring Hiring committees tend to respond differentially to outgroup applications and ingroup applications, in ways that perpetuate those groupings.20 What psychological mechanisms underlie these responses? The experimental evidence speaks more clearly to the responses themselves than it does to their cognitive underpinnings. But there are some natural hypotheses. First, the positive features of the ingroup applicants could be weighted more heavily than positive features of outgroup applications – in which case responsive dispositions are involved (presumably, bad evidential responses). Second, negative features of the application could be selected for further processing, while positive features are anti-selected – in which case selective dispositions are involved. I’m going to suggest that this second type of mechanism gives us an example of the kind of selective disposition we were looking for.

Both kinds of differential responses are predicted by the general principle that forming and maintaining beliefs are sometimes regulated by valences that attach to information when you take it in. Valences can take the form of a felt positive and negative affect. Once we factor in a potential role for affect in belief formation and maintenance, we can identify several ways in which positive features of outgroup

20 Steinpreis et al 1999, Goldin and Rouse 2000, Dovidio 2000.

19 applications could be systematically anti-selected, without generating the type of awareness that information is being ignored that on its own would ill-found the belief that the applicant is unqualified.

More specifically, in a search for information aimed at evaluating things, information has to be coded as strengthening or weakening the application. Coding those features could be mediated by felt valences. But what kind of information do the valences attach to? Suppose that instead of being mediated by the valences that match the features – positive valence from positive features, negative valences from negative feature – instead the valences that are felt are determined by match or mismatch with expectations. Then, given an expectation that outgroup applicants tend to be unqualified, positive affect will be generated by match with that expectation, and negative affect by mismatch with it.

A version of this principle is that felt discomfort (negative affect) tends to result from encountering information that is incongruent with what you already believe. That discomfort is cognitive dissonance.21 When cognitive processing is regulated by relieving or avoiding cognitive dissonance, then information that generates negative affect will be avoided, and if it is not avoided, the information will be discounted.22 Of course no one is doomed to read applications by putting a premium on avoiding cognitive dissonance. But the hypothesis that sometimes they are read in something like this way helps to explain the patterns of exclusion we find, and fits with what we know about the role that affect can play in our cognition.

Peter Railton has recently described the dynamics of affect and information using a simple analogy:

“If one likes a bit of food, one becomes more interested in information about it and more receptive to it. Where’s the harm in that? Well, if one likes an idea, one becomes more interested in it, and receptive to, information about it. If one dislikes an idea, the opposite. This makes wishful thinking endemic to the human mind. Deliberate self-deception is not required, just the natural tendency toward increased attention to, and easier acceptance of, information favoring what we like. …Belief is by its nature evidence-sensitive, but it also by its nature liable to prefer some information over others. And so we can be led down the primrose path by our very own selves.” – “The Value of truth, the Value of belief”

21 Festinger 1956, Hart 2009, Simon et al 1995. 22 Discounted, or scrutinized more thoroughly than congenial information, in a search for reason to discount it.

20 This description is so apt that one might wonder how far it generalizes. What if you have good grounds for liking an “idea” that sets off the belief-forming dynamics of interest and receptiveness? Railton articulates a sense in which those dynamics encode a preference to keep believing what we already believe, and to that extent, the dynamics belong to the overarching phenomenon of wishful thinking. Railton’s suggestion that such dynamics are set off by “liking an idea” leaves out an important type of trigger. As we saw before, one might not like an idea (e.g., that the world is going to end), and yet still be invested in it. Investment in an idea you don’t like can trigger the same dynamics of selection and uptake that Railton describes. The Outgroup hiring case is a potential example of an idea that subjects may dislike, yet be invested in maintaining – not necessarily out of malice, but because the belief is tied to so many others belonging to their conception of the outgroup.

8. Three ways to covertly select evidence How exactly can sensitivity to positive features of outgroup applicants mediate their anti-selection? And how could valences mediate the selection of information on the basis of which applications are assessed, without generating any awareness of the selection effect? This could happen in at least three ways.

Misclassification. Suppose you read that an outgroup applicant for a peer- counseling position was captain of the swim team, where this feature is independently measured as a strong feature of an application.23 But the qualification doesn’t make it on to a list of ‘pro’ side of a list weighing the candidate’s pros and cons, due to absence of positive affect that goes with processing it - even though a normal amount of information regarding the feature is taken in. Positive features are anti-selected for classification as positive. This route to anti-selection differs from classifying a feature as positive, but giving it less weight, compared to the same feature found in an ingroup application. There, the feature would make it on to the list of pros, but would not appear at the top of the list. What could explain the lack of positive affect, given that the search task is to find positive features of the applicants? By unconsciously assimilating the information’s incongruence with expectation that applicant is unqualified, you respond to the resulting negative affect that comes to the surface by not classifying the feature as positive.

Termination. In a second route to anti-selection, processing of the information about positive features of the application stops, because it doesn’t generate positive affect. Someone who relied on positive affect as a rough partial guide to positive features might stop reading about all the responsibilities that being captain involved, upon finding that they feel nothing positive when reading that X was captain of the

23 This much of the example is from Dovidio’s (2000) study about hiring.

21 swim team. The cutoff at that point is just an arbitrary example. Instead of cutting it off after learning that X was captain of the swim team but before reading through the attendant responsibilities, processing could get cut off after reading that X was a captain, but before learning what X was a captain of. At an extreme, each part of the application could get just a cursory look, without processing much of the information in it, and then end up in the ‘middling’ pile.

Unconsciousness or Forgetting. A third route to anti-selection would consist only of unconscious perception. If full processing of a feature and its ultimate classification as positive remained unconscious in its entirety, while conscious assessment of the application is guided only by conscious information. A close cousin of this route is to forget immediately uncongenial information once it is taken in. Here there is no need to appeal to negative affect of cognitive dissonance to explain how outgroup applications get rejected . It is enough that the applications seem to the reviewers at the time of decision to contain few positive features, if any.

In any of these three psychological configurations, there need be no felt aversion to the outgroup application, or to the applicant, of the sort illustrated by the aversion toward the hat-wearer in the café. Instead, aversion to uncongenial information influences when to end inquiry into one part of the application move on to new information.

The approaches I’ve labeled Termination and Unconsciousness could also operate in a perceptual case. Let’s say that a perceptual setting for F’s (e.g., squares) is a space in which, if it contains an F (square), you can experience the F (experience squares as square) by looking at it. A local F is an F in your perceptual setting. Suppose you hope that all the squares in a crowded visual display are red. Earlier, ‘rationalizing desire’ was defined in relation to a belief. But a desire can be rationalizing, relative to experience as well. Like a belief that P, an experience with content P can be rationalized by a desire that P, a desire to believe P, a desire that Q or a desire to believe Q, where the subject takes Q to support P, or to explain away apparent counterevidence to P.

The hope that all squares in the display are red will be a rationalizing desire, relative to an experience that all squares in the display are red.24 Your red-square hope could make you:

24 Could an experience have such universally quantified contents? Imagine seeing an otherwise empty egg carton with three eggs in it. Your experience does is not neutral on whether there are any other eggs in the egg carton besides the ones you see. It is difficult for contents to reflect this aspect of the experience without universally quantifying over something – locations, eggs, egg-shaped compartments, etc.

22 • disposed to stop unconscious processing of local non-F G’s (non-red squares) that would normally lead to G-experiences (square-experiences) – compare to Unconsciousness or Forgetting. • disposed to stop looking at local G’s (squares) that ambiguously or indeterminately appear F (red) – compare to Termination.

Both of these dispositions could operate without generating awareness that non-red squares are being ignored. I’ve argued that selective dispositions can operate without generating awareness of the selection. Now we can finally ask: when a selective disposition operates in this way, and when it explains why someone believes that P, can the dispositions ill-found the belief?

9. Does the implicit preference in the outgroup hiring case ill-found beliefs? Suppose that S on the hiring committee forms the belief that outgroup candidate X is a weak candidate, on the basis of examining X’s application, finding it not to have many positive features, as a result of a desire-dependent disposition to avoid encountering uncongenial information that operates in any of the three ways just described. Where P = X is weak applicant, and Q = outgroup candidates tend to be weak, S has a preference to keep believing Q. Because she has this preference, she is disposed to avoid encountering positive features of X’s application, and this disposition explains why she overlooks those features. S ends up believing that X is a weak candidate, manifesting her disposition to classify applications that seem to have mainly negative features as weak applications. By hypothesis, the disposition to avoid uncongenial information explains why S believes that X is unqualified. Does this disposition make any pro tanto contribution to ill-founding that belief?

You may find yourself with a strong intuition that the belief that X is unqualified is ill-founded. But the case exhibits a basic puzzle that challenges this intuition. We’ve met the puzzle before. Prejudice clearly makes an impact on S’s belief that P, but its impact is mediated by the evidentially good disposition: the disposition to classify applications that seem to lack positive features as weak applications. So is the belief made rational by its dependence on this evidentially good disposition, or is it made irrational by its dependence on the prejudice – the prejudice that selects the body of evidence over which the evidentially good disposition operates?

In response to the puzzle, the original intuition may evolve into an intuition that dependence on prejudice is not vindicated by dependence on evidentially good dispositions. But we can get underneath this intuition in ways that complicate its status as supporting the thesis that selective dispositions can be ill-founding.

Evidentially good dispositions can exist alongside different, evidentially bad dispositions. Let’s say you have a good disposition, if you are disposed to classify candidates as strong, weak, or intermediate, in accordance with whether upon

23 examining their application, you assess it as having strong, weak, or intermediate qualifications. You might simultaneously be disposed to weigh positives features of ingroup candidates more heavily than the same features in outgroup candidates, and that disposition. The same basic psychological mechanism could easily underlie both selectively overlooking positive features of the application, and discounting such features when they are brought into the open. Frequently people avoid information that they don’t know how to handle, and perhaps the intuition is tracking this fact. If the intuition about the outgroup hiring case is really an intuition about this type of mixed case, then it doesn’t address the key normative question, which is whether the selective disposition by itself makes any pro tanto contribution to ill-founding.

But we don’t have to rely exclusively on intuition, because we can make progress with the key question by returning to the three mechanisms of anti-selection. In all three cases, the selective disposition is mediated by bad evidential (responsive) dispositions. Assuming that those evidential dispositions are ill-founding, the selective disposition is too.

In the misclassification route to anti-selection, the belief that being captain of the swim team does not belong on the list of pros is not well-supported by the file- reader’s evidence, because they have evidence that it is a strong qualification – they have just read through all the responsibilities the captain has. By hypothesis, in assessing ingroup candidates, they would classify it that way themselves.

In the case where information processing is cut off upon finding no positive affect at initial stages, you have less information that the feature is positive, compared to what you have when you keep reading about all of the responsibilities of the swim team captain. If you get so far as examining a feature but don’t take in much information about it, you should suspend judgement on whether the feature is positive, unless background information justifies pushing you one way or the other. If nothing in the background can play that role, then concluding that the feature is not positive is ill- founded.

What if taking in information and classifying it as positive remains entirely unconscious, or is quickly forgotten? Does being unconscious or forgotten remove the information from the domain of evidence that you have? Arguably – though I won’t argue it here – the domain of evidence is not so limited. In some cases where unconscious perception guides beliefs and behavior, it seems to retain some of its rational impact on the subject’s beliefs.25 But if (as per Feldman and Conee) unconscious or forgotten evidence doesn’t retain any rational force, then the file-reader doesn’t have evidence that the outgroup candidate’s file has the forgotten (or unconsciously remembered) positive feature, and so this information would not defeat the negative

25 For defense of this view, see Siegel and Silins (forthcoming).

24 assessment of the candidate. And in that case, it is hard to see how the resulting belief would be ill-founded, given the assumption that it is formed only on the basis of conscious information indicating few if any positive features. For all I’ve argued so far, the belief that X is unqualified could be well-founded.

Is there another route to ill-foundedness, independent of whether unconscious or forgotten information counts as evidence? Everyone can agree that in the Outgroup hiring case, evidence is not collected carefully or thoroughly. At least sometimes, gathering evidence in this way violates a norm that the inquirer is subject to. I’m going to leave aside exactly what content and structure evidence-gathering norms have. Whatever those norms are, the Outgroup hiring case raises the question of how they relate to epistemic norms governing doxastic attitudes. If you come to have a belief that P by inquiring in a way that violates norms of gathering evidence that you are subject to, does that ill-found your belief?

On the one hand, it might seem straightforward that beliefs can be ill-founded by improper evidence-gathering. Improper evidence-gathering can explain why the subject forms a belief that P as opposed a different belief, and beliefs are epistemically appropriate or inappropriate depending in part on how they are formed. So improper evidence-gathering seems exactly the type of process that could bear on a belief’s epistemic status.

On the other hand, if the epistemically inappropriate beliefs are exactly those formed properly in response to the evidence you have, then a belief cannot be made epistemically inappropriate by the fact that there is evidence that a subject doesn’t have. On this picture, improperly gathering evidence does not violate the epistemic norms governing doxastic attitudes.

To assess the issue, what’s needed is a way to delineate the boundaries of distinctively epistemic appropriateness and inappropriateness of belief-formation well enough to determine if properly forming beliefs in response to whatever evidence you have exhausts it. To sharpen the issue, we can compare two cases of belief-formation in the red square case, in which the subjects respond to the same information, but come to have that information in different ways. In the hopeful scenario, the subject’s hope that all squares in the display are red makes her overlook the non-red squares (say by cutting off processing of those squares before the information that they’re non-red squares can become conscious), whereas in the bad luck scenario, the subject simply overlooks the non-red squares by accident. Both subjects who see the display end up seeing only the red squares (along with other shapes in the display), and so both reach the same generalization: all the squares in the display are red. (We can assume that both subjects have the sense that they looked carefully enough to have good grounds for generalizing).

25 Is there anything epistemically irrational about the hopeful subject’s belief? Does the fact that her red-square hope limited which evidence she had ill-found her generalization? You might try to explain away the impression that the hopeful subject’s generalization is irrational, by finding a type of non-epistemic irrationality to assign to the hopeful subject, such as the prudential irrationality that might go with systematically overlooking some squares when you’re trying to find them all. But this type of strategy seems unpromising in this case, since the subject’s red-square hope could still influence what she sees, even if she wasn’t trying to find all the squares. Moreover, avoiding the red-squares seems prudentially rational – not irrational - relative to a preference for avoiding information uncongenial to her hope. A similar point applies to the outgroup hiring case. Relative to the goal of assessing applications and preferring to maintain the belief that outgroup applicants are unqualified, avoiding uncongenial information is prudentially rational. And if there are conflicting goals or preferences, then norm of prudence is harder to apply.26 So this strategy for immunizing the epistemic status of beliefs against improper evidence-gathering do not seem promising.

Besides considerations about what kind of flaw is exhibited by evidence- gathering in the Outgroup hiring case, another type of consideration suggests that evidence-gathering can ill-found beliefs. To see it we need to focus on an aspect of the route to belief that we have glossed over so far.

In both the Outgroup and the red square case, the belief (All squares in the display are red, X is unqualified) results from a transition from considering multiple bits of evidence from parts of a whole. From a series of experiences of red squares and non- squares, the subject concludes that all the squares in the display are red. From a series of what appear to be mediocre qualities, the reader concludes that the application is weak. These transitions are generalizations about a whole on the basis of local bits of evidence about its parts. In both cases, a desire (red-square hope, or a preference to maintain the belief that outgroup candidates are weak) arranges for you to avoid getting evidence that would defeat this generalization. (Let’s assume the mechanism does not operate through misclassification or terminating search). Even if we assume that unconscious information has no defeating power, we can find a leftover ill-founding potential in the transition to the generalizations.27

What does it take for the transition to be epistemically appropriate? At one extreme, if you had a justified belief that you looked well and long enough to see all the

26 I’m indebted to Shantia Rahminian for extensive discussion of this type of strategy, which he uses to argue for epistemic reasons for furthering inquiry in Rahminian (ms). 27 The same points hold if you are looking and conclude something isn’t there. You look in the café and find: no Pierre. Here you generalize about the things in the café – none of them are Pierre.

26 squares, then that would presumably bolster its epistemic appropriateness. But do you need to have justification to think you looked well and long enough, for the transition to be appropriate? Or could it be enough simply to have looked well and long enough?

We don’t need to answer that question, because what matters in our cases is what it takes for the transition to be epistemically inappropriate. First, if you have justification to believe that you haven’t looked long and well enough, then that would presumably do it. That could easily be your situation, if you were aware of the selection effect. Second, if you have an unjustified belief that you have looked long and well enough, then that too has a good claim to making the transition epistemically inappropriate – even if you wouldn’t need justification to believe that you did look long and well enough, to be justified in generalizing.

In our examples, where the subjects aren’t aware of the selection effect, could they have such an unjustified belief? If it seems to them that they have examined the display of colors shaped normally, or that they read applications the way they’re accustomed to doing in searches that everyone takes to be normal, it seems plausible that they assume that they have looked long and well enough. And in those very same cases, that assumption could be driven by wishful thinking: the hope that all the squares in the display are red screens out the non-red squares, and the preference to maintain the belief that outgroup candidates are unqualified screens out (or leads to misclassifying) positive features in the file. When that happens, the assumption is arguably doxastically unjustified. More exactly, even if the length and quality of searching provides propositional justification for the assumption, or even if the state of seeming to have searched well enough does, one might make the assumption because one wants to be done searching, or because they want to avoid uncongenial information, or because they’re hungry. In those cases, the assumption is doxastically unjustified.28

Another example brings the position into focus. Imagine that on the third forkful of a nutty meal, you bite down with a suspicious crunch. It might be a small rock which could chip a tooth, or it might just be a hard nut. You try to locate the source of the suspicious crunch, chewing carefully in case there really is a rock in your mouth. You don’t find anything unusual, so you go back to chewing normally, teeth intact.

Let ~P = there is no rock in your mouth. What factors determine whether it is reasonable conclude ~P? These factors presumably include the experience (the feel, the sound) of the crunch, the length and quality of exploration, and perhaps the stakes

28 If seeming to have looked well enough is distinct from believing that one has, but you got to be in that seeming-state by exercising an ill-founding selective or responsive disposition, does the seeming state does even provide propositional justification for the belief? According to the epistemic symmetry argument of Part III, it doesn’t.

27 involved in chipping a tooth. Unlike the experimental context of the hiring case, where it has to be clear (to conduct the experiment) what counts as strong or weak evidence for a hiring verdict, here no one could plausibly identify with much precision what those interactions are. (Real hiring cases are like that too, as are real searches of crowded visual scenes). And yet we can identify clear cases of resuming chewing too soon (especially in the first person case, where we often have definite views about when it would be too soon to start chewing again). Putting these points together, we can’t obviously rule out the existence of cases in which you should keep searching for a rock, and yet you have strong evidence that there is no rock in your mouth.29

Now suppose that after failing to find any rocks in your mouth, you stop searching for a rock, because you’re (unjustifiedly) sure that nothing bad will happen to you today, and form the belief that there’s no rock in your mouth (~P). According to the position I’m describing, even if you have strong evidence for ~P, and you are sensitive to changes in that evidence, if you have an unjustified belief that you have looked well and long enough for a rock, that can ill-found your belief that ~P.

Analogously, the desire to maintain belief that outgroup applicants are unqualified could make you unreasonably confident that you’ve searched and scrutinized long and well enough, even if the conclusion you reach fits with the evidence you’ve covertly limited yourself to have. Even if you respond to that evidence perfectly well, the belief formed on the basis of that evidence could still be ill-founded.30

I’ve argued that the belief in the generalization in hiring, red square, and rockmouth cases can be ill-founded by doxastically unjustified assumption on the part of the subject that she has looked long and well enough to justifiably generalize.31 In

29 Suppose it impossible to be such that should keep looking while it is also rational to believe P (cf Friedman (2013), Holton forthcoming). Then the case as described might seem to be at odds with the “JJ” iteration principle for justification: J(Bp) iff JJ(Bp). But in the first instance what the description rules out is an interation principle about evidence: E(Bp) iff JE(Bp). The JJ principle is ruled out, only with the additional assumption that evidence exhausts justification. 30 A natural consequence of this position is that to correct the epistemic flaw, you would have to be disposed to scrutinize the application further – yielding an epistemic reason for action. For discussion of this idea, thanks to Shantia Rahminian and Selim Berker. 31 Laying the blame on the assumption brings out a difference between the position described here and Kelly’s (2008) position on belief polarization (a phenomenon where giving the same evidence to people with opposite attitudes toward the same proposition does not lead to convergence). Kelly argues that if both sides start out with well-founded belief and end up maintaining the belief by anti-selecting uncongenial information, that could leave each side’s belief well-founded. Since his account applies only to the case where people begin with a well-founded belief, the preference to

28 describing the position on these cases, I’ve spoken as if the evidence for the generalization (no rock in mouth, candidate is unqualified, all squares are red) could be sufficient to justify believing it, yet the belief is still ill-founded. But depending on what evidence is, the ill-founding factor could be either evidential or extra-evidential.

In the evidential form, the ill-foundedness reduces the force of evidence, relative to what it could be if the assumption were not unjustified. For instance, the individual red-square experiences would provide less support for believing the generalization that all squares in the display are red, in the hopeful scenario as opposed to the bad luck scenario. In the Rockmouth case, you’d have strong evidence to believe there’s no rock in your mouth if you weren’t overconfident, but your overconfidence reduces the force of the evidence you have. This evidential option is at odds with Feldman and Conee’s construal of evidence, where evidence you have at a time supervenes on mental sates you have at that time. A different evidential option says that for the local bits of evidence to support the generalization, you need justification to believe you looked well enough, and so a necessary element is missing when that assumption is unjustified. This evidential option is compatible with Feldman and Connee’s construal of evidence.

In the extra-evidential form of ill-founding, the force of the evidence is not reduced, but the subject has less rational support for believing the generalization than she could have, with the same evidence, if the assumption were not unjustified. The extra-evidential option is at odds with evidentialism.

This concludes my case for the Analogy Thesis. Where does that leave us? Recall the Simple Argument:

Premise: Beliefs can be ill-founded by wishful thinking. Premise: Wishful seeing is possible. Conclusion. Experiences can be ill-founded by wishful seeing.

If the Analogy Thesis is made true by selective and responsive dispositions, then wishful seeing can in principle take the form of modes of dependence on desire that can ill- found beliefs. To go the rest of way in defending the Simple Argument, what’s needed are reasons to think that selective and responsive wishful seeing do to experiences what the same forms of wishful thinking can do to beliefs: ill-found them.

maintain that belief is not as obviously ill-founding. In contrast, when the preference to maintain an unjustified belief (“nothing bad will happen today”, “outgroup candidates are unqualified”) drives the selection of evidence, it seems intuitively worse. Residual question: if a preference to maintain a justified belief makes you both anti-select uncongenial information, and makes you assume you have looked long and well enough at the congenial information you have, is that assumption ill-founding?

29

Part III. Epistemic Symmetry Suppose we ask: what explains why wishful beliefs (in the core phenomenon) are ill-founded by their dependence on a desire? Around this question, we can build a strategy for arguing that experiences are vulnerable to ill-founding by wishful seeing.

The strategy is this. First, consider a belief and an experience that arise by the same mode of dependence on desire (and that have the same contents). Then find the features of belief that explain why that mode of dependence ill-founds beliefs. Then argue that these features of belief are features of experience as well.

The central reasoning behind the strategy is that if a belief’s mode of dependence makes it ill-founded, because the belief has feature X, and experience E with the same mode of dependence has X, then that’s a reason to think that mode of dependence makes E ill-founded as well. Proposals that identify a X shared by beliefs and experiences are symmetrist proposals. In contrast, if belief B’s mode of dependence makes it ill-founded, because it has feature X and X is distinctive of belief, not had by experiences, then that gives little reason if any to think that wishful seeing can ill-found experiences. Proposals according to which X is distinctive of belief are asymmetrist proposals.

10. The Argument from Endorsement According to my symmetrist proposal, both beliefs and experiences can be ill- founded by responsive or selective modes of dependence on desire, because both beliefs and experiences are states of endorsing contents. Endorsement states are states that characterize how things seem to be, not how we take them to be in imagination or pretense, or how we take them to be for specific purposes, as we do in supposition (exploring consequences) or acceptance (practical purposes).

Endorsing contents differs from endorsing experiences. The upshot of endorsing an experience is a belief. When a belief that P is formed by endorsing an experience, the belief is formed on the basis of an experience that includes P in its content. Beliefs cannot be endorsed, whereas experiences can be.

The fact that experiences can be endorsed is the flipside of the fact that they can be disbelieved, or simply not believed – you might neither believe nor disbelieve your eyes. When experience that P is disbelieved (or not believed), its status as a state in which it seems to you that P can survive.32 Refusing to believe your eyes is in one respect like

32 It need not survive, if disbelieving the experience changes its content. For instance, Searle (1983) claims that upon learning that the you’re looking at a house-façade rather than at a house, your experience will be different. This could be a case where the original experience had the content that could (roughly) be expressed by saying “That is

30 refusing to believe testimony that you take to be sincere. Compare Evans’s remark in discussing perception that “our being placed in the appropriate information states by someone telling us a story does not depend upon our believing the story to be true.”33 He is observing that we don’t have to endorse our experiences, in order for experiences (as he would put it) to present us with information.

Evans’s analogy between having an experience and receiving testimony is partly apt, but not entirely. He is right that we can refuse to believe our eyes, thereby failing (or ceasing) to endorse our experiences, while still being presented, by our experiences, with the same information that we are refusing. Our relation to testimony is like that too. But even when that happens, the experience survives with a more committal character than we might have toward testimony that we take to be sincere but refuse to believe. The testimony comes from someone else, whereas our experiences are our own. If you refuse to believe your own experiences, then you are much closer to having contradictory beliefs than you would normally be just by refusing to believe a bit of testimony from someone else.34 You are closer, with respect to phenomenology: it might feel uncomfortable in the same way that it can feel uncomfortable to discover that you are committed to P and that you are also committed to not-P. In contrast, you need not feel any discomfort in listening to someone make the case for why cheese originated in New Jersey, even if you do not believe a word they say.

In being states of endorsing contents, what beliefs and experiences share is a endorsement of contents, and endorsement is a kind of commitment. Both experiential endorsement of P and belief that P is compatible with disbelieving P: in both cases you end up with something like contradictory beliefs.

Why think being an endorsement of content explains why wishful thinking makes beliefs conduct ill-foundedness? One way to think of ill-founding is that when a belief B1 ill-founds a belief B2, an epistemically good-making feature of B1 is missing. (This doesn’t mean that the good-making feature inheres in B1.) That feature is: providing rational support for B2. Recall Frank, who believes that his good features are evident to all. If Frank’s belief (B1) is well-founded, then it can well-found his interpretations of

a house”, and disbelieving it resulted in having an experience that was less committal on the ways in which the facing surface continue out of view. 33 Evans (1982), p. 123. Experiences will still share the relevant features with belief, even on many construals of them on which they don’t have contents. For instance, if experiences are relations to facts, or to clusters of (instantiated or uninstantiated) properties, these relata will still determine propositions, and the experiences can be seen as endorsements of those derivative contents. For discussion, see Siegel 2010 ch 2 on the Weak Content View. 34 If you believe P and disbelieve P, then that is arguably always irrational, but if your experience has content P and you disbelieve P, that is never irrational.

31 how others are responding to him (they’re embarrassed to express their admiration, etc), when combined with auxiliary assumptions (showing admiration for strangers can be embarrassing). But for Frank’s belief B1 to support those other beliefs, B1 has to be endorsement. At least, there are some other beliefs B2, such that B1 has to be an endorsement to provide rational support for B2.35

In contrast to the symmetrist proposal, various asymmetrist proposals don’t explain why wishful thinking makes beliefs conduct ill-foundedness. There are many asymmetries between experiences and beliefs. Let’s consider the ones that seem best candidates for grounding epistemic asymmetries.

According to a first asymmetrist proposal, wishful thinking ill-founds beliefs, because they make beliefs irrational, whereas experiences can’t be irrational. This proposal is inconclusive at best. Wishful beliefs are ill-founded because of the way they depend on desire – and if the Analogy Thesis is true, then experiences can depend on desire in the same way (more exactly, in at least two of the same ways – the two main modes of dependence we find in the human brand of wishful thinking). In addition, ultimately, as was noted at the start, a natural conclusion to draw from the Simple Argument is that there is a sense in which experiences are made irrational by those modes of dependence. That conclusion of course has no dialectical weight here, but it brings out a way in which the first asymmetrist proposal is simply denying it.

The other asymmetry between beliefs and experiences that seems most relevant concerns the roles that belief can play in relations of rational support. A first proposal along these lines is that beliefs can never start off chains of justification, but experiences sometimes can. Another is that beliefs need to be rationally supported by other states, but experiences don’t.36 Perhaps these features of belief explain why they can be ill- founded by wishful thinking, while wishful seeing has no impact on the justificatory power of experience.

How exactly is the hypothesis that beliefs cannot originate chains of justification related to the fact that they can be ill-founded by wishful thinking? Consider this line of thought: a belief could not be ill-founded by wishful thinking, if it could in principle originate a chain of justification. But beliefs can never originate chains of justification. So

35 Suppose that low credal states are not endorsements of content, and suppose that Frank has only low credence in the proposition that his many virtues are evident to casual observation, or suppose he only imagined it, rather than believing it. Then it wouldn’t support the other outright beliefs. Alternatively, perhaps low credal states really are endorsements of content, a la Joyce (1998). Or perhaps there are no credal states. 36 Huemer (2013)

32 the inability to originate chains of justification is at least an enabling condition for being vulnerable to ill-founding by wishful thinking.

By itself, being unable to originate chains of justification does not illuminate why beliefs are ill-founded by wishful thinking. Suppositions cannot originate chains of justification either, but they are not ill-founded by wishful thinking. If being unable to originate chains of explanation plays any role in an explanation, it is compatible with the argument from endorsement.

A response to both putative asymmetries is that they are not perfectly general, because some beliefs do originate chains of justification, and don’t need rational support from other states. Consider one of the principles that guide perceptual judgments about of object permanence: objects move as bounded wholes.37 Spelke calls this the Coherence constraint. This constraint influences which judgments we make in response to what we see, such as our judgments about how surfaces we see continue out of view. These judgments (I’m assuming) are often justified, and the coherence constraint plays a role in justifying them. Yet there is strong experimental evidence that the principles are not themselves gained from experience. If our reliance on them takes the form of belief, then they seem to be beliefs that originate chains of justification.

One might suggest that our reliance on the Coherence constraint does not take the form of belief, any more than the constraints on the visual system do – such as the constraints that allow us to compute edges from light contrasts.38 Our visual processing may conform to principles, without our believing them. None of the usual diagnostics of belief seem to apply. We are not disposed to judge that certain patterns of light contrast indicate edges. They are brought to bear only on a highly limited range of incoming information – namely, information about light contrasts taken in through vision. (That is what makes them good candidates for intramodular information).

But there is an important difference between this constraint and the principles that guide judgments of object permanence. The latter principles do meet many of the diagnostics of belief. We can draw on them to reach verdicts about hypothetical circumstances, without perceptual input. Suppose you are told that a ball is about to come out of a tube without touching anything else, and asked where its parts will be relative to one another once it comes out. Will its parts will scatter and separate, or will they continue to constitute a ball? The choice is clear. You don’t need imagery or perception to have an opinion about this question.39

37 Spelke et al 1995. 38 For discussion of this issue regarding the visual system see Burge 2010. 39 An opponent might still resist by granting that we believe the Coherence constraint, but holding that we encode the same information twice over, and in two different ways: once in a belief, which is formed in response to experiences, and again more primitively,

33

Putting these points together, we seem to believe the Coherence principle, and it is a good candidate for originating chains of justification, and so the putative inability to originate chains of justification is a poor candidate for explaining why beliefs are vulnerable to ill-founding by wishful thinking.

Does it really matter, though, whether there are beliefs that originate chains of justification? The asymmetrist might propose that even if there are beliefs that do this, such as our belief in the Coherence constraint, the rational support offered by that belief does not derive from the mere fact that we have the belief. And therein lies another putative asymmetry with experiences: sometimes, when experiences provide rational support, they provide it just by the fact that you are having an experience. But it is never the case that when a belief provides rational support, it provides it just by virtue of the fact that you have the belief.40

Some epistemological conservatives about belief might reject this thesis about beliefs. In addition, it seems a live option that certain combinations of belief taken together do exert rational pressures inherently – such as structural rationality requirements.41 But let us grant for the sake of argument that no rational pressures inhere in any beliefs. How could this observation help explain what makes beliefs vulnerable to ill-founding by wishful thinking?

Let’s focus on the rational support provided by an experience with content P for believing P. An asymmetrist might argue this way: if it is irrational to believe P on the basis of an experience with content P, then an amount of rational support for P that inhered in the experience had to be removed. In contrast, if a belief fails to provide rational support for P because it is irrational, it is never because some inherent support the belief provided for P was removed. So, according to the asymmetrist, if wishful seeing ill-founds experiences, it can’t do so for the same reasons that wishful thinking ill- founds beliefs.42

in a form that is not belief -- and that it is only the primitive form that originate justification. In regarding the object-permanence principles as taking the form of intramodular information, as well as beliefs, this position needs to explain: why do we end up believing object permanence principles, but not every other piece of putative intramodular information, such as rules of c-command? The natural explanation for the difference is that we believed the principles all along. 40 Thanks for Matt McGrath for formulating the proposal this way. 41 For instance, if you believe P and you believe (If P then Q), then together these beliefs rationally pressure you either to conclude Q or give up P. 42 Thanks once more to Matt McGrath for discussion of this line of thought.

34 The asymmetrist’s line of thought relies on the thesis that such rational support inheres in the experience:

Inherence thesis: If wishful seeing makes it irrational to believe, then a source of rational support for P that inhered in the experience is reduced or removed.

The inherence thesis is stronger than the thesis that just by having an experience with content P, you can have reason to believe P. It takes a stand on the features in virtue of which the experience provides reason to believe P, by saying that those features inhere in the experience, as opposed to being features that explain why you have such an experience as opposed to have a different experience, or no experience at all.

In reply, if the Inherence thesis is true, then any epistemic impact of wishful seeing on experience differs structurally from the epistemic impact made by wishful thinking on belief. But it does not point to any difference in why the modes of dependence on desire lead to such epistemic compromises. Nor does it point to any feature of belief, not shared by experience, that explains why wishful thinking ill-founds beliefs. For all it says, beliefs are vulnerable to ill-founding, because they are endorsements of contents.

11. Two overgeneration worries I’ve argued that in their selective and responsive modes, wishful thinking ill- founds beliefs, and wishful seeing ill-founds experiences. We could think of these modes of dependence as irrational modes of dependence.

According to the Argument from Endorsement, the vulnerability to ill-founding of both experiences and beliefs and is explained by a feature they share. One might then expect experiences and beliefs to be epistemically symmetrical in many other respects as well. And that might seem to be the wrong result. Forget about wishful seeing – what about regular old experiences? Those, let’s suppose, have an a-rational etiology. But don’t we find an asymmetry with beliefs that arise in comparably a-rational ways? Suppose God zaps you into believing that Frank is great, or that all the squares in the display are red. Intrasubjectively, there is nothing like reasoning or any of the usual modes of belief-formation. There is just a divine zap. If this belief would be unjustified, and experiences are epistemically symmetrical with beliefs, then doesn’t the Argument from Endorsement that regular old experiences are ill-founded as well?

A different overgeneration worry concerns beliefs that are well-founded by other beliefs. If experiences could depend on expertise in the same ways that beliefs do, then the Argument from Endorsement would seem to predict that those experiences are well-founded in the same sense – and does that even make sense?

We can represent these two overgeneration worries as the second and third lines in a chart:

35

Intrasubjective etiology Experience Belief Irrational ill-founded ill-founded Arational OK ill-founded Rational ? well-founded

In reply to the first worry, let us sharpen it. The symmetry argument does not predict that if a belief that arises from an a-rational zap is ill-founded, then any experience that arises in the same way is also ill-founded. What it predicts is that if a- rationally zapping a belief into a mind makes that belief ill-founded, then a-rationally zapping an experience into a mind should ill-found the experience as well. It would challenge the symmetry argument, if there was a case like this:

The fact that S has belief B and S* has experience E are both explained by the same a-rational factors, those factors ill-found B, but intuitively E is not ill- founded at all.

For instance, suppose God zaps S into believing that all squares in the display are red, and zaps S* into experiencing that all squares the display are red. (To simplify, let’s suppose S doesn’t also see the display, or have any experience that characterizes it). What is the epistemic status of S’s belief, and what gives it that status?

Of course the case is underdescribed. A-rational zaps per se do not generally make a pro-tanto contribution to ill-founding – if they did, then we’d have to convict our belief in the principle of non-contradiction, and our belief that objects move as bounded wholes (Spelke’s Coherence constraint), as being as least partly ill-founded. And there are versions of the zap case where S’s belief is arguably well-founded. Suppose she has a well-founded belief that she has a way of knowing about the colors of the shapes in the display, or a memory (even a fake one) of seeing the display. More generally, you could be zapped into believing P, when you have lots of rational support for P but hadn’t (pre- zap) drawn the conclusion that P. If the belief you got by the zap was subsequently controlled by your reasons, then arguably the belief is well-founded.

At a minimum, the challenger needs a version of the case where the belief is intuitively ill-founded, so that we can ask whether the zap is to blame. The strongest case seems to be one where S concludes, from her zapped belief that all the squares in the display are red, that she must have a way of knowing what color the squares in the display are.43 But here what makes the B ill-founded is the fact that S has no rational

43 The challenger needs to give S the higher-order belief, because without it the zapped belief is arguably defeated (and therefore ill-founded) by its lack of integration with S’s

36 support for it, and B (unlike the Coherence constraint) needs rational support in order to be well-founded. Getting a belief through a zap is a convenient way to get into a situation where you have a belief without having any rational support for it. But zapping does not itself constitute being in that situation. So it is not to blame for B’s ill- foundedness.

The chart is right that a-rational processes (like getting zapped) can lead to beliefs that are ill-founded, and that analogous a-rational processes lead to experiences that aren’t ill-founded. But since the a-rational formation processes do not isolate the factors that ill-found the beliefs, the symmetry argument does not predict that analogous processes ill-found experiences.

The large residual question in the vicinity is whether experiences are like (most) beliefs in needing rational support, in order to provide it. It seems they don’t, and the symmetry argument does not predict otherwise.

The second overgeneration worry raises the question whether there could be another argument like the one with which we began:

Another Simple Argument P1*. Beliefs can be well-founded by expertise. P2*. Expert seeing is possible. Conclusion: Experiences can be well-founded by expertise.

This argument raises parallel issues to the ones that brought us up to this point. They are too complicated to discuss here.44

References Bengson, J. (Forthcoming) “Presentation and Content: Critical Notice of The Contents of Visual Experience. Nous. Ditto, P. et al. (1998) “Motivated Sensitivity to Preference-Inconsistent Information” Journal of Personality and Social Psychology 75: 53-69. Dovidio, J. and Gaertner (2000) “Aversive Racism and Selection Decisions: 1989 and 1999”. Psychological Science. Vol. 11, No. 4. Ehrlich, D. et al. Postdecision exposure to relevant information. Journal of Abnormal & Social Psychology vol. 54 issue 1 January 1957. p. 98-102 Evans, G. (1982) The Varieties of Reference. Oxford University Press. Feldman (2003) Epistemology. Upper Saddle River, NJ: Prentice Hall. Festinger, L., & Maccoby, N. (1964). On resistance to persuasive communications. The Journal of

other doxastic states – either her belief that she has no way of knowing what color the shapes have in the display, or agnosticism about whether she has a way of knowing that. 44 For extensive discussion, thanks to Eric Mandelbaum, Jim Pryor, Nico Silins, Scott Sturgeon, Jonathan Vogel, and especially Matt McGrath.

37 Abnormal and Social Psychology, 68(4), 359. Festinger, L. (1962). A theory of cognitive dissonance (Vol. 2). Stanford university press. Festinger, L. (1956) When Prophesy Fails. Fodor, J. (1983). The Modularity of Mind. Cambridge: MIT Press. Friedman, J. (2013) “Suspended Judgment” Philosophical Studies 162:2. Giner-Sorolla (1997) “Selective use of heuristic and systematic processing under defense motivation” Personality and Social Psychology Bulletin 23, 84-97. Goldin, C., & Rouse, C. (1997). Orchestrating impartiality: The impact of" blind" auditions on female musicians (No. w5903). National Bureau of Economic Research. Hart, W. et al (2009) “Feeling Validated versus Being Correct: A meta-analysis of selective exposure to information” Psychological Bulletin 135, 555-588. Hirschberger, C. (2006) “Terror management and attribution of blame to innocent victims: Reconciling compassionate and defensive responses”. Journal of Personality and Social Psychology 91: 832-844. Holton, R. (forthcoming) “Intention as a Model of Belief”. In Vargas and Yaffe, Eds. Rationality and Social Agency: Essays on the Philosophy of Michael Bratman. Oxford University Press. Huemer, M. (2013). “Epistemological Asymmetries between Belief and Experience” Philosophical Studies 162, 741-748. Johnston, M. (2006) “Better than mere knowledge: the function of sensory awareness” In Gendler and Hawthorne, eds. Oxford University Press. Johnston, M. (2004) “The Obscure Object of Hallucination” Philosophical Studies 120.113-83. Joyce, J. (1998) A nonpragmatic vindication of probabilism. Philosophy of Science 65 (4):575-603. Kelly, T. (2008) “Disagreement, Dogmatism, and Belief Polarization” Journal of Philosophy 105. Kornblith (1983) “Justified Belief and Epistemically Responsible Action” 92:1, 33-48. Kvanvig, Jonathan. (2003) “Propositionalism and the Perspectival Character of Justification.” American Philosophical Quarterly 40.1, pp. 3-18. Lerner, Melvin J. (1982), The Belief in a Just World: A Fundamental Delusion (New York, NY: Plenum Press, 1982) Lieberman, M. D., Ochsner, K. N., Gilbert, D. T., & Schacter, D. L. (2001). Do amnesics exhibit cognitive dissonance reduction? The role of explicit memory and attention in attitude change. Psychological Science, 12(2), 135-140. Lord, Ross and Leper (1979) “Biased assimilation and attitude polarization: The effects of prior theories on subsequently acquired evidence”. Journal of Personality and Social Psychology 37: 2098-2109. Mandelbaum, E., and Ripley, D. (2012). “Explaining the Abstract/Concrete Paradoxes in Moral Psychology: NBAR Theory”. Review of Philosophy and Psychology 3 (3): 351-68. DOI: 10.1007/s13164-012-0106-3. Molden, D. and Higgins, E. (2012) “Motivated Reasoning”. In Oxford Handbook of Thinking and Reasoning. Eds. K. Holyoak and R. Morrison. Oxford University Press. Moscowitz, Gollwitzer et tal (1999), “Preconscious control of stereotype activation through chronic egalitarian goals” Journal of Personality and Social Psychology 77: 167-184. Railton, P. (ms) “The Value of Truth and the Value of Belief”. Rahminian, S. (ms) “Epistemic Reasons for Action”. Santioso, R. Kunda, and Fong (1990) “Motivated Recruitment of Autobiographical Memories” Journal of Personality and Social Psychology 59: 229-241. Scanlon, T. (2007) “Structural Irrationality”. In Common Minds: Essays in Honor of Philip Pettit. Eds. Brennan et al. Oxford University Press. Sherman, J.W. et al (2005). Prejudice and stereotype maintenance processes: attention, attribution, and individuation. ” Journal of Personality and Social Psychology 89. 607-622. Siegel, S. (2010) The Contents of Visual Experience. Oxford University Press. Siegel, S. (2013) “The Epistemic Impact of the Etiology of Experience”. Philosophical Studies 162. Siegel, S. and Silins, N. (forthcoming) Attention and Perceptual Justification. Festschrift for Ned Block. Eds. Pautz and Stoljar. MIT Press. Simon, L., Greenberg, J., & Brehm, J. (1995). Trivialization: the forgotten mode of dissonance reduction. Journal of Personality and Social Psychology, 68(2), 247. Shweder, R. et al 1997. The ‘Big Three’ of Morality (Autonomy, Community, Divinity) and the ‘Big Three’

38 Explanations of Suffering. In Morality and health, eds. A. Brandt and P. Rozin. New York: Routledge. Spelke, E., Phillips, A. and Woodward, A. (1995) Infants’ Knowledge of Object Motion and Human Action. Causal Cognition: A Multidisciplinary Debate. Clarendon Press/ Oxford University Press, Oxford, pp. 44– 78. Steinpreis, R. E., Anders, K. A., & Ritzke, D. (1999). The impact of gender on the review of the curricula vitae of job applicants and tenure candidates: A national empirical study. Sex roles, 41(7-8), 509-528. Tagliacozzo, R. (1981). Smoking behavior and exposure to “dissonant” information. Substance Use & Misuse, 16(7), 1149-1159.

39