![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
ABSTRACT
A distinctive feature of Juan Comesaña's epistemological account is the possibility of an agent possessing a false proposition as evidence. Comesaña argues that there are a number of theoretical virtues of his account once we accept this possibility, however, one might expect that there are particular vices of his account as well. Littlejohn and Dutant (2021. “Even if it Might not be True, Evidence Cannot be False.” Philosophical Studies 179 (3): 801-827. https://doi.org/10.1007/s11098-021-01695-0) claim that a reflective agent who accepts Comesaña's view is rationally compelled to update their credences differently than unreflective agents, or else they will be guilty of a problematic kind of doublethink: evidence in general can be false, but my evidence is never false. I argue that parallel reasoning applies to any epistemological view that says that it is possible for a rational agent to be wrong about what their evidence is. When a rational agent reflectively accepts that they could be wrong about their evidence they are rationally compelled to update their credences differently than unreflective agents, or else they will be guilty of a problematic kind of doublethink: even though rational agents in general are wrong about their evidence, I'm never wrong about my evidence. Reflecting on a particular challenge to Comesaña's view ultimately reveals a general challenge to fallibilism.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 More precisely, the view is that the degree of epistemic support for any proposition relative to an agent is determined by an objective evidential probability function conditional on the agent’s evidence. For Comesaña, one’s evidence is all and only the propositions one is basically justified in believing. For Williamson, one’s evidence is all and only the propositions one knows.
2 In Littlejohn and Dutant (Citation2021), they present three arguments against Experientialism. The first argues that Experientialism implies that one’s evidence could be inconsistent (see also the discussion in Comesaña Citation2020a, §6.6) and that this is a serious defect of the view. The second argues that Experientialism implies that reflective agents ought to update their credences in a different way than unreflective agents. The third argues against the purported motivation for having one’s evidence be constituted by worldly propositions instead of appearance propositions. I only address the second argument in this paper, so even if I am correct that their second argument does not single out a unique difficulty for Experientialism, this would not undermine their other arguments against Experientialism.
3 Strictly speaking, Comesaña does not accept that evidence can be false, but rather that we can have false evidence. The problem for Littlejohn & Dutant, then, is about an agent who comes to believe that it is possible that a proposition P is part of their evidence yet P is false. See Comesaña (Citation2020a; Citation2023) for clarifications on this subtlety.
4 I think White (Citation2023) pithily hits on this point when he says that what ‘Comesaña calls false evidence the rest of us might call counterfeit evidence – non-evidence which is perhaps innocently mistaken for evidence.’
5 The use of ‘fallibilism’ and ‘infallibilism’ varies in the literature, and my usage differs from other discussions of infallibilism. For example, sometimes it is used to denote the view that knowledge requires infallible belief (e.g. Dutant Citation2016), while other times it is used to denote the view that knowledge that P requires that one’s evidence entails P (e.g. Brown Citation2018). According to my usage, fallibilism is the view that no doxastic state is so secure that it is not open to coherent doubt, so even a Factivist would be a fallibilist in this sense due to the fact that they reject KK and luminosity. Importantly, this implies that beliefs about one’s evidence are always open to coherent doubt. Difficulties in delineating what counts as a fallibilist view are not new. For instance, some contextualists about knowledge consider their view to be a version of fallibilism (e.g. Cohen Citation1988), while others describe their view as a version of infallibilism (e.g. Lewis Citation1996).
6 I suspect that this issue is also connected to the problem of exogenous defeaters (see Weisberg Citation2009; Citation2015 and Comesaña Citation2020a; Citation2020b) which also seems like it is a problem for everyone, but I do not get into that issue in this paper.
7 Note that an appearance proposition is about a mental state, not about an external object. So that the jelly bean looks red is about the agent’s mental state, not about the jelly bean.
8 Huemer (Citation2001).
9 This is the version of Factivism argued for in Williamson (Citation2000), however, fully fleshed out versions of Factivism can differ in what they will say about Jim in S2 and S3 (see, e.g. Williamson Citationforthcoming a; Littlejohn Citation2012; Citationforthcoming; Lasonen-Aarnio Citationforthcoming). The strongest version of Factivism would be akin to the kind of epistemological disjunctivism defended by McDowell (Citation1998a; Citation1998b; Citation1998c) and would say that in S2 and S3 it is not that Jim’s evidence is constituted by appearance propositions. Rather, Jim has no evidence regarding the color of the jelly bean – he has no justification for believing incorrectly, only an excuse for believing incorrectly. Regardless of which flavor of Factivism one prefers, my arguments in this paper should still apply.
10 See Comesaña and McGrath (Citation2016) for an extended defense of these claims. See also Littlejohn and Dutant (Citation2021) as one of their lines of criticism that I do not address in this paper takes aim at these claims from Comesaña & McGrath.
11 Here I am assuming what Comesaña (Citation2020a) calls Fumerton’s Thesis: rational action presupposes rational belief.
12 Indeed, the Factivist seems to be forced to say that in situations like this Jim would be irrational no matter what he believed and no matter what he did (see Comesaña 2020a, ch. 4).
13 This is not how Littlejohn & Dutant structure their argument, but I think this reconstruction makes salient what their objection to Experientialism amounts to.
14 I have added this extra step to Littlejohn & Dutant’s argument because in their example they use appearance propositions. Yet a key aspect of Comesaña’s view is that, absent any defeaters, one’s evidence is not constituted by appearance propositions. So I do not think that we should model Jim as updating on appearance propositions from the outset in an argument against Comesaña’s view.
15 Appearance propositions, by their very nature, do not rule out counter-possibilities. If it appears that R, this leaves open that it is not R; using appearance propositions in one's reasoning or inferences involves giving credence to counter-possibilities. Using the pair of propositions RT and RF seems to serve the same purpose of giving credence to counter-possibilities, which is why I say the use of RT and RF is analogous to the use of appearance propositions.
16 Here I am ignoring the complication of borderline cases, which would be a situation where it looks red and it is red, but the belief that it is red is too unsafe to count as knowledge.
17 I got the term ‘counterfeit evidence’ from White (Citation2023).
18 Note that this seems to imply that Comesaña has two problems since one could rerun my reasoning on Comesaña’s view. However, this does not undermine my claim that the reasoning from Littlejohn & Dutant, if taken seriously, generalizes to fallibilism in general.
19 Or between any agents for that matter, though there is a sense in which the demands of rationality may take different forms for different creatures. Our eyes and ears are often sources of basic justification, but less so our noses; yet a beagle’s nose will be a primary source of basic justification for them. We do not naturally have a sense for magnetic fields, but perhaps futuristic body modifications will allow a sensitivity to magnetic fields which will provide people with basic justification. In each of these scenarios, the demands of rationality are the same, even though the shape of those demands may differ.
20 Littlejohn and Dutant (Citation2021, 819).
21 A Factivist pushing back on the legitimacy of this parallel would strike me as particularly odd once we consider what Factivism sometimes demands of reflective agents. Littlejohn & Dutant suggest that it is unreasonable for an Experientialist to ignore the possibility that their evidence will be false, yet Factivism has an even more counterintuitive conclusion. It says there are cases where an agent ought to ignore the overwhelming probability that their evidence is not-E, and conditionalize on E (Williamson Citation2014). (Or, more precisely since Williamson would object to conceiving of his view as involving acts of conditionalizing (see Williamson Citationforthcoming b b), the agent knows that it is almost certainly the case that their epistemic support for propositions is not determined by E, however, the degree of epistemic support is determined by E. So the agent’s doxastic attitudes ought to match the degree of epistemic support provided by E, even though they know that it is almost certainly the case that such matching would be irrational).
For example, consider again Tim, our reflective Factivist. Suppose that Tim takes a break from rereading Knowledge and Its Limits, and looks at a particularly austere clock. Using some formal tools he learned from Jaakko Hintikka, he rationally concludes that it is .99999 probable that his evidence is not-E. Yet unbeknownst to Tim he knows E. Surely, it would be reasonable, given that it is .9999 probable that his evidence is not-E, for Tim not to update on what he unknowingly knows. Yet the Factivist says that it would be unreasonable for Tim not to conditionalize on E. And not just unreasonable, but downright irrational. So I think that it would be odd to maintain that it is unreasonable to ignore the mere possibility of false evidence, but that it is reasonable to ignore the overwhelming probability of true evidence.
This footnote may appear to be belaboring a mere tu quoque, but it is not. The objection I was responding to was of the form: ‘it is more plausible to think that the combination of Experientialism and Changing Demands leads to implausible consequences when compared to the combination of Factivism and Changing Demands.’ The considerations in this footnote, I think, cast doubt on this. The Factivist might point out that they have learned to live with the variety of counterintuitive implications of their view, but the point here is not simply to point out a counterintuitive implication. Rather, it is to show that from a theory-neutral standpoint, one cannot reasonably hold that the justification for (E2) is more compelling than the justification for (K2).
22 This attitude certainly sounds bad, and I will give a more charitable interpretation of it in the next section. But even setting that aside, I think there is a precedent for it in the philosophical literature. An analogous position has been defended in the disagreement literature. Permissivists argue that there can be permissive cases where two people who have the same body of evidence can rationally disagree on a proposition. A difficulty that permissivists face is that it is hard to see how someone who accepts permissivism can justify maintaining their confidence in a proposition if they come to believe that they are in a permissive case (see White Citation2005). While some permissivists have tried to explain how one could both rationally believe that they are in a permissive case and rationally hold on to their belief (e.g. Schoenfield Citation2014), other permissivists have argued that one is never justified in believing that one is in a permissive case even though one believes that permissive cases are possible (see Smith Citation2020).
23 A similar point is made in Schoenfield (Citation2017). She uses this point, plus considerations of accuracy, to argue in favor of a modified form of conditionalization.
24 Consider Schoenfield’s (Citation2017) argument that conditionalization does not (in general) maximize expected accuracy, but rather accuracy is maximized by a modification of conditionalization that she calls conditionalization*. Interestingly, Schoenfield concludes that thinking that rational agents maximize accuracy goes hand in hand with thinking that there exists a privileged, luminous class of propositions relative to every rational agent. Furthermore, she is motivated by an aversion to an attitude very similar to one described in this paper: ‘every rational agent must antecedently be certain that any proposition P that could be true of her experience (and which it is possible to learn about) is a proposition that she will learn exogenously whenever P is true and that there are no other propositions that she could exogenously learn’ (p. 1167). So the considerations of this paper would not be the first to suggest that, perhaps surprisingly, accepting fallibilism in general seems to commit one to accepting an unreasonably high level of confidence that, in spite of my fallibilities, things will not go astray for me in the future. In this paper, I am suggesting that we should accept this seemingly unreasonable attitude, whereas Schoenfield suggests that we ought to give up on fallibilism – or, at least, that we give up on a kind of thoroughgoing fallibilism that would reject what she calls Luminous Infallibility: ‘There is a class of propositions concerning an agent’s situation, such that, for any agent S, if S is rational, these propositions will be true of S if and only if she is certain of them’ (p. 1180).
25 This may be surprising given that their disagreement on certain epistemological issues runs deep, but they do share many of the same features, including: evidence is propositional, worldly propositions can be evidence, an agent can permissibly rule out ∼P possibilities if and only if P is part of their evidence, the degree of epistemic support for a proposition is determined by an objective evidential probability function, and rational agents are not luminous with respect to their evidence.
26 Thanks to Juan Comesaña and Timothy Kearl for helpful feedback.