This chapter considers one popular way to account for cases of resistance as cases of evidence one should have had, where the normative failure at stake is taken to be either (1) a breach of social normativity (Goldberg Reference Goldberg2018) or (2) a breach of moral normativity (Feldman Reference Feldman, Conee and Feldman2004). I argue that the social normative option is too weak, in that it allows problematic social norms to encroach on epistemic normativity, and that the appeal to moral oughts fails both on theoretical grounds – in that it cannot accommodate widely accepted epistemic conditions on moral blame – and on extensional adequacy.
3.1 The Social ‘Should’
In recent work, Sandy Goldberg has taken up the task of developing an account of the normativity of evidence one should have had and normative defeat. One key thought that motivates Goldberg’s project is that social roles – for instance, being a medical doctor – come with normative expectations. These normative expectations may be, and often enough are, epistemic. For instance, there is a social epistemic expectation that medical doctors are up to speed on the relevant literature in their field. Another key thought is that to believe that p justifiably one must live up to all of these legitimate expectations. Doctors who fail to be up to speed with the most recent research in their field are not justified in their corresponding beliefs, in virtue of being in breach of the social expectation associated with their role. For instance, a doctor who believes that stomach ulcers are caused by stress, in ignorance of the widely available evidence that suggests that they are caused by bacteria, is not justified to believe that ulcers are caused by stress. As such, Goldberg grounds the normativity of evidence one should have had in the social expectations associated with the believer in question’s social role.
It is easy to see that Goldberg’s key thoughts promise to give us the ideal resources to handle cases of resistance to evidence. Take, for instance, the case of Professor Racist from Chapter 1. Recall that this fellow is biased against people of colour, and, 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. In virtue of occupying his social role, by the first key thought, Professor Racist is subject to normative expectations that are associated with this social role. In particular, he is subject to the expectation to fairly distribute his care and attention in the student population. Since Professor Racist doesn’t live up to this expectation, by the second key thought, he does not believe justifiably that the Black students are not active in his class.
This is a very rough description of how Goldberg aims to deal with the kind of cases of resistance to evidence that we are concerned with here. Even so, here is a worry that arises immediately: social expectations can be legitimate social expectations, but also illegitimate social expectations. Women, for instance, are often illegitimately expected to carry most of the household burden and to underperform in leadership roles. If so, it would seem as though social expectations cannot play the normative grounding role that Goldberg wants them to play, since they seem to require further normative unpacking themselves: we seem to need further normative notions to help distinguish between epistemically legitimate and epistemically illegitimate social expectations.
If so, the question that arises is: aren’t the social epistemic expectations that Goldberg appeals to in order to explain intuitive epistemic failure grounded in epistemic norms? And if so, won’t we have to invoke the relevant epistemic norms in the final analysis of what goes wrong in cases of resistance to evidence? As a result, doesn’t Goldberg’s story remain very much at the surface, too much so to offer a satisfactory account of resistance cases?
To see why one might think this, suppose that social epistemic expectations are grounded in epistemic norms. If so, the reason why there is a social epistemic expectation that doctors be up to speed with the literature is grounded in an epistemic norm that applies to doctors and that requires them to be up to speed with the literature. Crucially, however, it is precisely these epistemic norms that we need to explain if we are to give a satisfactory account of the epistemic impermissibility of resistance to evidence. It may appear, then, that Goldberg’s treatment of evidence one should have had does little more than appeal to a symptom of the norms that need to be explained by an adequate account of what is wrong with resistance to evidence. As a result, it may also appear that Goldberg’s treatment doesn’t cut deep enough to offer a satisfactory account of evidence resistance.
While this worry seems prima facie legitimate, it is ultimately unfounded. The reason for this is that Goldberg develops a view that reverses the standard direction of explanation between norms and expectations. According to Goldberg, epistemic norms are explained in terms of social epistemic expectations rather than the other way around. If Goldberg is right about this, the above worry can be laid to rest. His account cuts exactly as deep as it needs to.
At the same time, a lot hinges on the credentials of Goldberg’s account of epistemic norms. Goldberg defends a view he calls ‘coherence-infused reliabilism’. According to this view, very roughly, one’s belief that p is prima facie proper if and only if it is held by a process that one is permitted to rely on and that satisfies a reliability and a minimal coherence-checking condition.
Goldberg observes that we are deeply social creatures who are engaged in practices of information sharing and joint action. These practices are supported by a rationale in that opting out of them would be practically irrational for us. Crucially, these can only be supported by this kind of rationale if we are entitled to certain expectations. More specifically, Goldberg argues that we must be entitled to expect others to live up to the requirements of coherence-infused reliabilism. We must expect them not to form beliefs in unreliable ways, and we must expect them to ensure coherence. If we couldn’t expect them to form their beliefs in these ways, it would not be rational for us to engage in the kind of cooperative ventures in which we rely on the truth of others’ beliefs for success. For instance, suppose you and I wanted to move a sofa. If you couldn’t expect me to reliably form beliefs about where the sofa is and not to have incoherent beliefs about the matter, it would not make sense for you to embark on this venture with me.
In this way, the fact that we are engaged in information sharing and joint practical ventures and the fact that there is a rationale for this presuppose that we are entitled to have certain expectations of one another. These expectations ground epistemic norms. In particular, one important norm that they ground is the norm that specifies the conditions for prima facie proper belief.
What about ultima facie proper belief and evidence one should have had? To explain cases like these, Goldberg appeals to general expectations that go beyond the explicit normative criteria at issue in prima facie proper belief and that may serve to disqualify a prima facie proper belief from being ultima facie proper.
Goldberg argues that there is independent reason for thinking that these general normative expectations do exist, and that they can and often do the work in the way that he needs them to do. By way of support, Goldberg considers a number of examples. Suppose your firm is hiring, and you are currently interviewing a number of applicants. The explicit criteria for the job provide one important standard for your evaluation, perhaps the most important one. However, beyond the explicitly stated criteria, there are also general expectations, including, for example, that candidates be appropriately dressed. Or suppose that you are on the committee that awards the Nobel Prize. Again, while the explicit criteria for your evaluation provide an important standard for your evaluation, there are general expectations, including that nominees mustn’t be Nazis. Crucially, job applicants and Nobel Prize nominees who do not live up to these general expectations may be disqualified because they don’t. If I show up in flip-flops, shorts, and a vest to your job interview, you may not give me the job even if I meet all of the criteria explicitly mentioned in the job description. Similarly, even if a certain person produced amazing science, if it transpires that they are an all-out Nazi, they should not be awarded a Nobel Prize.
Goldberg’s thought is that we find these general normative expectations in cases of epistemic assessment, too. Most importantly for present purposes, one relevant expectation is that one play one’s social epistemic roles properly. In the case of a medical doctor, to play this role properly is to remain up to speed with the relevant literature. As a result, while a doctor who fails to do so may satisfy the conditions for prima facie proper belief that p, they fail to live up to the general normative expectations that come with their role as a practicing doctor.
Similarly, the thought could go, there are social epistemic expectations on taking up easily available evidence that explain the impermissibility intuition in the resistance cases. The characters in Cases 1–7 fail to live up to these social expectations. In turn, this failure disqualifies their beliefs from being proper, just as the underdressed job applicant was disqualified from getting the job and the Nazi scientist was disqualified from winning the Nobel Prize.
3.2 Worries for Social Epistemic Normativity
While Goldberg’s account may look promising at first glance, there is reason to think that it remains ultimately unsuccessful. I will argue that there are two main problems that Goldberg’s view encounters due to the social grounding of epistemic normativity: the first has to do with the scope of epistemic normativity; the second is a normative strength problem.
3.2.1 The Scope Problem
To bring the scope problem with Goldberg’s account into view, notice first that, since on this normative picture epistemic norms are grounded in social expectations, which, in turn, are grounded in reliability constraints that are cooperation-generated, the scope of epistemic normativity only reaches as far as our rationale-supported practices of information-sharing and joint action. This is a theoretically heavy burden to carry: it amounts to a claim that epistemic normativity strongly co-varies with a particular subset of practical normativity: since, plausibly, the rationality at stake in the information sharing and joint action that Goldberg appeals to is (or, at least, can be, and often will be) practical rationality, it will follow on Goldberg’s view that, for all x epistemic practices, x is epistemically permissible insofar as it is practically rational to the aim of information sharing and joint action. This is an extremely strong normative co-variance proposal.
To see why this is a problem, consider a society that has practices of sharing information and acting jointly on a wide range of issues. Suppose, furthermore, that these practices are supported by a rationale in the way envisaged by Goldberg. The result that we get is that members of this society are entitled to expect others to form beliefs reliably and minimally coherently on this range of issues. But now suppose that this society also has a practice of not sharing information and acting jointly on certain issues. To take an example that is close to home, let’s suppose that they don’t have the practice of sharing information and acting jointly on cases of sexual assault. Since there is no practice of sharing information and engaging in joint action, members of this society cannot expect others to form beliefs reliably and minimally coherently on this issue, nor to be sensitive to the corresponding testimonial evidence, at least not if Goldberg is right and this expectation is grounded in our practices of sharing information and joint action. But if it is practice-generated expectations that explain epistemic normative standards, the result that we get is that whatever epistemic norm there may be that requires (or at least permits) members of this society to trust the word of others will not extend to the word of victims of sexual assault. As a result, in this society, the word of victims of sexual assault need not be uptaken, nor can it defeat beliefs in the innocence of sexual predators. And that, clearly, is the wrong result. It cannot be that we diminish the epistemic status of the testimony of victims of sexual abuse simply by tuning up the degree of sexism in a society (no matter how many practical benefits the sexist practices in question may generate).
The problem Goldberg encounters here is grounded in the absence of certain social practices. A similar problem arises from the presence of bad social practices. Consider a community of agents that have a social practice of actively distrusting the testimony of victims of sexual assault. This practice not only fails to give rise to epistemic expectations – it also gives rise to bad epistemic expectations. For instance, one expectation that this practice gives rise to is that those who claim to have suffered sexual assault are not to be believed. If it is practice-generated expectations that explain epistemic normative standards, the result that we threaten to end up with here is that the word of victims of sexual assault can permissibly be disregarded (in other words, members of this community threaten to end up having standing defeaters for the word of victims of sexual assault simply as a result of having a bad social practice). And, of course, this result is even worse for Goldberg’s view.
Before moving on, I’d like to consider some rejoinders on behalf of Goldberg.
A first route of resisting this result that Goldberg might explore is that the practices of sharing information and acting jointly on a range of issues entitles you to have expectations that are universal rather than restricted to the range of issues in question.
Unfortunately, there is reason to think that this route is ultimately not viable. One reason for this is that legitimate social epistemic expectations of the kind Goldberg envisages will be environment dependent: it is legitimate for me to expect people to know a lot about the history of Eastern Europe if I’m in Eastern Europe, for instance, but less so if I’m in Canada. Since environments can restrict the issues on which one can have legitimate social epistemic expectations of others, it follows that our practices of sharing information and acting jointly only entitle one to expectations restricted to a range of issues rather than to universal expectations. And if Goldberg is right and it is certain expectations we are entitled to have that determine epistemic standards, then the reach of epistemic standards is limited also. By the same token, the prospects of resisting this problematic result by holding that the expectations have universal reach are not bright either.
One might wonder, secondly, whether Goldberg’s reliability constraint cannot help with this problem. After all, testimony from sexual assault victims is notably highly reliable.
Unfortunately, on Goldberg’s view, it cannot, for two reasons: first, because the reliability of a practice is not enough to warrant its existence on Goldberg’s view – it also needs to be grounded in the cooperation rationale. Since we can easily imagine a world where this is not so, we get the result that the absence of the practice of trusting victims of sexual assault is unproblematic. How about the practice of actively distrusting them? Can’t Goldberg insist that the practice of distrusting the word of victims of sexual assault is not reliable, nor supported by a rationale?
Again, the answer here is ‘no’. First, this is because disbelieving is not plausibly subject to reliability constraints in the way believing is: I can unproblematically fail to believe a lot of propositions that are true, whereas I cannot unproblematically believe a lot of propositions that are false. Second, this is because normativity is modal: even if this practice is not, as a matter of fact, supported by a rationale, insofar as the practice may be supported by a rationale – in that it may be practically irrational to opt out of it – it may, on Goldberg’s view, generate legitimate social expectations. Consider a world in which sexual assault is widespread in that most adult men engage in it. In that case, it may well be practically irrational for them to opt out of this practice. At the same time, it may also be practically irrational for women to opt out – say, because this opting out is punished severely. In addition, it may be that abandoning or changing the practice is practically catastrophic not just for each individual human, but for humanity as a whole. To take a particularly drastic illustration of this point, suppose there is a powerful evil demon who will extinguish all of humanity if they abandon the practice of distrusting the word of victims of sexual assault. It is easy to see, then, that even bad practices can be supported by a rationale in Goldberg’s sense, in that opting out individually or abandoning or changing the practice as a whole is practically irrational. By the same token, Goldberg cannot hope to avoid the problem even in its second incarnation by appealing to the absence of a rationale.
Again, the underlying problem for Goldberg’s view is the normative co-variance claim. On his account, epistemic normativity strongly co-varies with (a subset of) practical normativity: since, plausibly, the rationality at stake in information sharing and joint action that Goldberg appeals to is (or, at least, can be, and often will be) practical rationality, it will follow, on Goldberg’s view, that epistemic permissibility will co-vary with practically rationality to the aim of information sharing and joint action. Since we can easily imagine cases in which what is beneficial for information sharing and joint action departs from what is epistemically permissible, the view is bound to get such cases wrong.
Furthermore, note that one does not even have to come up with very far-fetched examples to illustrate this point. We do, as a matter of fact, live in a world where many societies have a practice of disbelieving women and people of colour. We can imagine that one might even come up with a practical rationale for these practices – having to do, for example, with division of labour. Nevertheless, gender- and race-based epistemic injustice remains epistemically problematic.
Furthermore, research in cognitive psychology (e.g. Nisbett and Ross Reference Nisbett and Ross1980, Kahneman et al. Reference Kahneman, Slovic and Tversky1982, Gilovich et al. Reference Gilovich, Griffin and Kahneman2002) notably indicates that human beings tend to rely on heuristics when engaged in probabilistic reasoning, with these heuristics making people prone to commit elementary probabilistic fallacies. Also, according to error management theory (Haselton and Buss Reference Haselton and Buss2000, Reference Haselton and Buss2009, Haselton and Nettle Reference Haselton and Nettle2006), the fallibility of human cognition, at least in many cases, is the result of natural selection. Evolutionary psychologists argue that, given the limited information and computational power with which organisms must contend, an inference mechanism can be advantageous if it often enough (for biological purposes, such as survival) draws accurate conclusions about real-world environments, and if it does so quickly and with little computational effort. The heuristics humans rely on in probabilistic reasoning, some of these psychologists maintain, are mechanisms of just that sort.
Note that it is plausible that these evolved epistemically deficient practices are beneficial for both biological and social evolution – otherwise, it seems implausible that they would have been selected to begin with. Indeed, it seems plausible that relying on heuristics like those discussed above will be beneficial to the aim of information sharing and joint action – due to limited information and computational power. If so, Goldberg’s view will predict epistemic permissibility in all of these cases of intuitive epistemic failure.
To put the worry in more theoretical terms, here is the problem: if our model predicts that epistemic norms are grounded in the rationality of our practices of information sharing and joint action, and if the latter are (very plausibly) aimed at the survival of our species, then our model predicts that epistemic norms will track survival norms. Our belief-producing processes, for instance, will only be as reliable as needed for survival. However, there is nothing to ensure that the socially and biologically set reliability threshold will coincide with the epistemically needed reliability threshold. That is, the threshold of reliability required for epistemic purposes may well be higher than what is needed for our practices of information sharing and joint action, and in turn for biological benefit. Socially and biologically reliable enough need not coincide with epistemically reliable enough. Similarly, epistemic norms for sensitivity to evidence and for evidence gathering may set the threshold for epistemic permissibility differently than social and practical norms for joint action and survival.
3.2.2 The Strength Problem
A second problem for Goldberg’s account concerns the strength of the resulting epistemic normative requirements. In particular, the fact that, on his account, epistemic normativity is sourced in social expectations generates failures of extensional adequacy due to it being too weak to capture the distinction between epistemic shoulds: that between the synchronic ‘should’ of epistemic justification and the diachronic ‘should’ of responsibility in inquiry.Footnote 1 Goldberg’s view shares this important theoretical lacuna with Williamson’s E = K. Again, proceeding responsibly in inquiry (e.g. pursuing worthwhile questions) is one thing; synchronically responding well to available evidence is another. However, plausibly, both are governed by epistemic shoulds and accompanied by the corresponding social expectations.
To see this, remember the slightly modified version of the Friendly Detective case from Chapter 2: this time around, Dave and his colleague, Greg, were sent to investigate the crime scene. 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. As a result, he does not believe that the butler did it. In contrast, as we’ve already seen, 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.
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, and they both fail to meet the social expectations associated with their roles. Compatibly, 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.
This problem is a normative strength problem for Goldberg’s view: an account in terms of social expectations is too weak to individuate the relevant epistemic normative demands, in that it overgeneralises. At the same time, this strength problem, coupled with the scope problem identified above, serves to further suggest that the main underlying issue is the normative co-variance claim between the social and the epistemic: we sometimes (practically rationally) socially expect people to phi when they epistemically shouldn’t phi, and, conversely, other times we fail to expect people to phi when they epistemically should phi. By the same token, social normativity seems ill-suited to accommodate the epistemic impermissibility of resistance data that we want explained.Footnote 2
3.3 Problems for the Moral ‘Should’
We have seen that Goldberg’s view, accounting for what is intuitively epistemically amiss in resistance cases in terms of evidence one (socially) should have had, runs into trouble due to its underlying strong normative co-variance claim for the epistemic and the practical.
One might wonder, alternatively, whether the resistance cases we have been worried about aren’t really cases of moral failure rather than cases of genuine epistemic failure to begin with. On this account of the data, the intuition of impropriety in the resistance cases has a non-epistemic normative source: we think, for instance, that George is doing something wrong in the Testimonial Injustice case because he’s doing something morally wrong in not listening to the female passer-by: epistemic injustice, the thought would go, is the stuff of intellectual ethics, not of theory of knowledge proper. However, our intuitions are not fine grained enough to see the difference: theory is needed. Indeed, here is Ernie Sosa on this topic:
[T]he theory of knowledge […] is the department wherein we find the core issues of knowledge […] in the history of epistemology, by contrast with the wisdom of inquiry, and with the intellectual ethics wherein we find issues of epistemic justice and epistemic vice, broadly conceived.
Here also is Richard Feldman:
It’s surely true that there are times when one would be best off finding new evidence. But this always turns on what options one has, what one cares about, and other non-epistemic factors. As I see it, these are prudential or moral matters, not strictly epistemic matters.
I don’t find this move particularly plausible: the failure in question in Cases 1–7 is a genuinely epistemic failure. Here are a few reasons to think so: first, it is hard to see how, in the cases that exhibit morally problematic features, these could be instantiated without bad epistemic underpinnings. After all, one thing that the vast majority of the theorists of blameFootnote 3 strongly agree with is that there is an epistemic condition on moral blame: very roughly, moral blameworthiness implies that one is not epistemically blamelessly ignorant that one is doing something wrong. But this suggests that in the morally pregnant cases above, for example, the sexist and the racist are doing something epistemically wrong as well. Otherwise, if they were epistemically blameless, they could not be morally blameworthy. But they are.
Second, and most crucially, while some of these cases exhibit ethically problematic features, others do not. To the contrary, some of these cases (e.g. the case of Mary the wishful thinker and that of the friendly detective) can be plausibly construed as cases of moral success while remaining intuitively problematic with regard to the lack of evidence uptake. This suggest that the source of the intuition is, indeed, epistemic failure (absent other normative constraints at the context). Take, for instance, the case of Mary, the optimistic spouse: when her partner, Dan, spends more and more evening hours at the office, she’s happy that his career is going so well. When he comes home smelling like floral perfume, she compliments his taste in fragrance. Finally, when she repeatedly sees him having coffee in town with his colleague, Alice, she is glad he’s making new friends. She never considers the question as to whether Dan is having an affair. Is Mary justified to believe as she does that Dan is a faithful, loving husband? Clearly not. Note, however, that it’s hard to find moral flaws with Mary’s epistemic ways: after all, many moral philosophers (and a good number of epistemologists, e.g. Stroud 2006) agree that we owe more trust to our friends and family than to people we have never met: if so, Mary’s suspension is morally impeccable but epistemically problematic.
3.4 Conclusion
This chapter has looked into the option of explaining the impermissibility datum in resistance cases via appeal to social or moral normativity. I have argued that a social expectations-based account of the epistemic impermissibility of resistance is too weak to explain cases of epistemically bad social expectations, sourced in practical considerations pertaining to cooperation. Further on, I looked at the plausibility of explaining resistance cases away as pertaining to the moral rather than the epistemic domain, and I argued this doesn’t work on both theoretical and empirical grounds: first, a view like this fails to accommodate a widely accepted epistemic condition on moral responsibility. Second, since some of the resistance cases we have been looking at are cases of clear moral success, the view will be unsatisfactory on grounds of extensional adequacy.
Open access