<i>Why We Doubt: A Cognitive Account of Our Skeptical Inclinations</i>

In Why We Doubt: A Cognitive Account of Our Skeptical Inclinations, Ángel Pinillos makes a strong case for the importance of skeptical doubt, not merely to philosophy but to matters of great practical importance. Contrary to the views of Hume and Peirce, skeptical doubt can arise outside the study; it need not be “fictitious” or “paper” (viii). When it arises, it can have an impact on our willingness to act, individually and collectively. Sometimes its effects are deleterious, such as leading one to think creatively and “outside the box.” But it can lead us astray, such as when it is encouraged, honestly or dishonestly, by those peddling conspiracy theories.If Pinillos is right about the importance of skeptical doubt, which he plausibly is, we should try to give an account of it as a real-world phenomenon. What Pinillos attempts to provide is a cognitive account. Following David Marr, he takes this to mean providing three levels of description: algorithmic, functional, and implementational. Chiefly his concern is with the first two levels. There is a place for philosophy in this project. Philosophers have amassed a rich set of data concerning when skeptical doubt arises, which can be used to test proposals about the relevant algorithm, and they have sketched candidate algorithms themselves. Moreover, once we uncover the relevant algorithm, we can ask a range of normative questions about it, which philosophy can help us answer: Is the algorithm rational to follow? How does it relate to genuine epistemic norms?Pinillos is not attempting to explain all doubt, only skeptical. He glosses this as inclination to deny that we know things, where the denial is in “rational tension with some of our ordinary commitments” and where the inclination arises from a “special source” that involves consideration of “a scenario in which there is a possibility of error” (1). Later in the book, epistemic feelings play an important role as well. One might wonder whether this gloss encompasses Pyrrhonian doubt or suspension, which needn’t be associated with considering particular possibilities of error. But Pinillos is on firm ground in seeing reflection on particular error-possibilities as a fecund source of doubt, capable of inclining us toward a variety of skepticisms.What algorithm leads us to skeptical doubt? To answer this, we shouldn’t look to introspection (compare the limits of introspection to tell us about how we arrive at a “feeling of knowing”). The algorithm is a “black box,” which we crack into only using experiment and inference. Yet Pinillos’s proposal in chapter 2 won’t come as a shock to the cognoscenti. Skeptical doubt arises when we have a belief that p, take seriously a certain possibility in which not-p, and apply the following principal basis sensitivity (PBS) heuristic: we check to see if our “principal basis” for the belief that p is sensitive to p—that is, whether, if p were false, this basis would be in place; if we judge it insensitive, we have an epistemic feeling as if something is wrong with our belief and become inclined to think we don’t know p (65). Sensitivity is the key to skeptical doubt—principal-base sensitivity, not Nozickian belief-sensitivity.To see how the PBS algorithm is supposed to work, consider a case in which I believe my car is parked in the lot and believe, at least implicitly, that it hasn’t been stolen. Suppose now I take seriously the possibility that it has been stolen. I identify my principal basis: that I parked my car in the lot. I then check for sensitivity. If my car were stolen, would it still be the case that I parked in the lot? Yes. This verdict results in doubt—a feeling that something is wrong and an accompanying inclination to think I don’t know it hasn’t been stolen. The doubt can then spread to affect further appropriately related beliefs, such as the belief that my car is in the lot.This is an example of “local” skepticism. The account is more complicated for “global” skepticism. Consider applying the PBS heuristic to my belief that I am not a brain in a vat (BIV). The hard part is figuring out my principal basis. A principal basis is a rational ground that is also the “main cause” of one’s belief. Pinillos thinks we go through a “trial and error” procedure here (83). We try out different candidate bases and find that they are all insensitive to our not being a BIV, and then conclude that all candidate principal bases are insensitive. Doubt ensues.For the BIV case, it’s important that we don’t see our principal basis as “externally individuated”—for example, a fact about the world outside us, such as I parked the car in the lot. If I found my basis for my belief that I’m not a BIV to be I live in St. Louis with my family, then PBS wouldn’t produce doubt, since I can see that this basis is sensitive to I’m not a BIV. Pinillos argues that in determining our principal basis, we consider what reasoning would be good, and we judge this sort of (Moorean) reasoning isn’t good (70). It would be interesting to test this account empirically. Do we go through this sort of trial and error and then generalize when we consider the BIV hypothesis and experience doubt? (Notice that the “belief-sensitivity” account avoids postulating this sort of trial and error. We simply can see that if we were BIVs, we would still think we weren’t [in the same apparent way].)Pinillos’s account of our “internalizing tendency” in judging our bases might seem to give us a path to skeptical doubt, irrespective of sensitivity. If thinking about BIVs gets to me to think that my principal bases for believing I’m not a BIV and for my perceptual beliefs are merely my experiences and apparent memories, then I’m well on the way to skepticism, before any question of sensitivity arises. This is not a problem for Pinillos, so long as he has a good cognitive account of our tendency, when thinking of global skeptical possibilities, to take, perhaps inaccurately, our principal bases to be internal. In light of Hugo Mercier and Dan Sperber’s account according to which the function of giving reasons is persuasion, Pinillos takes it as unsurprising that our assessments of good reasoning reflect this function: if it wouldn’t persuade the relevant opponent, we treat it as poor reasoning (87). This is a good start, but more is needed. When we consider the flat earth conspiracy theory, we presumably don’t have the same tendency to see our bases as limited to things compatible with what flat-earthers believe.There are familiar counterexamples to evidence-sensitivity as a necessary condition for knowledge. Pinillos takes a subtle but reasonable line on this issue in chapter 12. On the one hand, since he is giving a cognitive account and not proposing a necessary condition on knowledge, he can (and does) grant that sensitivity isn’t required for knowledge. On the other, since he’s invoking the PBS rule to explain our skeptical doubts, he needs to argue that PBS-based predictions of doubt are generally correct.Consider the fifty students in my logic class. Suppose I take seriously the possibility that all of them get an A on the final exam. My principal basis for believing they won’t all get A’s concerns their abilities and past performance. In the peculiar event that all my students were to get A’s on the final exam, I would have this same basis for believing they won’t all get A’s. So it would seem that PBS would predict that I’ll be inclined to think I don’t know. But no such thing happens. Discussing a similar example of Vogel’s, Pinillos claims that we go in for backtracking reasoning here, which vindicates sensitivity (247). But we can slightly alter the possibility: all fifty students get A’s on the final exam, despite the mediocre abilities of many of them, and without cheating. If this were the case, plausibly, I would have the same relevant evidence. But I’m still not inclined to think I don’t know they won’t all get A’s. Or consider the proposition that my mothers’ decades of expressions of love to me are all insincere. If this were the case, my principal basis—those expressions of love—would be in place. Yet I’m not tempted to think I don’t know her declarations are sincere. One wonders if more than mere insensitivity is needed in the black box—for example, whether explanatory considerations matter.1Suppose PBS is the algorithm. What’s its function? Pinillos describes a broad and a narrow function. The broad function, discussed in chapters 3–4, is a metacognitive one of a familiar kind: to monitor and control our first-level reasoning. PBS produces a metacognitive skeptical feeling that helps us flag possible problems so that we can take steps to avoid them. Here Pinillos’s commitment to the idea that knowing p suffices for the appropriateness of using p as a premise in reasoning plays an important role. Inclined to think we don’t know, we might stop using p in reasoning, in a way that is perfectly rationalizable (and doesn’t require a black box).The narrow function of PBS, discussed in chapter 5, is to approximate a correct but hard-to-use norm. PBS is a shortcut, taking its place alongside other shortcuts posited in psychology, such as the availability, representativeness, and the “take the best” heuristics. Pinillos’s account is as follows. Moving from the premise that one’s principal basis E is insensitive to P to the conclusion that one doesn’t know P—is a user-friendly way of approximating the following correct norm: If (i) you know for certain that E is insensitive to P, and (ii) you don’t yet have rational credence sufficient for knowing P, then gaining only E as evidence will not give you knowledge that P. (138)(This in turn is grounded in a Bayesian norm that replaces [i] with “if Crinit(E|∼P)=1” and [ii] with a more precise rendering in terms of a rational credence falling below a knowledge-marker.) Pinillos’s main point is that PBS ignores our prior rational confidence in omitting anything like (ii); this exposes us to error in the relatively rare cases in which (ii) doesn’t hold.This story applies nicely to many cases of skeptical doubt. For I am not a BIV and my car was not stolen away, clause (i) of the norm applies, but we ignore the fact that our prior credence was already good enough for knowledge, and so embracing the doubt PBS generates leads us to error. But the account runs into problems with cases of nakedly statistical evidence. In these cases, (i) isn’t satisfied.Suppose I have a ticket in a lottery I know to be fair and to have only one winner. I don’t know much at all about its size. You tell me you saw the announcement and you seem to remember that the winning number ends in an 8 (or maybe it was a 6). Mine ends in a 2. This information isn’t enough for knowledge that my ticket lost, it seems. We wouldn’t want to accept a norm that tells us that my rational credence in this case can’t rise substantially and even become enough for knowledge, if I learn the lottery has ten million tickets. The norm Pinillos cites might seem to imply otherwise, but it doesn’t. Prior to getting the evidence that there are ten million tickets, I don’t know with certainty—nor do I even have a high rational confidence—that the ticket’s being one of ten million isn’t sensitive to my ticket will lose. Clause (i) doesn’t apply. For I don’t know—at the earlier time—that if my ticket were to win, my ticket would be one of ten million. I only know that, whatever the size of the lottery is, if my ticket were to win, the lottery would be of that size. This suggests another way PBS oversteps the correct norm, in addition to ignoring prior rational credence: in cases involving statistical evidence, we treat our current knowledge of insensitivity as if it were in place before we aquired the evidence.2 This suggestion seems available to Pinillos.I’ve given only the barest summary of this rich book. I’ve not touched on the illuminating chapters in which Pinillos applies his account to conspiracy theories and the doubt associated with obsessive-compulsive disorder. All in all, Pinillos’s book is a major contribution to empirically informed epistemology and a model for its philosophically sophisticated use of ideas and results from cognitive science to help address long-standing philosophical problems.I thank Ángel Pinillos and Allan Hazlett for helpful comments on a draft of this review.