A Defense of Doxastic Defeaters

Abstract

A defeater is, very broadly, a consideration that reduces or completely takes away justification from a subject’s belief about a certain proposition. According to a widely endorsed view, justifiers and defeaters require evidential support. However, a number of philosophers argue that unjustified beliefs can serve as defeaters as well. Call the former type of defeater an ‘evidential defeater’ and the latter a ‘doxastic defeater’. Doxastic defeaters are highly controversial. First of all, they seem to be flatly incompatible with evidentialism. Moreover and more alarmingly, if we accept that unjustified beliefs can be defeaters, we have to accept that unjustified beliefs can serve as justifiers as well. A further unwelcome implication is that epistemically irresponsible subjects could immunise themselves from defeat by generating their own defeater-defeaters. Problems like these have led philosophers to reject doxastic defeaters altogether. This paper argues that doxastic defeaters are intelligible given a dualistic conception of rationality.

1. Introduction

A defeater is, very broadly, a consideration that reduces or completely takes away epistemic justification or rationality from a subject’s belief about a certain proposition. ¹ According to a widely endorsed view, both justifiers and defeaters require evidential support. In the following, we will presume that a belief has evidential support if it is probable given the evidence. ² Call a defeater that requires evidential support an ‘evidential defeater’. The distinguishing feature of an evidential defeater is that it can be derived from one’s evidence. However, some authors argue that unjustified beliefs can also serve as defeaters – so-called ‘doxastic defeaters’. ³ For instance, your unjustified belief that you are a brain in a vat constitutes a doxastic defeater for your justified belief that you have hands. The distinguishing feature of doxastic defeaters is that they are unjustified doxastic states.

Although a number of philosophers defend the existence of doxastic defeaters (see, e.g., Bergmann 2006; Goldman 1986; Greco 2010; Lackey 2008; Pryor 2004), they are still highly controversial for several reasons. First of all, they are apparently incompatible with evidentialism. Secondly and more alarmingly, if we accept that unjustified beliefs can be defeaters, we have to accept that unjustified beliefs could serve as justifiers as well (Alexander 2017). A further unwelcome implication is that epistemically irresponsible subjects could immunise themselves from defeat by producing their own defeater-defeaters (Casullo 2018). Problems like these have led some philosophers to reject doxastic defeaters altogether (e.g., Alston 2002; Casullo 2018; and Graham & Lyons 2021).

This paper purports to provide a rationale for doxastic defeaters. Contrary to the received view, it will be argued that unjustified beliefs as well as justified beliefs can serve as defeaters – at least given the dualistic assumption that they respond to two distinct kinds of rationality: substantive and structural rationality. The upshot is that evidential defeaters target substantive epistemic rationality, whereas doxastic defeaters target structural epistemic rationality. The plan for the rest of the paper is as follows. Section 2 presents a rationale for doxastic defeaters based on dualism about rationality. It is argued that evidential defeaters fit well with a substantive account of epistemic rationality, and that doxastic defeaters fit well with a structural account of epistemic rationality. Finally, in section 3, it is explained how the problems regarding unjustified defeaters can be mitigated once doxastic defeaters are viewed in combination with a structural account of epistemic rationality. Section 4 concludes.

2. A rationale for doxastic defeaters

In the philosophical discussion over epistemic defeat, the focus has mostly been on the defeater, while significantly less attention has been devoted to the relevant defeatee at issue. Some philosophers make a distinction between evidence that takes away justification and evidence that takes away knowledge or warrant (e.g., Plantinga 1993). But not much has been said about what sense of justification or rationality is affected by defeat. As already mentioned, there are two main threads in our talk and thoughts about rationality: substantive and structural rationality. According to the substantive account of rationality, a belief is rational if it is the correct response to one’s reasons. In contrast, according to the structural account of rationality, a belief is rational if it coheres with one’s other beliefs and attitudes.

Reasons are broadly speaking considerations that count in favour of certain attitudes or actions (Scanlon 1998). ⁴ Reasons can be interpreted in different ways. For instance, it is common to make a distinction between objective and subjective reasons. Objective reasons are considered to be reasons that depend on facts or how things actually are, whereas subjective reasons are considered to be dependent on how the subject perceives that things are. The distinction between objective and subjective reasons is problematic for reasons that we cannot discuss in closer detail here (Fogal & Worsnip 2021). A more promising idea is an ‘evidence-relative understanding of reasons’. ⁵ In this view, reasons are dependent on one’s possessed evidence. In general, it is assumed that, in order to possess evidence, the evidence has to be within one’s epistemic reach. This means, roughly, that one has to have access to the relevant evidence and also the cognitive ability to evaluate it in order to possess the evidence (Lord 2018). For instance, the mere fact that there is evidence for believing that your building is on fire does not give you any reason to believe that your building is on fire unless you also have access to the relevant evidence (in this case, smoke, heat, etc.) Moreover, since ought implies can, it is also reasonable to presume that you have the cognitive capacity to make the proper inference from the evidence in order to possess the evidence.

What matters given a structural account of epistemic rationality is instead that you have a coherent set of beliefs and attitudes. According to an influential suggestion, one has to meet certain structural requirements in order to be rational in the structural sense (Broome 2013). Structural requirements of rationality are usually framed as a system of rules much like in chess, etiquette, or law. A typical structural requirement on practical rationality is means-end coherence. Roughly, to take the necessary means to one’s ends. For instance, rationality requires that if S intends an end E, and believes M is a necessary means to E, she intends M. A typical structural requirement on epistemic rationality is doxastic enkrasia. Very roughly, to align one’s beliefs with one’s beliefs about what one ought to believe. For instance, rationality requires that if S believes that her evidence supports p, she believes p. What rational requirements have in common is that they regulate what combinations of attitudes are required in order for you to be rational. On the most plausible interpretation of this view, you are not rationally required to have any particular attitudes – what matters is instead that your attitudes conform to a certain pattern or structure. ⁶

Although philosophers tend to accept the distinction between substantive and structural rationality, they disagree about what to think about how they relate to each other. Some authors argue that structural rationality can be either eliminated or explained in terms of substantive rationality (see, e.g., Kiesewetter 2017; Lord 2018). Others argue that substantive rationality can be either eliminated or explained in terms of structural rationality (e.g., Broome 2013). Both these approaches are forms of rationality monism. Wedgwood (2017) argues that substantive and structural rationality coincide. If that is the case, the distinction between substantive and structural rationality evaporates. Still others philosophers make a sharp distinction between reason and rationality (structural rationality). The former is considered to be a normative notion and the latter is merely apparently normative. ⁷ The main problem with this view is to explain the apparent normativity of rational requirements. By contrast, dualism about rationality is the view that substantive and structural rationality are two genuine and distinct normative areas of evaluation (see, Worsnip 2018; Worsnip 2021; Fogal & Worsnip 2021). That is, neither notion of rationality can be reduced to the other, nor do they coincide.

The best argument for dualism about rationality is based on the intuition that a subject can make separate rational mistakes that are of different kinds. Fogal & Worsnip (2021: 313) call this: ‘a counting intuition’ about rationality. Here is an example. Imagine a person, Tom, who believes that he is Superman. He also believes that Superman can fly. However, Tom does not believe that he can fly. It seems that Tom is guilty of two rational failings. First, it is irrational for Tom to believe that he is Superman since it is hard to see how this could be the right response to his evidence. Second, although Tom believes that he is Superman and that Superman can fly, he fails to believe that he can fly. This is obviously incoherent. So, he commits another rational mistake as well in virtue of being incoherent. Tom is, therefore, guilty of two distinct rational mistakes: (i) Believing that he is Superman; and (ii) failing to believe that he can fly. These mistakes are different in kind. The first mistake is caused by the fact that Tom fails to correctly respond to his evidence. The second mistake is caused by the fact that Tom fails to make his beliefs coherent. The two mistakes correspond with the two types of rationality that are discussed above.

However, what matters for our purposes is that dualism about rationality opens up for an account of defeat that includes both evidential and doxastic defeaters. Evidential defeaters appear to fit well with a substantive account of epistemic rationality (especially given an evidence-relative understanding of reasons). In this view, defeaters are epistemic reasons that speak against holding certain beliefs. By contrast, doxastic defeaters are a poor fit to a substantive account of epistemic rationality. Remember that doxastic defeaters are not necessarily true or justified. According to most accounts, reasons are based on either facts or evidence. Therefore, it is difficult to see how doxastic defeaters could qualify as epistemic reasons. On the other hand, doxastic defeaters appear to fit well with a structural account of epistemic rationality. What matters for structural epistemic rationality is that one’s beliefs and attitudes cohere with each other – whether or not the relevant attitudes have evidential support does not have to be crucial in any sense.

For instance, suppose again that you unjustifiably believe that you are a brain in a vat. This belief is supposed to be a doxastic defeater for your belief that you have hands. Why is that? Your belief that you are a brain in a vat neither appears to be justified or true. The reason why your belief that you are a brain in a vat constitutes a doxastic defeater, according to a structural account of epistemic rationality, is that it creates an incoherence among your beliefs and attitudes. You cannot simultaneously and coherently believe that you are a brain in a vat and that you have hands. Obviously, if you are envatted, you cannot possibly have any hands, and vice versa; if you have hands, you cannot be envatted.

At this point, one might ask why your belief that you are a brain in a vat is a doxastic defeater for your belief that you have hands, and not the other way round? Peter Graham & Jack Lyons (2021) argue that there can be no such thing as doxastic defeaters. The main reason for this is that they believe that it is arbitrary whether a certain belief is a doxastic defeater or not. They concede that structural rationality requires some accommodation but it does not require that one gives up any belief in particular. As long as one gives up either one’s belief about p or the potentially defeating belief, one will regain coherence amongst one’s attitudes. For instance, suppose that S believes p and that the justificatory source behind her belief that p is q. But then, S comes to believe that there is something wrong with q. At this point, S appears to have an incoherent set of beliefs. It is not coherent for S to believe: (i) p; (ii) the justificatory source for my belief that p is q; and (iii) there is something wrong with q. If there is something wrong with q, surely it cannot be a justificatory source for p. However, note that at this point, S could regain coherence by simply giving up her belief that p. If there is something wrong with q, then S should not believe p on the basis of q. But there is also another way for S to regain coherence. S could just as well give up the belief that there is something wrong with q. In that case, S could coherently believe p. So, it appears to be arbitrary whether or not the undercutter defeats S’s belief that p – at most, there is a potential undercutting defeater for one’s belief about p.

If the argument is sound, doxastic defeaters are not actual defeaters but merely potential defeaters. I think, however, that the argument rests on a misunderstanding about the required duration of defeat. Graham & Lyons (Ibid) rightly point out that certain combinations of beliefs are rationally impermissible rather than specific beliefs within these combinations according to a structural account of epistemic rationality. As already stated, S can escape incoherence in two ways: either by giving up her belief about the target proposition or by giving up the putative defeater. But note that even if S gives up the putative defeater instead of giving up her belief about p, the damage to the relevant belief is already done. The moment S comes to believe that there is something wrong with the believed source of justification behind her belief about p, she has a structural undercutting defeater for her belief about p. Subsequently giving up the defeating belief will not change this retrospectively. At most, it enables S’s belief to regain coherence with her other beliefs and attitudes. For this reason, we should not consider the relevant doxastic defeater to be merely a potential defeater. Instead, it might be considered to be a temporary doxastic defeater since it can subsequently be given up. However, temporary doxastic defeaters are still actual defeaters and not merely potential defeaters.

3. Problems regarding unjustified defeaters

So far, I hope to have demonstrated how a structural account of epistemic rationality can provide a rationale for doxastic defeaters. The time has come to return to the objections raised against doxastic defeaters that were mentioned at the outset. I will argue that the objections are misguided or that they at least can be mitigated given that we combine doxastic defeaters with a structural account of epistemic rationality. The most problematic feature of doxastic defeaters is probably the implication that unjustified beliefs (i.e., beliefs that lack evidential support) can be defeaters. Unjustified defeaters are considered to be implausible for several reasons. First of all, one might ask why beliefs that lack evidential support should have any rational impact at all. For instance, according to evidentialism, one should believe in accordance with one’s evidence. If this view is correct, one should pay no attention to beliefs that do not belong to one’s evidence, e.g., beliefs that lack evidential support. Alexander calls this ‘the problem of rational inclusion’ (2017: 895).

Most philosophers agree that unjustified beliefs cannot be part of one’s total evidence and this is also what follows from an evidence-relative conception of epistemic reasons. In order for a consideration to count as a reason to believe p, this reason has to be based on one’s possessed evidence. A noteworthy exception is Michael Bergmann (2006) who thinks that unjustified beliefs can be part of one’s total evidence. Unfortunately, he does not offer much argument in favour of this view. Note, however, that what is rational to believe, according to a structural account of epistemic rationality, does not ultimately depend on one’s total evidence. What matters is instead the relationship between one’s beliefs and attitudes. Whether or not a belief is based on one’s evidence is irrelevant from this perspective. As long as one holds a certain belief, then that belief will partly determine whether one’s beliefs and attitudes cohere or not. A belief does therefore not have to belong to one’s total evidence in order to have an epistemic impact on what it is rational to believe provided that we combine doxastic defeaters with a structural account of epistemic rationality. As a result, the alleged problem of rational inclusion becomes a pseudo-problem.

Second, regardless of whether unjustified beliefs should be included in one’s total evidence or not, one may still ask why unjustified beliefs should have any rational impact. It does not follow from the fact that an unjustified belief is evidence that it should determine what you should believe in relation to the relevant matter at issue. Alexander calls this the ‘problem of rational import’ (2017: 895). Note that the problem of rational import is distinct from the problem of rational inclusion. After all, a possibility could be that unjustified beliefs can count as evidence but that they still are unable to have any substantial rational impact. Unjustified beliefs might be systematically neutralised by other beliefs or evidence, so-called defeater-defeaters. For instance, suppose that you believe that you have hands and then unjustifiably come to believe that you are a brain in a vat. As a result, it seems that you have a defeater for your belief that you have hands. Suppose, however, that the reason why your belief that you are a brain in vat is unjustified in the first place is that you have plenty of other beliefs that imply that you are not envatted. Then it seems that your belief that you are a brain in a vat is defeated beforehand by other beliefs that are included in your total evidence, i.e., defeater-defeaters. As a result, it is difficult to see how an unjustified defeater could have any rational impact on what one is rational to believe in the light of the rest of one’s evidence.

Notice that the reasoning above seems to presuppose an understanding of rationality in terms of evidence and goes well with a substantive account of epistemic rationality. However, something similar appears to hold even if we assume a structural account of epistemic rationality. For instance, your belief that you are a brain in a vat would not cohere very well with beliefs like that you wear gloves to protect your hands from the cold, that you have nails on the fingers of your hands, etc. However, not all unjustified doxastic defeaters are as aberrant as the belief that you are a brain in a vat. Coming to believe on the basis of pure gut feeling that the person who gave you a certain piece of information is untrustworthy appears to pass the test in a much better way. Suppose, for instance, that you think that the person in question looks suspicious (without having any particular evidential support for this belief). Believing that a certain person is untrustworthy might very well be coherent with your other beliefs and attitudes in a way that believing that you are a brain in a vat fails to be. The problem of rational import does therefore not exclude the existence of doxastic defeaters. What is decisive is instead whether the putative defeater coheres well with one’s other beliefs and attitudes or not. As it stands, this appears to be a live option. So, it can at least not be excluded that unjustified beliefs can have a rational impact given a structural account of epistemic rationality.

Third, given that unjustified beliefs can serve as defeaters, it should follow that they can serve as justifiers as well. This is a strange result since it is widely assumed that ‘a belief can justify only if justified’ (Alexander: 891). Moreover, if we accept that unjustified defeaters can serve as justifiers, we will get implausibly low standards of justified beliefs (Casullo, 2018: 2901). In the same spirit, William P. Alston (2002) has argued that accepting unjustified defeaters implies an overly subjective view of rationality. Alston makes reference to Plantinga who holds that a radically irrational belief can serve as a justifier for another belief. As an example, Plantinga uses Descartes’ madman, M, who believes that his head is made of glass. If M believes that wearing a football helmet at all times protects his glass head from being shattered, it is rational for him to wear a football helmet at all times.

One line of response might be to grant that unjustified beliefs cannot serve as justifiers but maintain that they still can serve as defeaters. In fact, this seems to be the position of most defenders of doxastic defeaters (see, among others, Goldman 1986; Bergman 2006). The problem with this line of response is that it seems to be arbitrary (Alexander: 895). Why should an unjustified belief be able to serve as a defeater but not as a justifier? The position seems to be particularly arbitrary if we take into consideration that a defeater-defeater might equal a justifier. However, from a perspective of structural epistemic rationality, there is neither something peculiar about the fact that unjustified beliefs can serve as defeaters nor about the fact that they can serve as justifiers. For instance, if M believes that his head is made of glass, it follows that he is justified to believe that he needs to wear a football helmet at all times. ⁸ Notice that ‘justified’ in this case refers to justified in the structural sense and not the substantive sense of epistemic rationality. In contrast, the reason why M’s belief that he needs to wear a football helmet at all times is not justified in the substantive sense is that M cannot plausibly possess any substantial evidence that supports believing so.

Of course, one might still insist with Alston (2002) that this makes rationality overly subjective, in other words, to reject the idea of structural epistemic rationality altogether. However, this type of objection seems to beg the question against the suggestion that has been proposed here. Remember that our point of departure is the hypothesis that evidential defeaters are considerations relevant given a substantive account of epistemic rationality, whereas doxastic defeaters are considerations relevant given a structural account of epistemic rationality. If one does not think that a structural account of rationality is a plausible form of rationality in the first place, one will neither accept the arguments presented here. On the other hand, in that case, it needs to be explained why we tend to perceive cases of structural irrationality as paradigmatic cases of irrationality. ⁹

Fourth and finally, accepting unjustified beliefs as defeaters implies that epistemically irresponsible subjects can immunise themselves from defeat (see Casullo 2018; Klenk 2019). It is easy to come up with a defeater that defeats all conflicting beliefs if it is not required that the defeater is justified, i.e. without any need for evidential support. For instance, suppose that I consult the tea leaves in my cup and reach the conclusion that my head is made of glass. For some reason, I also form the belief that all other sources of evidence that give rise to beliefs that stand in conflict with my belief that my head is made of glass are untrustworthy. In this way, I am able to make my belief that my head is made of glass immune to defeat.

This is an unwelcome consequence since being dogmatic and stubborn should not be able to make one entirely immune from defeat. However, given the distinction between substantive and structural epistemic rationality, it does not follow that such subjects are immune to defeat all things considered. Unreflective subjects might make themselves immune to defeaters for structural epistemic rationality but they are not immune to defeaters for substantive epistemic rationality, i.e., evidential defeaters. It is sufficient that there is a reason for an unreflective subject to believe that there is something wrong with the source of justification or the justificatory process in order for her to be irrational in the substantive sense. On the other hand, remember that in order for a subject to have an evidential defeater, she must also possess the relevant evidence. And in order to possess the relevant evidence, it has to be within her epistemic reach. If the subject in question lacks the general ability the interpret the evidence in the proper way, she might evade an evidential defeater as well. However, I do not think that this result is particularly worrisome or controversial. Subjects that lack basic cognitive abilities will also lack the ability to form rational beliefs and ipso facto be immune to defeat. For similar reasons, they might also lack the ability to hold justified or rational beliefs to start with.

4. Conclusion

The main purpose of this paper has been to provide a rationale for doxastic defeaters. The paper defends a dualistic approach to epistemic rationality and defeat. It has been argued that doxastic defeaters are best understood in combination with a structural account of epistemic rationality. It has also been shown how the problems often raised against doxastic defeaters are misguided or at least that they can be significantly mitigated given this rationale.