The chapter argues that the main extant views of the nature of evidence one has lack the resources to account for the impermissibility of cases of resistance to evidence. I first examine classic internalist, seemings-based evidentialism and argue that it fails to account for evidence resistance. This, I argue, is an in-principle problem: internalist evidentialism cannot recover from this because it is internalist.

I move on to externalist views of evidence, starting with factive externalism (i.e. Williamson’s (Reference Williamson2000) evidence is knowledge (E = K) ), and I argue that, since resistant cognisers don’t take up the relevant facts in the world to begin with, the view fails to predict epistemic impermissibility in resistance cases. I also look at and dismiss several ways in which the champion of E = K might attempt to account for what’s going wrong in resistance cases (i.e. via employing notions such as epistemic dispositions one should have had and epistemic blameworthiness), and I argue that the view faces difficulties. Finally, I move on to less radical, non-factive externalisms and investigate the potential of prominent reliabilist views – indicator reliabilism (Comesaña Reference Comesaña2020) and virtue reliabilism (Turri Reference Turri2010, Sylvan and Sosa Reference Sylvan, Sosa and Star2018, Sosa Reference Sosa2021) – to account for the phenomenon of resistance. I argue that these views are too agent-centric to successfully account for resistance cases.

2.1 Evidence Internalism

Evidence matters: the concept of evidence is central to epistemology, the philosophy of science, the philosophy of law, the ethics of responsibility. Outside philosophy, the concept of evidence is highly employed as well: lawyers, judges, historians, scientists, economists, investigative journalists, and reporters, as well as ordinary folk in the course of everyday life, talk and think about evidence a lot.

Both within and outside of philosophy, what we care most about is not just the nature of evidence alone, but rather what it is for a subject to have evidence. We care, as it were, about evidence had. That makes sense in philosophy because we are interested in the quality of our beliefs and our actions, and the latter will mostly be affected by the evidence we have. Outside of philosophy, evidence one has bears relevance to one’s legal status, professional performance, decisions, policies, voting, plans, etc. Evidence one has, the thought goes, but less so evidence one does not have, will influence all of one’s walks of life.

It is interesting to note – in line with the main scholarly source on the nature of evidence, the Stanford Encyclopedia of Philosophy article on the issue (Kelly Reference Kelly and Zalta2016) – that the ways in which we think of evidence outside and within philosophy are strongly incompatible with each other. In philosophy, we disagree a lot about the nature of evidence, but one thing that the vast majority of theorists have always assumed is that the having relation is somehow related to the limits of one’s skull: one has evidence, on this received view, when one uptakes it ‘in one’s head’ – be it via seemings, beliefs, knowings, etc. In contrast, outside of philosophy, the having relation has never been about the skull: just try to tell a judge that you had no evidence that the butler did it, even though he did it right in front of you, because you couldn’t believe your eyes; see how that goes down.

Of course, one might think, what’s the surprise there? Experts know best in all domains – that’s what semantic externalism teaches – and philosophy is not an exception. The way in which us laymen conceptualise ‘depression’ is likely different from the way in which psychiatrists do; that’s fine, we’re wrong, and the experts are right. The same goes for evidence one has.

In what follows, I will argue that this way of thinking about the issue at hand is mistaken: in particular, I will propose that the philosophical conception of the having relation – as having to do with the limits of one’s skull – fails on extensional grounds – having to do with failing to account for the impermissibility of evidence resistance – and, as a result, it also fails to fulfil its central function in predicting accountability and legal responsibility.

Let’s start with a classic: according to internalist evidentialist, phenomenal conceptions of evidence, a subject S’s evidence consists (roughly) in what it seems to S to be the case. This view has a notable tradition: Russell, for instance, thought of evidence as sense data, mental items of one’s present consciousness with which one is immediately acquainted. Similarly, Quine thought that evidence consisted of the stimulation of one’s sensory receptors. Finally, and more recently, according to Connee and Feldman (Reference Conee and Feldman2004), one’s evidence consists exclusively of one’s current mental states.

The view accommodates our intuition in New Evil Daemon cases: the recently envatted brain-in-a-vat version of myself, the thought goes, is, intuitively, just as justified as I am to believe that she’s typing on her laptop right now. An evidentialist account of justification, in conjunction with a phenomenal view of evidence, vindicates this intuition.

A classic problem, however, for this way of thinking has to do with seemings with bad etiologies: sometimes, our seemings are based in wishful thinking and racial bias rather than proper cognitive mechanisms. When this happens, the phenomenal conception predicts – against intuition – that we have evidence for our corresponding beliefs. In turn, this problem renders the phenomenal conception of evidence incapable to distinguish between epistemically permissible evidence rejection and problematic evidence resistance. Take the science sceptic Neda again: in the good case, it seems to Neda that vaccines are unsafe because of reliable testimony that they are unsafe. In the bad case, her seemings are sourced in an irrational fear of needles. The phenomenal conception of evidence has trouble distinguishing between the good and the bad cases.

2.2 E = K

According to the prominent, knowledge-first view of having evidence (Williamson Reference Williamson2000), for any subject S, S’s evidence is S’s knowledge. Since knowledge implies belief, and since all of the protagonists in Cases 1–7 from Chapter 1 lack the relevant beliefs, E = K will predict that the subjects in question lack evidence: for example, Bill, the fervent supporter of President Dump, does not believe, and therefore does not know, that Dump is a bad president; furthermore, he does not believe, and therefore does not know, any of the statements by the media, etc., that suggest as much, and thus, on this view, he has no evidence that Dump is a bad president. And the same will hold for all of the protagonists of Cases 1–7. In this, E = K cannot make good on the resistance intuition – at least not when unpacked as resistance to evidence one has. Furthermore, several knowledge-first theorists explicitly embrace this result: according to people like Hawthorne and Srinivasan (Reference Hawthorne, Srinivasan, Christensen and Lackey2013), for instance, short of knowing, one should withhold belief.

One alternative way to account for our cases within an E = K framework would be by employing the notion of being in a position to know in order to account for evidence that is easily available but not possessed by the agent. Plausibly, the thought would go, the Dump supporter is in a position to know that Dump is a bad president: that’s what explains our intuition that he’s failing epistemically when he fails to form the corresponding belief.

Of course, a lot will hinge on how the relevant notion of ‘being in a position to know’ is spelled out: importantly, the relevant notion should be E = K-friendly (i.e. it should be compatible with the thought that evidence one has amounts to knowledge). Consider, first, a view on which I am in a position to know that p if and only if there is evidence for p available to me, and evidence is available to one just in case it consists of facts that follow from or are made probable by one’s extant knowledge. On this view, Bill is in a position to know p: ‘Dump is a bad president’ in virtue of the fact that it follows from his other extant knowledge – such as his knowledge that presidents shouldn’t lie, shouldn’t make racist and sexist comments, etc., together with his knowledge that Dump engages often in all of the above.

Unfortunately, this view will not deliver the needed result if we describe the case as one in which Bill’s system of (false) beliefs about Dump being a great president is perfectly coherent (in that Bill either doesn’t believe that lying, etc., are bad or doesn’t believe Dump lies, etc.), although unjustified: p will not follow from any piece of knowledge Bill has. To bring this point into even sharper relief, consider also the Perceptual Non-responsiveness case: what is the relevant piece of knowledge here?Footnote ¹

Here is one alternative E = K-friendly way to unpack being in a position to know: S is in a position to know that p if and only if, were S to believe that p, S would know that p. Bill, then, on this account, is in a position to know that Dump is a bad president if and only if, were he to form the relevant belief, he would come to know that Dump is a bad president.Footnote ²

The problem with this account is that if, on the one hand, we keep Bill’s psychology otherwise fixed, and all that changes is his forming the relevant belief, it will fail to constitute knowledge in virtue of its acute incoherence with the rest of his belief system. On the other hand, if, in order to assess Bill’s actual epistemic situation, we go and look at the closest world where Bill’s psychology is radically different, such that, indeed, were he to form the belief that Dump is a bad president, it would constitute knowledge, our account of being in a position to know becomes too strong.Footnote ³ To see this, consider Alvin Goldman’s (Reference Goldman and Tomberlin1988) benighted cogniser – let us call him Ben. This fellow lives on a secluded island where he’s been taught that reading astrology is an excellent way to form beliefs and where he has no access to any clue to the contrary. Plausibly, there is no evidence available to Ben for p: ‘astrology is an unreliable way to form beliefs’, nor is he in a position to know it. However, at the closest world where things are different enough (say that Ben leaves his benighted community), such that now he believes the relevant proposition, he knows it. As such, the account construed along these lines will mistakenly place Ben in the same boat as the Case 1–7 protagonists, in spite of the fact that Ben has no way to access information of the unreliability of astrology.

One last move available to the defender of E = K is to argue that what is present in Cases 1–7 and explains resistance intuition is potential evidence: evidence that Bill, the Dump supporter, would have had, had he not had bad epistemic dispositions.Footnote ⁴ Since, plausibly, one should have good epistemic dispositions rather than bad epistemic dispositions, the view predicts that Bill is in breach of an epistemic ‘should’. Williamson (Reference Williamson2000, 95 and in conversation) gestures at a view like this.

One important problem with this move, however, is that it is both too weak and too strong.

To see why the view is too weak, note that a version of the E = K account thus construed will miss an important distinction between synchronic and diachronic epistemic shoulds: the distinction between the synchronic ‘should’ of epistemic justification and the diachronic ‘should’ of responsibility in inquiry.Footnote ⁵ Proceeding responsibly in inquiry (e.g. thoroughly searching for evidence diachronically) is one thing; synchronically responding well to available evidence is another. However, both are governed by epistemic shoulds.Footnote ⁶

To see this, think back to the Friendly Detective case. Say that, this time around, Dave is investigating the crime scene with his colleague, Greg. Greg is rather lazy and distracted: he briefly looks around, fails to find any evidence at the crime scene, and concludes that there’s no evidence to suggest that the butler did it. In contrast, Dave is extremely thorough, but, at the same time, a close friend of the butler. Dave finds conclusive evidence that the butler did it at the crime scene but fails to form the corresponding belief.

I submit that both Dave and Greg are rather rubbish detectives, in that they fail to conduct their inquiry well – they are both in breach of the diachronic epistemic should of inquiry. Also, both Dave and Greg display pretty bad epistemic dispositions: Greg is a sloppy epistemic agent, while Dave fails to believe what the evidence supports. Compatibly, I submit, there is an important epistemic difference between Dave and Greg: Dave, but not Greg, is aware of all of the evidence in support of the hypothesis that the butler did it and fails to form the relevant belief nevertheless; Dave is resistant to available evidence.

The view, then, is too coarse grained to do the work needed to account for this datum. What is needed is a principled way to identify the epistemic dispositions and the corresponding epistemic should that matter in resistance cases.

To see why the view is also too strong, note that one need not have bad epistemic dispositions in order to fail, epistemically, in the way in which, for example, Bill, the Dump supporter, does: it can be a one-off affair. Maybe Bill is an excellent epistemic agent in all other walks of life: it’s only this particular belief – that Dump is a bad president – that he refuses to form against all facts speaking in favour of it.Footnote ⁷

2.3 Reliable Indicators

We have seen that strong, factive externalism struggles to accommodate the resistance data. In what follows, I will look at non-factive, reliabilist externalisms, in search for the normative resources we need to this effect. In this section, I take on Juan Comesaña’s (2010, Reference Comesaña2020) reliabilist view of evidence. In the next section, I will look at virtue reliabilist (e.g. Sylvan and Sosa Reference Sylvan, Sosa and Star2018, Sosa Reference Sosa2021) account of reasons to believe and Turri’s virtue reliabilist account of propositional warrant.

It is surprising to see just how very few fully fledged non-factive externalist accounts of the nature of evidence and defeat are available in the literature. Comesaña’s account is one that supplies this lack. The view falls squarely in-between the main camps on the market when it comes to the study of evidence: it is less demanding than factive views of evidence à la Williamson, in that, on Comesaña’s account, one can have evidence that is false. It is, however, more demanding than internalist views, in that experiences will only provide their content as a reason for belief when belief in the content is ultima facie justified. In this, Comesaña’s account promises both to reap the benefits of the main competitors and to avoid all of their downsides. To see this, consider the following case:

Understandably, Tomás is disappointed – but was he irrational in acting as he did? Juan Comesaña’s answer is: obviously not. According to Comesaña, Tomás’s action was rational because it was based on the rational belief that the candy-looking object Lucas was offering him was a candy. Tomás’s belief was rational because it was based on evidence that, in Comesaña’s view, is constituted by those propositions that Tomás is basically justified in believing by his experiences.

According to this account, then, an experience provides its content as a reason when the subject is justified in believing its content. The justification in question in the account, importantly, (1) is non-factive and (2) must be ultima facie: if an experience of the subject S provides them with prima facie justification for believing its content but this justification is defeated by something else S is justified in believing, then S does not have the content of the experience as evidence (Comesaña Reference Comesaña2020, 119).

(Early) Comesaña favours a reliabilist account of justification. If coupled with his view of evidence, this will render the latter stronger than internalism about evidence, in that only those contents of experience that are believed based on a reliable process will qualify as evidence.

In contrast to E = K, the view accounts for the intuition of rationality in CANDY: Tomás’s belief that Lucas is offering him candy will come out as justified and thereby as a proper part of his body of evidence. The view, when coupled with reliabilism about justification, also scores points over the internalist, in that it nicely accounts for a normative difference that we want our view of evidence to predict between Tomás and, for example, a wishful thinker or a biased cogniser: after all, wishful thinking and forming beliefs based on biases are not reliable processes, therefore the contents of the thus generated experiences will not constitute evidence.

At first glance, the account also promises to deliver the result we want in several of the resistance cases. Take, for one, Testimonial Injustice:

• Case #1. Testimonial Injustice: Anna is an extremely reliable testifier and an expert in the geography of Glasgow. She tells George that Glasgow Central is to the right. George believes women are not to be trusted; therefore, he fails to form the corresponding belief.

Since on the view under consideration an experience provides its content as a reason just in case the subject is justified in believing its content, and since, by stipulation, Anna is an extremely reliable testifier, the content of George’s experience of her telling him that Glasgow Central is to the right constitutes evidence. Anna’s testimony provides George with a reason to believe Glasgow Central is to the right, which he fails to take up. Similarly, the account correctly predicts that the Dump supporter has evidence that Dump is a bad president, which he ignores, that Mary has evidence that her husband is cheating, that the detective has evidence that the butler did it, etc. All of these people have experiences with the relevant contents that are reliably generated, the contents of which thus count as evidence on Comesaña’s view.

Unfortunately, on closer inspection, it turns out that granting the indicator reliabilist success on resistance cases is a bit premature. In particular, as I’m about to argue, the reliabilist treatment of the case of Professor Racist (and, relatedly, of any cases with a similar structure; i.e. cases where no experience of the facts at stake is present) is problematic at two crucial junctures. Here is the case again for convenience:

• Case #6. Misdirected Attention: Professor Racist is teaching college-level maths. He believes people of colour are less intelligent than white people. As a result, whenever he asks a question, his attention automatically goes to the white students, such that he doesn’t even notice the Black students who raise their hands. As a result, he believes Black students are not very active in class.

First, note that, against intuition, indicator reliabilism will predict that there is no evidence for Professor Racist that the Black students are active in class. After all, he has no experience with this particular content; therefore, he has no reliably generated experience with this content either. I take it that this is not a great result in itself. More generally, I take it, if our epistemology predicts that, simply because they ignore the facts, there is no evidence for racists and sexists that, for example, Black people and women are to be trusted, that they are deserving of good treatment, etc., we should probably go back to the drawing board.

Furthermore, note that the case of Professor Racist is not unique in this respect: we can modify all of the other cases along the exact same lines (i.e. ramping up the epistemically bad features) to get the same wrong predictions. Here is how: first, we can make it such that our characters not only don’t form the relevant beliefs because of sexism, politically motivated reasoning, etc., but they don’t even host the corresponding experiences. Say that George, for instance, in Testimonial Injustice, not only doesn’t believe what Anna says, but he doesn’t even register that she said anything at all due to his sexist bias: he just zones out when women speak. In all cases like these, contra intuition, indicator reliabilism will predict absence of evidence. Furthermore, the view now has the unpalatable consequence that tuning up epistemically bad properties can lead to an improvement in an agent’s epistemic position. Making sexist George more sexist such that he not only discounts the female passer-by’s words, but he doesn’t even listen to her when she kindly provides him with directions to the Glasgow Central will amount to an improvement in his overall epistemic state. I find this consequence highly problematic.

Second, consider a variation of the case in which sexist George systematically mishears what he is being told by female speakers in general, such that whenever he encounters disagreement, he hears agreement. It is perhaps even harder to believe that this trait should lead to an improvement in his epistemic position towards the relevant propositions.

Third, note that we can now even drop the gender-discriminatory component of the case. We may suppose that George simply mistakes disagreement by anyone for agreement. Again, it’s implausible that, as a result, George should be insulated from epistemic normativity: clearly, George is not justified in his beliefs.

2.4 Virtuous Reasons

This section investigates whether virtue epistemology has the resources needed to account for what is going wrong in Cases 1–7. For the most part, virtue epistemologists distance themselves from talk of evidence. However, they have other resources that they could employ to explain what is going wrong in Cases 1–7: the market features well-developed virtue-theoretic views of reasons to believe (Burge 2013, Sylvan and Sosa Reference Sylvan, Sosa and Star2018) and propositional warrant (Turri Reference Turri2010). According to these authors, broadly speaking, competences come first in epistemic normativity. I will examine these accounts in turn.

According to Kurt Sylvan and Ernie Sosa (Reference Sylvan, Sosa and Star2018), a fact is an epistemic reason to believe for S just in case it is competently taken up and processed by S. At root, then, reliable epistemic competence is doing the epistemic warranting work, even when reasons are involved. The only way in which a reason can have any epistemic normative standing is if it is competently taken up and processed by the agent: “[We] think […] claims [about reasons supporting a species of justified belief] could only be true if possession and proper basing are themselves grounded in a deeper normative property of competence” (Sylvan and Sosa Reference Sylvan, Sosa and Star2018, p. 557).

In turn, epistemic competences are traditionally unpacked as dispositions to believe truly (Sosa Reference Sosa2016) or know (Miracchi Reference Miracchi2015, Kelp Reference Kelp2018, Schellenberg Reference Schellenberg2018).

The view, whether construed along truth-first or knowledge-first lines, is too weak to account for what is going wrong in cases of resistance to evidence: Think back to the case of Bill, the Dump supporter; on this view, we get the result that there are no reasons for Bill to believe that Dump is a bad president, since he is not uptaking the relevant facts (i.e. media testimony, Dump’s own actions, etc.) via his cognitive competences. The same will hold for all of Cases 1–7: there will be no epistemic reasons for sexist and racist subjects to believe women and Black people; there will be no reason for Alice to believe that there is a table right in front of her; there will be no reason for Mary to believe that her partner is cheating; and finally, there will be no reason for Detective Dave to believe that the butler did it. All of these facts fail to constitute epistemic reasons on this view, since they are not competently processed by the subjects.

But can’t the virtue theorist appeal to these epistemic agents’ lack of competence to explain the poor epistemic status of beliefs that they do hold and account thereby for the impermissibility intuition?Footnote ⁸ For instance, can’t the virtue theorist argue that what is going on in cases like Political Negligence is that Bill is an epistemically incompetent believer, which results in him not being justified in his belief that Dump is a good president. This, the virtue theorist may argue, is enough to explain the intuition of epistemic impermissibility; we don’t also need to predict that there are reasons for Bill to believe that Dump is a bad president.

Two things should be said about this: first, note that it need not be that Bill is a rubbish epistemic agent overall. Indeed, maybe Bill is actually an extremely reliable believer, including about political matters. It’s only on this particular instance that he gets it wrong (after all, competences need not be infallible but merely reliable; thereby, their presence and manifestation are compatible with occasional failures). If so, the virtue theorist cannot appeal to lack of competence to explain this datum.

Second, I take it that it is independently problematic if a view predicts that there are no reasons for Bill to believe that Dump is a bad president; that is, independently of the epistemic status of his belief that Dump is a good president. To see this, consider a variation of the case in which Bill just doesn’t have any belief on the issue, in spite of all the media reports, Dump’s own actions, etc. It still seems as though there is something epistemically impermissible about Bill’s doxastic behaviour. However, since Bill is not forming any belief on the matter of Dump’s fitness for office, he isn’t forming any incompetent belief either.

One way to go for the virtue theorist here would be to blame the impermissibility on the availability of propositional warrant. The thought would go something along the following lines: what triggers the resistance intuition has to do with warrant that one has but that one fails to update on.

Unfortunately, resistance cases also generate problems for the virtue-theoretic view of propositional warrant, and for pretty much the same reason why they generate problems for virtue-theoretic accounts of reasons: because virtues come first in the relevant analysis. According to John Turri, for all p, p is propositionally warranted for a subject S iff S possesses at least one means to come to believe p such that, were S to form the relevant belief via one of these means, S’s belief would be doxastically warranted. In turn, doxastic warrant is unpacked in terms of epistemic competence: S is doxastically warranted to believe p iff S’s belief is the product of a reliable belief-formation competence of S’s.

On this view, since sexists, racists, and wishful thinkers are, by definition, people who lack the dispositions to form true or knowledgeable beliefs on the relevant issues, we get the counterintuitive result that these subjects lack propositional warrant and thus are not doing anything wrong, epistemically, in not forming the relevant beliefs.Footnote ⁹

What are we to do? Here is one move the virtue theorist might want to make here: dispositions can fail to manifest themselves when ‘masked.’ Consider the fragility of a vase. When in a room filled with pillows, the vase is still fragile, although its disposition to break cannot manifest itself. Similarly, virtue theorists could argue, Bill has an epistemic ability to form the relevant true belief about Dump, but it’s ‘masked’ by the presence of many incompatible – though false – beliefs about Dump. Similarly, sexist George’s epistemic competences are masked by his sexism, Professor Racist’s by his racism, and so on.

There are two problems with this move, however. First, the view thus construed overgeneralises, for it, once more, threatens to mistakenly place Goldman’s benighted cogniser and the protagonists of Cases 1–7 in the same epistemic boat. After all, Ben the benighted cogniser is the straightforward epistemic counterpart of a vase in a room full of pillows: were he to move to a friendlier epistemic environment, he would employ the right kinds of methods of belief formation. In this, he has a masked disposition to do well, epistemically.

Second, factors that ‘mask’ dispositions are commonly believed to be environmental factors (Choi and Fara Reference Choi, Fara and Zalta2018) – recall again the vase in the room full of pillows – rather than factors somehow ‘internal’ to the item in question; indeed, when the problem lies within the object itself – say that we inject all of the pores of the vase with glue, for instance – the more plausible diagnosis is lack of disposition – no fragility – rather than masked disposition. However, in many of the Cases 1–7, it is the subject’s own mental state (biases, wishful thinking, etc.) that interferes in the formation of the relevant beliefs.Footnote ¹⁰

In a nutshell, then, since the virtue theorist conceives of epistemic normativity as sourced in an agent’s competences, and since Cases 1–7 are cases of incompetent belief formation by stipulation, the virtue theorist has difficulties explaining the datum at hand.Footnote ¹¹

2.5 Conclusion

This chapter has examined the prospects of some of the most popular contemporary internalist and externalist theories of evidence, reasons to believe, and propositional warrant to account for what is going wrong in cases of resistance to evidence. I have first argued that evidence internalism suffers from in-principle difficulties. Further on, I have shown that the belief condition on evidence possession generates inescapable difficulties for the E = K view, according to which one’s evidence is one’s knowledge. Still further on, I looked into indicator reliabilism, and I found that it lacks the normative resources needed to explain resistance to evidence as it predicts – against intuition – that biased cognisers lack evidence speaking against their biased beliefs just in virtue of dogmatically ignoring it. Finally, I have examined virtue epistemological accounts of reasons and propositional warrant, and I found a common culprit that prevents these accounts from accommodating impermissible resistance to evidence: on these views, epistemic virtues constitute the bedrock of epistemic normativity. Unsurprisingly, when virtues are missing or inactive in the case at stake, there are no normative resources available to explain epistemic impermissibility. Since that is precisely what is the case in resistance cases, I have argued, virtue epistemology is too agent-centric to accommodate the phenomenon we are discussing.