What kind of reason does incoherence provide?

Abstract

In this commentary, I raise a few questions about Schmidt’s argument against (R-E): whether facts about incoherence are directly reasons for suspension on particular propositions, as opposed to reasons against sets of attitudes; whether (R-E) should really be formulated in terms of a broad category of “doxastic attitudes” that includes transitional attitudes like suspension; and whether incoherence-based reasons really must fit into the category of “epistemic reasons,” as opposed to be a more general category of right-kind reasons. Though my questions reflect some skepticism about the specifics of Schmidt’s argument, I conclude that it succeeds in what I take to be its broader aim.

1 Introduction

Thanks to Eva Schmidt and an anonymous reviewer for helpful feedback on an earlier draft of this commentary.

In “Facts about incoherence as non-evidential epistemic reasons,” Eva Schmidt argues against the following principle:

(R-E) All epistemic reasons for doxastic attitudes towards p are provided by evidence concerning p.

Schmidt argues against (R-E) by counterexample. The counterexamples she provides involve cases where facts about the incoherence of a set of doxastic attitudes constitute reasons to suspend on all the (dis) believed propositions in the incoherent set. She contends that these incoherence-based reasons to suspend are properly epistemic reasons for doxastic attitudes towards some proposition p, and yet are not constituted by evidence concerning p. It follows that (R-E) is false. After rejecting (R-E), Schmidt briefly suggests as its replacement a view on which “what makes epistemic reasons specifically epistemic is that they guide the subject in appropriately regulating her intellectual conduct, that is, with a view to the truth” (Schmidt 2023, 22).

Schmidt’s argument is interesting and original, and addresses an important issue in normative epistemology that has not received enough attention. There is significant debate about whether practical considerations can be genuine normative reasons for doxastic attitudes, or only properly epistemic considerations. Something like (R-E) is often assumed in such debate. As Schmidt notes, simply assuming (R-E) ignores questions about incoherence-based reasons for doxastic attitudes. But before considering the substantive merits of Schmidt’s argument, it will be necessary to clarify the target of the argument. By “epistemic reasons,” Schmidt means “considerations that favor a subject’s doxastic attitudes in an epistemic rather than a practical manner” (2). An epistemic manner of favoring a doxastic attitude towards p has something to do with a consideration’s relationship to the truth of p, as opposed to practical considerations, which bear on whether the attitude is good or useful, and have no particular relationship to the truth of p.

In this commentary, I’ll raise a few questions about Schmidt’s argument against (R-E): whether facts about incoherence are directly reasons for suspension on particular propositions, as opposed to reasons against sets of attitudes (§1); whether (R-E) should really be formulated in terms of a broad category of “doxastic attitudes” that includes transitional attitudes like suspension (§2); and whether incoherence-based reasons really must fit into the category of “epistemic reasons,” as opposed to be a more general category of right-kind reasons (§3). Though my questions reflect some skepticism about the specifics of Schmidt’s argument, I conclude that it succeeds in what I take to be its broader aim.

2 What are facts about incoherence reasons for?

Schmidt’s counterexamples to (R-E) are all cases involving incoherent sets of doxastic attitudes. She doesn’t commit to any particular account of incoherence, but characterizes incoherent sets of attitudes as ones where it is irrational to hold all of the attitudes together.² She offers three counterexamples: 6/49 Lottery, Marple and Poirot, and History vs. Philosophy. In 6/49 Lottery, Lola believes of each individual lottery ticket that it will lose but also believes that it is not the case that every ticket will lose. Each of these beliefs seems to be supported by Lola’s evidence. But they are jointly inconsistent, and so incoherent. Schmidt argues that this fact constitutes a reason for Lola to suspend on each of the beliefs in the incoherent set, even though it is not evidence regarding any of the believed propositions.

Marple and Poirot is a case of misleading higher-order evidence. Marple, whom Poirot knows to be a reliable testifier, has told Poirot that the evidence indicates that the vicar did it (v). However, on the basis of his own assessment of the evidence, Poirot disbelieves v. So, Poirot both believes the evidence indicates v and disbelieves v, which is an incoherent set of doxastic attitudes. Here again, the thrust of the counterexample is that each attitude in the incoherent set is supported by Poirot’s evidence, and yet the fact that the set is incoherent constitutes a reason for Poirot to suspend on each proposition. These first two cases are sufficient for my purposes, so I’ll set the third case, History vs. Philosophy, aside.

It’s crucial to Schmidt’s contention that these cases are counterexamples to (R-E) that they have the following structure: the subject holds an incoherent set of (dis)beliefs. For each individual attitude in the incoherent set, the fact that the set is incoherent constitutes an epistemic reason to suspend on all of the relevant propositions, even though it isn’t evidence for or against any of those propositions. Thus, facts about incoherence are epistemic reasons for doxastic attitudes towards p that are not provided by evidence concerning p.

I won’t dispute Schmidt’s claim that facts about incoherence constitute reasons. Instead, I’ll raise a worry about what these facts constitute reasons for. Consider 6/49 Lottery. While Schmidt describes the fact that Lola’s beliefs are incoherent as if it is a single reason to suspend on all the propositions in the set, it can’t really be a single reason on her view. Strictly speaking, the incoherence fact must constitute many reasons: a reason to suspend on p₁, a reason to suspend on p_2, a reason to suspend on p₃, and so on for each proposition in the set. This must be the case, on Schmidt’s view, because otherwise, incoherence facts wouldn’t constitute reasons for doxastic attitudes towards p, where p stands in for any proposition the subject currently (dis) believes in the incoherent set.

But does the incoherence fact really constitute a discrete reason for an individual’s attitude of suspension towards each of the propositions in question? Here’s one worry about going this way. Consider Marple and Poirot. On Schmidt’s view, it seems the fact that this set is incoherent must constitute two discrete reasons: a reason to suspend on v, and a reason to suspend on [the evidence indicates v]. And according to Schmidt, these reasons outweigh Poirot’s evidential reasons to believe each proposition. It follows that Poirot would have the attitude toward v he epistemically ought to have if he suspended on v, and would have the attitude toward [the evidence indicates v] he epistemically ought to have if he suspended on [the evidence indicates v]. If he suspended on one proposition but not the other, he would have the attitude he epistemically ought to have towards that proposition, but not the other.

But would responding to one of these reasons but not the other really constitute a response to the relevant incoherence fact in this case? Thinking of the incoherence fact as constituting multiple discrete reasons, each for suspending on a different proposition, seems to decompose the normative force of incoherence in a way that loses what is distinctive about it. Intuitively, the incoherence fact constitutes not a collection of discrete reasons for individual doxastic attitudes towards the particular propositions in question, but rather a single reason against the incoherent set of attitudes, constituted by a holistic assessment of the relations between them. In other words, the incoherence fact seems to constitute a reason against a whole set of attitudes rather than a set of reasons for individual attitudes. But if that’s right, then incoherence-based reasons are not actually reasons for doxastic attitudes towards p, in the sense referred to by (R-E).

This may be why it’s so natural to describe the incoherence fact as constituting a single reason. Now, perhaps it’s still true that, in the absence of any assessment of which attitude in the incoherent set has gone wrong, the only way of complying with the incoherence-based reason is to suspend on each attitude, as Schmidt puts it, “with an eye to figuring out how to resolve the tension between [one’s] attitudes” (10). But those reasons to suspend on each attitude seem to be derivative of a more fundamental reason against holding an incoherent set of attitudes, given the background fact that one has no assessment of which attitude has gone wrong. Thus, in cases where this background fact doesn’t obtain, there won’t be reasons to suspend on all of the attitudes.

To illustrate this, imagine Poirot instead knows Marple to be a very unreliable testifier with regard to what the evidence indicates, and yet believes her anyway. It still seems right to say that Poirot has an incoherence-based reason not to hold the attitude set {belief [the evidence indicates v], disbelief[v]}. But now it seems quite odd to insist that Poirot has a reason to suspend on v and a reason to suspend on [the evidence indicates v], each of which are equally weighty. Given that Poirot clearly epistemically ought to give up his belief that the evidence indicates v, it seems he can comply perfectly well with the reason against the incoherent attitude set that way, without having to suspend on v. In fact, it seems he epistemically ought not to suspend on v, given that v is supported by his evidence. Thus, it isn’t clear that facts about incoherence generate reasons to suspend on all of the attitudes in the incoherent set. Instead, it may be that they provide reasons against holding the entire set of attitudes, which can sometimes be followed by pinpointing the defective attitude and giving up only that one.³

3 Should (R-E) be formulated in terms of doxastic attitudes in general?

My next question about Schmidt’s argument against (R-E) concerns the formulation of (R-E) itself. As Schmidt formulates it, (R-E) is about epistemic reasons for doxastic attitudes in general, where this category includes at least belief, disbelief, and suspension. But all of her counterexamples to (R-E) concern reasons for suspension. (R-E) is untouched when it comes to (dis)belief. This is important because, as Schmidt herself notes (14), suspension is in all of her cases what Staffel (2019) calls a transitional attitude.⁴ A transitional attitude towards p doesn’t involve settling the question whether p, and is compatible with further inquiry as to whether p. A terminal attitude, by contrast, involves settling the question and is the terminus of doxastic deliberation. Belief and disbelief are terminal attitudes.

While Schmidt briefly attends to this distinction, it doesn’t feature in her presentation of the case for (R-E). But it is worth considering whether the case for (R-E) really supports the principle as Schmidt has rendered it, or whether there are important differences between transitional and terminal attitudes that should have restricted its scope in the first place. This is especially important given that the case for (R-E), as she presents it, focuses on reasons for belief rather than reasons for suspension.

According to Schmidt, the primary case for (R-E) is that it “may seem to be true by definition” (3). In other words, it might be thought that (R-E) just follows from the definitions of “epistemic reason” and “evidence.” Given that epistemic reasons have been defined as considerations that favor doxastic attitudes in a distinctively epistemic manner, it is a natural next step to identify this distinctively epistemic manner of favoring with bearing on the truth of a proposition. But this sounds like what pieces of evidence are: considerations that bear on the truth of a proposition (say, by probabilifying it⁵). So, it seems to follow that epistemic reasons for doxastic attitudes are just pieces of evidence.

That epistemic reasons and evidence go together in this way seems especially difficult to deny in the case of (dis)belief. To (dis)believe p is to settle for oneself the question whether p is true, and the only considerations that can be relevant to this question are considerations that bear on whether p is true. Nothing else could be a reason of the right kind for belief. Thus, the case for (R-E) seems quite clear when applied to the terminal doxastic attitudes of belief and disbelief. Is the case just as clear when it comes to suspension, a transitional doxastic attitude?

I don’t think so. As transitional attitude, suspending on p doesn’t involve settling the question whether p is true. Unlike in the case of belief, where we can define the right kinds of reasons for belief as ones that bear on the question the settling of which amounts to believing (see, e.g., Hieronymi 2005), it’s just not as obvious how to separate out the right and wrong kind of reasons (recall that Schmidt initially characterizes the epistemic manner of favoring by contrasting it with how practical reasons favor). Thus, the simple and straightforward argument above for the conclusion that epistemic reasons for belief are considerations that bear on whether p is true can’t be transposed to suspension.

Of course, this doesn’t mean epistemic reasons for doxastic attitudes towards p aren’t considerations that bear on the truth of p. Insofar as there are such things as properly epistemic reasons for suspension, it seems right to say they have something to do with truth. But the point is that the powerful case for (R-E) discussed by Schmidt applies much more clearly and straightforwardly to terminal attitudes like belief and disbelief. Grouping these attitudes toward p with transitional attitudes like suspension as “doxastic attitudes” may make it seem natural that there should be some unified sense of “epistemic reason” that applies to all of them in the same way. But once we recognize how different terminal and transitional attitudes are, it isn’t obvious that the category “doxastic attitudes” groups like with like when it comes to the reasons for those attitudes.

In light of this, we might ask whether proponents of (R-E) should be understood, as Schmidt does, as advocating for a version of it that applies to suspension in addition to (dis)belief in the first place. Perhaps what they primarily have in mind is the claim that all epistemic reasons for (dis)belief that p are provided by evidence concerning p. And that claim is untouched by Schmidt’s counterexamples, all of which have to do with reasons for suspension. Either way, it seems important to note the mismatch between the strongest case for (R-E) applying most clearly to reasons for (dis)belief, and Schmidt’s counterexamples to (R-E) applying only to reasons for suspension.

Now, Schmidt might respond that her counterexamples are still applicable because, as she has already claimed, reasons to suspend on p are ipso facto reasons against believing p (13-14).⁶ However, I don’t think this response would be sufficient given the distinction raised above. Reasons to suspend on p are reasons against settling the question whether p at all, whereas reasons against believing p constituted by evidence against p are reasons to settle the question whether p in the negative. In light of this, something seems to be lost if we take reasons for suspension qua transitional attitude to be interchangeable with reasons against belief qua terminal attitude.

4 Are incoherence-based reasons really epistemic reasons?

My final question regarding Schmidt’s view is about whether we should really, ultimately think of incoherence-based reasons as specifically epistemic reasons. Later in her paper, she discusses a dilemma for her view: either incoherence-based reasons can be identified with evidence after all, or we should think they are not actually epistemic reasons. When discussing the second horn, Schmidt notes that one problem for the view that incoherence-based reasons are epistemic reasons is that “incoherence is not a purely epistemic issue, but concerns theoretical and practical deliberation/attitudes equally” (18). Many paradigmatic cases of incoherence, such as akrasia and means-ends incoherence, involve a mix of doxastic and practical attitudes.

Schmidt’s response to this problem is to hold that incoherence-based reasons are epistemic reasons only in cases of incoherent doxastic attitudes in particular, because such incoherence is a purely theoretical failing. In mixed cases, by contrast, the incoherence is a moral or practical failing, so incoherence-based reasons would be moral or practical reasons. The idea here seems to be that the akratic or means-end incoherent sets of attitudes involve a moral/practical conclusion-state (an intention), so the failing in an incoherent set is moral/practical.

I’m skeptical of this response for two reasons. First, there seem to be mixed cases where the failing is not intuitively a moral/practical failing. Consider cases of means-ends incoherence involving an irrational means-ends belief. Imagine a friend is having a bad day, and I intend to cheer them up. Bizarrely, I believe that the only way to cheer them up is to gift them a saucer of mud. But I don’t intend to gift them a saucer of mud. Given that the locus of my failing here is in my bizarre means-ends belief, my failing seems to be a theoretical one, not a practical one. And insofar as there is some failing that consists solely in the incoherence of the set of attitudes, and not in any individual attitude, it doesn’t seem to be distinctively practical or theoretical. Thus, it’s not clear that we can categorize mixed cases of incoherence as distinctively moral/practical failures, independent of any assessment of the individual attitudes. That categorization seems to be provided by a specific attitude that has gone wrong, not the mere fact of incoherence.⁷

The second reason I’m skeptical is that the distinction between the moral/practical and the theoretical doesn’t seem to be exhaustive. Imagine I hope that I will live forever even though I believe that’s impossible. This is an incoherent set of attitudes but doesn’t fit into the aforementioned distinction at all. Assuming there’s an incoherence-based reason here, is there some category like “epistemic” or “practical” that it fits into? And does it matter whether there is? I worry that the more we look into the diversity of cases of incoherence, the less plausible this strategy of trying to sort incoherence-based reasons into categories like “epistemic,” etc., becomes. Why not instead hold that incoherence-based reasons transcend such boundaries?

Of course, much here depends on what’s at stake in characterizing incoherence-based reasons as distinctively epistemic in the cases Schmidt is interested in. Part of what she wants to show is that reasons-first epistemology doesn’t require the identification of epistemic reasons with evidence. But if the category of epistemic reasons is only interesting because it rules out practical considerations as right-kind reasons for doxastic states, I’m not sure it’s necessary to hold onto the category when trying to capture incoherence-based reasons. It seems to me that the larger point could be made with the broader notion of right-kind reasons. Indeed, Schmidt herself evokes this framing when she writes that, “incoherence-based reasons exhibit the ‘earmarks’ of right-kind reasons for doxastic attitudes” (20).

Right-kind reasons for believing p are standardly thought of as considerations that bear on whether p is true. Why is this? An attractive answer is that beliefs are correct if and only if they are true, and right-kind reasons are considerations that bear on the correctness of attitudes (see, e.g., Singh 2021). Now, consider my earlier proposal that incoherence-based reasons are in first instance reasons against incoherent sets of attitudes, rather than reasons to suspend on each individual attitude in the set. In an incoherent set of beliefs, not all the attitudes can be true. So, even though it might not be natural to categorize incoherence-based reasons as being provided by the evidence, they do seem to have something to do with the fact that the incoherent set must include at least one instance of incorrectness. Thus, they seem like right-kind reasons against the incoherent set of attitudes.

This explanation of why incoherence-based reasons are right-kind reasons can, in principle, be extended to mixed cases. On the assumption that, given the correctness conditions of other attitudes like intention, incoherent sets of attitudes will be guaranteed to contain at least one incorrect attitude, incoherence-based reasons against these sets of attitudes will also be right-kind reasons. This is, of course, a significant assumption. But I think it’s a plausible one (and I defend a similar view in Singh n.d.).

The question, then, is whether it might be better to sidestep all the messiness raised by the imperfect match between the two categories of incoherence-based reasons and epistemic reasons. Schmidt seems most interested in classifying incoherence-based reasons as epistemic reasons because of the pressure this puts on evidentialism. But it isn’t clear to me that the upshot of Schmidt’s argument relies on targeting evidentialism. Say we let evidentialists have (R-E). It seems to me that the thrust of Schmidt’s argument can still be preserved, by arguing the following. Incoherence-based reasons are right-kind reasons against incoherent sets of attitudes. These incoherent sets sometimes include doxastic attitudes. Therefore, incoherence-based reasons will sometimes bear on how we manage our doxastic attitudes. Notice that establishing this doesn’t require rejecting (R-E) or evidentialism, and it avoids the complications Schmidt notes for her view involving mixed cases. It’s unclear to me, then, why the argument needs to be framed in opposition to (R-E) and evidentialism.

5 Concluding remarks

I’ve raised three broad questions for Schmidt that I hope invite further expansion and clarification of her view. Each of them puts pressure on some aspect of her argument. The first question puts pressure on whether we should understand incoherence-based reasons as reasons for individual attitudes, rather than reasons against the entire incoherent set. The second question puts pressure on whether suspension is relevantly similar to (dis)belief to be grouped together in the way Schmidt does. And the third question puts pressure on whether the category of epistemic reasons in particular is really all that important for the spirit of Schmidt’s argument.

Despite not being sold on all of the specific moves that make up Schmidt’s argument and overall view, I think the paper achieves what I take to be its broader aim. Schmidt pushes back on the idea that evidence, conceived of as bearing on the truth of individual attitudes in some straightforward way (like probabilification), is the only kind of thing that’s relevant to how we manage our doxastic attitudes. Facts about incoherence are relevant too, even though they aren’t easily construed as evidence. Responsiveness to these incoherence-based reasons is important for keeping our doxastic and other attitudes in order.

I agree with all of this, and think Schmidt is absolutely right to draw further attention to incoherence-based reasons. I’m not convinced that these are reasons for individual attitudes, or that they are distinctively epistemic reasons in any meaningful sense. But these finer points are only important insofar as the goal of the paper is to argue against evidentialism. If the broader aim is to make a case for incoherence-based reasons as non-evidential but still right-kind reasons that are relevant to our doxastic attitudes, then I think Schmidt succeeds at that, regardless of her paper’s bearing on evidentialism.