Portal de Periódicos da CAPES

Beliefs as dispositions to make judgments

2023; Wiley; Volume: 106; Issue: 3 Linguagem: Inglês

10.1111/phpr.12993

1933-1592

Anna‐Sara Malmgren,

Philosophical Ethics and Theory

I'm very pleased to have the opportunity to comment on Declan Smithies’ thought-provoking, creative, and ambitious book: The Epistemic Role of Consciousness (‘ERC’).1 Unfortunately I can only discuss a small selection of the issues that it covers here. I'll focus on the conception of belief that Smithies defends in Ch. 4–5 (and further in Ch. 7–10): beliefs as dispositions to cause judgments—specifically, as one-track dispositions to cause phenomenally conscious judgments. Smithies’ conception of belief has two noteworthy components: first, the claim that beliefs are fully constituted by dispositions to cause (conscious) judgments (132).2 Second, the claim that judgments have phenomenology: there's ‘something it's like’ to make a judgment (4). Indeed, Smithies holds that judgments have both attitude- and content-specific phenomenology—that's to say that the judgment that p has a certain phenomenal character in virtue of being a judgment that p (rather than, say, a desire that p) and another, interlocking, character in virtue of being a judgment that p (rather than, say, a judgment that q). Let's label each component for easy reference: I'll discuss both components, as well as the arguments Smithies gives in their support. It'll emerge, as we go, why this conception of belief is important to his overall project in ERO. The arguments can be divided into those that support the full conception (one–track disposition & j–phenomenology), and those that specifically support j–phenomenology. Smithies argues, first, that the conception of beliefs as one-track dispositions to make conscious judgments (best) explains the distinction between beliefs and sub-doxastic states (124–38; §2–3 below). Second, he argues, it explains Accessibilism about justification (148–52, 165–75; §4), which he thinks can be independently motivated (223 ff.). Next he argues that j–phenomenology explains phenomenal contrast cases featuring judgments (140–4; §5) as well as the ‘introspective datum’ that “we can sometimes know by introspection alone which judgments we're currently making” (145–8; §6). The best explanation of this, Smithies argues, is that judgments have phenomenal content and force. (And the best explanation of our introspective knowledge of what we believe is that beliefs are one-track dispositions to make such judgments [148].) In the literature Smithies primarily engages, both beliefs and sub-doxastic states are taken to be representational mental states that figure in the proximate causal-psychological explanation of belief (more generally: attitude) regulation and action. But it's controversial what unifies beliefs and sub-doxastic states respectively, what the principled metaphysical difference between them is, and what the normative difference is (or, some would say: what the normative significance of the metaphysical difference is—but that's not quite how Smithies thinks of the matter).3 On what might be considered ‘the received view’, due to Stephen Stich (1978) and Gareth Evans (1981), beliefs are inferentially integrated and consciously accessible; sub-doxastic states are neither. Smithies, however, argues that conscious accessibility alone grounds the distinction: that what marks off beliefs from sub-doxastic states is just that they're consciously accessible—or, rather: essentially consciously accessible—whereas sub-doxastic states aren't. How does he argue this? First, he argues that conscious accessibility explains the normative distinction between beliefs and sub-doxastic states; specifically: that what grounds the special normative role of belief is that conscious accessibility—glossed as a disposition to make judgments—is fully constitutive of belief. And this, in turn, implies a one-criterial explanation, in terms of conscious accessibility, of the metaphysical distinction too. (This must be the strategy since the argument otherwise begs the question against the two-criterial account.) Some clarifications are needed. Conscious accessibility. Strictly speaking, Smithies says, beliefs themselves can't be brought to consciousness—they're unconscious standing dispositions to make judgments—but their contents are consciously accessible as the contents of the judgments they're dispositions to make. And the fact that being thus disposed is essential to belief is enough for it to meet Smithies’ ‘intermediate criterion’ for conscious accessibility, according to which “an unconscious representational state is accessible to consciousness just in case it is individuated by its dispositions to cause conscious experiences” (131). He grants that sub-doxastic states too can (be disposed to) cause conscious experiences, but denies that it's constitutive of them (131–2). Inferential integration. Smithies says that “our beliefs are inferentially integrated with our other beliefs in the sense that they combine with each other to generate further beliefs in their deductive and inductive consequences, and they're mutually adjusted to avoid logical inconsistency and probabilistic incoherence” (126). In contrast, sub-doxastic states ‘don't combine with beliefs to generate further beliefs’ (or further sub-doxastic states), nor do they combine with sub-doxastic states in other sub-systems to generate further sub-doxastic states (125–6; cf. Stich 1978, 506–7). Smithies notes that beliefs also ‘combine with [a wide range of] desires to produce action’ (126; cf. Evans 1981). And he notes that integration is a relative measure: compared to beliefs, sub-doxastic states combine—or better: tend to combine—with a narrower range of other states (perhaps via a narrower range of inferential or ‘inference-like’ transitions) to generate a narrower range of output. But there's supposed to be a salient threshold which sub-doxastic states fall below, and which—on the received, two-criterial view—coincides with inaccessibility to consciousness. Smithies grants that beliefs are typically inferentially integrated, specifically: with other beliefs, but denies that this is constitutive of belief (128). The normative difference. Smithies operates with the dominant view of the normative difference: beliefs are (epistemically) rationally evaluable—hence capable of being justified/unjustified—and, when justified, capable of conferring justification;4 sub-doxastic states are neither. (Cf., “Beliefs are subject to norms of epistemic rationality, whereas sub-doxastic states are not” [123]; “beliefs can justify other beliefs whereas sub-doxastic states cannot” [131].) Smithies doesn't think that only beliefs can justify other beliefs: conscious perceptual experiences can as well, on his view, despite not being justifiable. But he doesn't consider the possibility that other mental states might fall in this category. Of course he's not alone in that; I just want to mention a different picture—one that I'm tempted by (and develop in my own work). On this picture, beliefs and certain other attitudes are rationally evaluable; these states can confer justification, but only when they're justified in turn. What makes them rationally evaluable is that they're both consciously accessible (although I understand this differently than Smithies),5 and directly revisable—or, more carefully, under our direct deliberative leverage: they're potential unmediated outcomes of first-personal deliberation (in virtue of the functional roles that are characteristic of the kinds of mental state they are). States that lack one or both of these properties aren't capable of being justified, but some can still play a justificatory role—e.g., conscious perceptual experiences, and, more controversially, representational states that are inferentially integrated with our propositional attitudes to a non-negligible degree.6 I also think that there are states like this: mental states that—unlike beliefs—aren't consciously accessible, nor under our direct leverage, but that are significantly integrated with our attitudes. But this is obviously much too brief: just a taste of a different picture. Suppose we grant the dominant view of the normative difference—roughly: that beliefs are justifiable justifiers; sub-doxastic states are neither. Smithies argues that this difference is grounded in a difference in conscious accessibility, understood as explained above: beliefs, but not sub-doxastic states, are fully constituted by dispositions to cause conscious judgments. On the one hand, conscious accessibility is not necessary for inferential integration. Zombies can have unconscious mental representations [that] are inferentially promiscuous. And yet the contents of these representations are not accessible to consciousness as the contents of conscious judgments. On the other hand, conscious accessibility is not sufficient for inferential integration. When patients with Capgras delusion believe their spouse has been replaced by an impostor, the contents of these beliefs are accessible to consciousness […]. At the same time, these ‘monothematic delusions’ tend to be inferentially isolated from other beliefs. (128) In Case 1, “Your representation of rule r is consciously accessible, but it's inferentially insulated from your other beliefs, including the conditional belief that if r is true Chomsky is mistaken. [Here] you're disposed to judge that r is true, but you're not supposed to infer that Chomsky is mistaken. This seems irrational, or anyway less than fully rational.” In Case 2, “Your representation of r is inaccessible to consciousness, but it's inferentially integrated with your other beliefs, including the conditional belief that if r is true Chomsky is mistaken; [here] you're not disposed to judge that r, but you're nevertheless disposed to judge that Chomsky is wrong. Again, this seems irrational, or anyway less than fully rational.” (129–30) From these thought-experiments, Smithies draws the moral that “the normative distinction between beliefs and sub-doxastic states is explained in terms of conscious accessibility rather than inferential integration”, [and that a mental representation] “isn't subject to epistemic norms unless its content is accessible to consciousness as the subject of a conscious judgment” (130). I have some worries about this. The problem cases Smithies invokes are heavily under-described, and it's unclear whether the given descriptions can be elaborated so as to pick out metaphysically possible scenarios. This might not matter too much, by Smithies’ own lights. His stated aim is just to show that the target features are conceptually distinct (and that epistemic evaluability goes with conscious accessibility, in the conceptually possible scenarios where they come apart). By this he can't simply mean that it's logically and analytically/definitionally possible that they come apart, since that's already clear from their independent characterizations. But in other chapters Smithies clarifies that he's neutral on the relationship between conceptual and metaphysical possibility, and that what he means, by saying that p is conceptually possible (or ‘conceivable’) is that p “cannot be ruled out on a priori grounds alone” (Ch. 1, 7–9). That's fair enough, but it limits the argument's import, and invites the response that the two-criterial view of the belief/sub-doxastic state distinction was never put forward—and isn't charitably interpreted—as conceptually necessary, but (at most) as metaphysically necessary. The mere conceivability of philosophical zombies and Capgras syndrome patients of the type Smithies describes doesn't bear on the view, read in the intended way. One might think that the latter example already supports a stronger conclusion (since there are actual patients with Capgras syndrome). According to Smithies, these patients have inferentially insulated but consciously accessible representational states—viz., their impostor-themed delusions. But note that this is likely false of all such patients in the actual world. Inferential integration is compatible with the presence of interference mechanisms—e.g., neuropathological abnormalities, or brute external force—which obstruct the transitions into which an integrated state would otherwise enter. (Otherwise, the suggestion that inferential integration is essential to belief would be a non-starter.) And delusional misidentification syndromes like Capgras are widely, and very plausibly, taken to be rooted in complex impairments of normal physiological and/or psychological functions.7 That's not to say that the notion of an interference mechanism at play (or the normal/abnormal function distinction) is well understood, or that there aren't unclear cases.8 But if the notion has explanatory utility at all, then phenomena like Capgras syndrome are prime candidates for being understood in terms of it. Perhaps there are possible agents without consequential impairments, but otherwise psychologically organized as stipulated by Smithies. But it's hard to see how reflection on this one case, under the bare-bones description he provides, could settle that. I have similar reservations about Case 1: why think that it's possible for your representation of r to be consciously accessible but truly inferentially insulated—rather than ‘cut off’ from normal pathways by interference mechanisms? Note, moreover, that this detail might affect whether the failure to draw the relevant inference seems irrational or not. There are natural ways of imaginatively filling out the case description, involving interference, on which it does—loosely put: interference mechanisms might exculpate, but they don't justify. However, if what we're asked to imagine is that the representational state exhibits the specified inertia in virtue of its proprietary place in the (normal human) overall cognitive architecture—like the traditional paradigms of sub-doxastic states arguably do—the matter is much less clear. I, for one, find it implausible that any of our rational norms (at least those that are ostensibly deontic) mandate operations that we're unable to perform without substantially reconfiguring our minds. This point is independent of the question whether the argument bears on (intended versions of) the account Smithies uses as his foil. Note also that Case 1 in particular—and the verdict that the depicted behavior is irrational—is crucial, for the argument to establish that the disposition to make judgments is wholly constitutive of belief. Smithies does also argue directly against what he takes to be the most promising competing models of belief (134–9; §5). And his explanation of Accessibilism turns on beliefs being one-track dispositions (§6). But the present argument too is supposed to support this aspect of the view. Next: suppose we grant the conceivability of Case 1 & 2, as well as the verdict that the agent's inferential disposition in each case is irrational. How exactly does this show that epistemic evaluability goes with conscious accessibility? Reflection on these cases, Smithies says, shows that “there are consistency (and/or coherence) constraints of epistemic rationality on combinations of beliefs, but not on combinations of beliefs with sub-doxastic states”, and that “[e]pistemic rationality requires your beliefs to be inferentially integrated with each other, but it doesn't require your beliefs to be inferentially integrated with your sub-doxastic states” (129). It's not entirely clear what's going on here. Inferential integration is an architectural causal-functional feature—being governed by, or satisfying, ‘structural’ rationality constraints is not. A state could in principle be integrated—i.e., tend to enter into transitions of the right sort, with a wide range of other states of the right sort—without satisfying any structural constraints (rather: without being a member of a set of states that do), or even being subject to them. Conversely, it could meet such constraints by fluke, divine intervention, or evolutionary design, although it's inferentially insulated. Moreover, structural constraints arguably have considerable independence from other rational norms and constraints, including norms of ‘justified-ness’; they restrict which states can rationally be combined, irrespective of justificatory status.9 So it would be good to see Smithies’ thoughts here spelled out in more detail. How does he understand the relations between inferential integration, structural constraints, and epistemic evaluability? Smithies criticizes some competing accounts of belief—including accounts on which beliefs are ‘multitrack’ dispositions “to cause bodily actions as well as mental acts of judgment” (134). But his critique of those (134–7) doesn't apply to views on which the constitutive features include broadly inferential, rather than overtly behavioral, dispositions. Smithies doesn't say much about such views—with the exception of the two-criterial account, on which inferential integration is among those features. But it's plausible that the overall functional role that constitutes belief (also) includes more specific inferential dispositions: dispositions to be created, revised, and eliminated in distinctive ways—e.g., in response to (representations of) considerations that pertain to the truth or falsity of the given content (rather than, say, the desirability or utility of its being true). Smithies rightly points out that “beliefs and judgments aren't perfectly sensitive to evidence” (137). I'm not suggesting that. But it strikes me as very implausible that someone could count as believing (rather than, say, wanting or hoping) that p without having even the slightest tendency to (re)adjust their stand on whether p in light of considerations of the kind that qualifies as evidence or epistemic reasons—at least absent interference. It's hard to spell out in informative terms what this comes to, and to pinpoint the limiting case, but I suspect I'm more optimistic than Smithies is that it can be done.10 On another view, no disposition—inferential or otherwise—is individually necessary for belief: rather, beliefs are clusters of dispositions. On one version of this view, certain disjunctions of dispositions are still essential (e.g., either the disposition to make judgments, or the disposition to be regulated in broadly ‘truth-aiming’ ways), but some minimum number of additional dispositions in the cluster are also required. On another version, any dispositions in it will do, provided it's sufficiently many. On a hybrid view, some specific dispositions are indeed essential, but a sufficiently large subset of further dispositions in the cluster are also needed. The hybrid view is particularly attractive to me, but it would be good to hear what Smithies thinks is wrong with any of these competitors to his account. one–track disposition also figures in Smithies’ explanation of Accessibilism. This is the thesis that certain facts about propositional justification—e.g., whether you have justification to believe that p—“are strongly luminous in the sense that [necessarily] you're always in a position to know whether or not they obtain” (226). Smithies argues that this is true only because those facts are determined by (facts about) mental states that are strongly introspectively luminous: you're always in a position to know, through introspection, whether or not you're in these states (148–9). This, in turn, is true only because all justification-determining states are phenomenally individuated: wholly individuated either by their phenomenal characters, or by their dispositions to token states with phenomenal characters. How does that ensure introspective luminosity? On Smithies’ account—the ‘Simple Theory’—of introspection, phenomenally individuated states are such that the very fact that you're in them makes it evidentially certain to you that you are (155–8): the fact that, e.g., I believe that p is an ‘immediate, infallible, indubitable, indefeasible, and safe’ (introspective) reason for me to believe that I believe that p, since beliefs are one-track dispositions to make phenomenally conscious judgments. I don't quite understand Smithies’ explanation of Accessibilism, since I don't understand the Simple Theory of introspection. However, if the only feature of beliefs that our (allegedly infallible, etc.) introspective reasons are tracking is the disposition to make judgments, then yes: the introspective luminosity thesis requires this disposition to be (not just necessary but) sufficient for belief. It also requires that judgments have phenomenal character/s. So let's look a little closer at that claim, j–phenomenology, before returning to the Simple Theory. To recap: according to Smithies, only phenomenally individuated states can provide justification; beliefs fall in this category because they're individuated by dispositions to cause phenomenally conscious judgments. The fact that judgments have distinctive, attitude- and content-specific, phenomenology also explains why ‘we can sometimes know by introspection alone’ what we're judging, or what we believe—and, indeed, why judgment and belief are strongly introspectively luminous (given that the relevant ‘cognitive experience’ is wholly constitutive of judgment, and the disposition to cause such experiences is wholly constitutive of belief), which is required for Accessibilism to be true. j–phenomenology is thus integral to Smithies’ explanation and vindication of Accessibilism. Suppose […] that two subjects entertain the same proposition, but one of them judges the proposition to be true, while the other suspends judgment. This makes a phenomenal difference because judgment has phenomenal force. Alternatively, suppose that two subjects entertain the same proposition, but one of them judges that it's true, while the other judges that it's false. This makes a phenomenal difference because judgment has phenomenal content. (141) Smithies expects his readers to agree that the difference in judgment in each example would or might make a phenomenal difference. (“Surely making a judgment sometimes makes a difference causally, if not constitutively, to the phenomenal character of your experience” [141].) He proceeds to criticize accounts of that difference in terms of different sensory experiences, and concludes that j–phenomenology provides the best explanation (141–4). Here's what I grant: that making a judgment sometimes makes an ostensibly causal difference to the phenomenal character of our overall experience. What, if any, such difference it makes partly depends on what's being judged—not because judgments have content-specific phenomenology, but because it can be, e.g., upsetting or gratifying or surprising to judge certain contents, and totally banal to judge others. Judgments can trigger various states or processes that do have characteristic, or at least typical, ‘feels’—e.g., angst or surprise or self-directed satisfaction, relief of puzzlement, feelings of subjective confidence or understanding.11 But these are all distinct psychological phenomena, some of which may be regularly associated with making judgments (for certain subjects, in certain contexts). As far as I can tell, there's nothing ‘it's like’ to make the judgments themselves, in isolation from any such concomitant phenomena. Presumably Smithies has some other phenomenal contrast in mind than that between, say, judging that p with a feeling of understanding or relief, and not judging that p with that feeling. But what is that contrast? I'm afraid I fail to recognize it in my own experience or imagination—just like I fail to recognize a particular phenomenal character therein that correlates with all-and-only judgments (or all-and-only judgments that it's raining, all-and-only judgments that children show scars as medals, etc.) More guidance—perhaps including some identifying details that differentiate the target phenomenology from those in close vicinity—would be helpful. There's a puzzling dialectical standoff at precisely this point, in the larger debate about cognitive phenomenology to which Smithies contributes.12 Some participants in that debate claim to be able to introspectively isolate the target phenomenal character—of, say, judgments, thoughts, or so-called ‘intellectual seemings’—in their experience. Others claim to be unable to do so. To those of us in the latter camp, the search turns up nothing—at any rate, nothing that's not plausibly attributed to various familiar, broadly emotional, states which sometimes accompany the target state or process. One diagnosis of the disparate self-reports is that only some humans enjoy cognitive phenomenology (of the target type/s). This can't be ruled out from the armchair, but it would be quite surprising. Another possibility is that subjects like me systematically miss, or misidentify, the target phenomenal character; alternatively, that those in the opposing camp systematically hallucinate and/or misidentify some other character as the target. This would be surprising too, but perhaps more tolerable—except on very stringent views about the epistemic security of introspection (like Smithies’). I'll end by addressing the role of one–track disposition & j–phenomenology in Smithies’ explanation of the datum that we sometimes know by introspection alone what we're judging, and of introspective luminosity (hence of Accessibilism). The best explanation of the datum, Smithies argues, requires j–phenomenology, and the best explanation of how we sometimes know by introspection alone what we believe, requires one–track disposition (145, 148). The explanations take the form of the Simple Theory, on which phenomenally individuated states necessarily ‘provide introspective reasons for beliefs about themselves’ (155); on this view, the fact that you're in some such state is itself a (maximally secure) reason to believe that you are. Given certain further assumptions (156–7), the Simple Theory underwrites the claim that phenomenally individuated states are strongly introspectively luminous. I'm unclear on what it is for the fact that you're in a state to be a reason to believe that you are. What makes this fact—any such fact—a reason to believe or do anything? I'm also unclear on why (all-and-only) phenomenally individuated states—including standing unconscious states, like beliefs, which are individuated by their phenomenal dispositions—provide such reasons? Relatedly: it's unclear what explanatory work j–phenomenology is doing, in the account of how we sometimes know by introspection alone what we believe, and of the introspective luminosity of belief. Why is j–phenomenology required—or even relevant—to explaining these things when, in Smithies’ words, “you can have standing introspective knowledge of beliefs when you're not making any judgment at all”, and “[t]he disposition to judge that p is both necessary and sufficient for having introspective justification to believe that you believe that p, although the act of judgment itself is neither necessary nor sufficient.” (180). To rephrase: the mere fact that I have the belief that p—i.e., the disposition to judge that p—is supposed to be an introspective reason for me to believe that I have this belief. What gives it that status? The fact that I have the disposition to fall over if someone shoots me in the head isn't a reason (of any kind) for me to believe I'm thus disposed. What explains the difference? And what work is j–phenomenology doing in that explanation, and in the explanation of the introspective luminosity of belief? It would be great to get some clarity on this too.