1. Introduction: Kinds of defeaters

Epistemic defeat is a familiar feature of our cognitive lives. Just as acquiring new information can lead to our having better justified beliefs, new information can also take that justification away. For example, suppose you saw your colleague at work an hour ago, and he tells you that he will be in the office for the rest of the day. On that basis, you believe what he says. Such a belief is epistemically justified. But now, suppose that a trusted, reliable authority tells you that they just saw that colleague leave the office and board a bus. You now believe the trusted, reliable authority. By believing their testimony, many want to say that you have thereby acquired a defeater for your belief that your colleague will be in the office all day, meaning that the belief is no longer justified (or at least it is now less justified than it was before).

It has become common to distinguish between different kinds of defeaters, each individuated in terms of how they defeat: in term of being believed, in terms of being what the subject should believe, and in terms of being true. This corresponds to the traditional three-way distinction between mental state, normative, and factual defeatersFootnote ¹. Mental state defeaters are understood to be a subject’s mental states and, in Jennifer Lackey’s words, are “a doubt or belief that is had by S and that indicates that S’s belief that p is either false or unreliably formed or sustained” (Reference Lackey2008, 44). These defeaters function in virtue of being had by the subject, and generally are taken to function regardless of their truth-value or epistemic status.Footnote ² A reliable and trusted informant telling you that you are experiencing an optical illusion is an example of such a defeater. If you believe what the informant says, this belief can defeat the justification your perceptual belief might have otherwise enjoyed. The case above about believing that a colleague will be in the office all day, which is then contradicted by the testimony of someone else you believe, is another example of acquiring a mental state defeater.

Sometimes justification is defeated not because of a belief that someone actually holds, but because of a belief that they should hold. This is a normative defeater, characterized as “a doubt or belief that S ought to have that indicates that S’s belief that p is either false or unreliably formed or sustained” (Reference Lackey2008, 44). Like mental state defeaters, these also function regardless of their truth-value or their epistemic status. For instance, suppose that due to wishful thinking and self-deception one cannot bring oneself to accept overwhelming evidence one has that counts against what one believes. This evidence might nonetheless constitute a normative defeater that rebuts or undermines one’s justification.

Finally, a factual defeater is commonly understood thus: “a true proposition such that if it were added to S’s belief system, then S would no longer be justified in believing that p.” (Lackey, Reference Lackey2008, 45–46n21) Unlike mental state and normative defeaters, this kind of defeater functions in virtue of being true. Factual defeaters were first introduced as part of an early diagnosis and response to the Gettier problem (1963) (e.g., Lehrer and Paxon, Reference Lehrer and Paxon1969).Footnote ³ Here is an updated version of a classic case that illustrates factual defeaters:

Notice that “Linus does not own a Lamborghini” is a true proposition such that if Dr. Lamb were to believe it, she would no longer be justified in believing as she does. Crucially, advocates of factual defeaters hold that the reason Dr. Lamb does not know that Linus owns the Lamborghini is because her belief’s justification is defeated by this true proposition. As such, they contend that this true proposition constitutes a factual defeater which explains why knowledge is absent.

In addition to three traditional kinds of defeaters, are there also social defeaters? Or if not a distinct kind of defeater, is there a kind of uniquely social defeat phenomenon that falls under one of the three general kinds distinguished above? Some have come to think so, convinced by cases given by Gilbert Harman which purport to show that one can have a justified true belief that fails to amount to knowledge because of the social environment. Here is one such case to illustrate:

From this case Harman concludes that: “This reveals the social aspect of knowledge. The evidence that a person has is not always all the evidence relevant to whether he knows. Someone else’s information may also be relevant” (172–73).

My aim in this paper is to explore the questions of whether there are social defeaters, or else a kind of uniquely social defeat phenomenon that falls under one of the three general kinds of defeaters distinguished above. I shall argue that despite initial appearances to the contrary, we have no reason to accept a special class of social defeater, nor any essentially social defeat phenomenon. We can explain putative cases of social defeat with our existing epistemological apparatus, including appealing only to the three kinds of defeaters commonly postulated. So while justification and knowledge undoubtedly have social dimensions, I shall argue that we have yet to see that there is an inherently social form of defeat.

2. What would it take for a defeater to be social?

It is not immediately obvious what would make something a social defeater, but we can begin by ruling out the idea that if a defeater has a social source, that would be enough to make it a social defeater in any interesting sense. For instance, suppose I believe that p, but you then tell me something I know to be inconsistent with p and I believe what you say. Plausibly I now have a defeater for my belief that p and the source of this defeater, your testimony, is a social one. Nevertheless, this is insufficient to make this a social defeater. Rather, it is just an instance of a regular mental state defeater.

What about Gilbert Harman’s widely discussed assassination cases? Might these help us identify what it would take for a defeater to be social? After all, these have been taken to be among the paradigm examples used to motivate the position that knowledge and justification are vulnerable to a kind of inherently social defeat.Footnote ⁴ Here is a variation of the case:

As in the first version of the assassination case, Harman suggests that this case also reveals a social aspect of knowledge. Other philosophers have followed suit in holding that cases like these reveal that social factors can destroy knowledge, although their analyses of how this happens differ (e.g., Pollock, Reference Pollock1986, 190–93; Meeker, Reference Meeker2004). I shall argue that there is nothing inherently social in this kind of case that requires special epistemological treatment. I propose a disjunctive analysis of the assassination cases: either they are cases of knowledge after all, and the luck present is a benign form of evidential luck, or else if knowledge is lacking, then this can also be explained by appealing to the notion of epistemic luck, without making essential use of anything social. I shall explain each of these options in turn.

As to the first proposal, I grant that one might be initially tempted to deny that Jill in Version 2 of the case, for example, knows that the political leader has been assassinated (I suggest that the same disjunctive analysis can be applied, mutatis mutandis, to ASSASSINATION Version 1). After all, it is (in some sense) a matter of luck that she has the justified true belief that she does, and knowledge is essentially nonlucky (that is surely a principle lesson of the Gettier problem). Jill could very easily have seen the misleading TV news broadcast and, in that case, she would believe that the leader survived the assassination attempt. So her justified true belief that the political leader was assassinated is too lucky to amount to knowledge, one might think.

But upon further reflection, it is clear that there are instances of lucky knowledge. We should not conclude from the Gettier problem that knowledge excludes any and all kinds of luck. Rather, knowledge is compatible with certain kinds of luck and not others. For example, Duncan Pritchard (Reference Pritchard2005), following Peter Unger (Reference Unger1968), has argued that knowledge is incompatible with specifically veritic luck, whereas what has been called evidential luck, by contrast, is no threat to knowledge at all. Very roughly, a belief is veritically lucky if, given how it was formed, the belief could have very easily been false. A belief is evidentially lucky, on the other hand, if it is a matter of luck that one has the evidence that one does (i.e., while one has the evidence that one does in the actual world, there are nearby worlds where one lacks that evidence).

Applying this distinction between different types of epistemic luck to Harman’s ASSASSINATION Version 2 case, for example, I suggest that we should hold that the kind of luck Jill’s belief enjoys is epistemically benign evidential luck. She is of course very lucky to have the evidence that she does, namely, the reliably produced accurate newspaper report. Many other people lack that evidence, and we can suppose that she herself might easily not have had it—she could easily have seen the TV news broadcast instead. But given the evidence she actually has, it is no accident at all that her belief is both justified and true. In short, I suggest that contrary to Harman’s original judgment, and contrary to possible initial appearances otherwise, Jill does know that the political leader has been assassinated. She does not lack knowledge, although obviously she almost does. Jill knows luckily, but not problematically so.Footnote ⁵

Suppose that despite the above considerations, one agrees with Harman that Jill lacks knowledge. Does that motivate the hypothesis that knowledge and justification are inherently social as Harman contends? I suggest that it does not. Appealing to the same distinction drawn above between veritic and evidential luck, we still have the resources to explain the judgment that Jill lacks knowledge without having to appeal to social considerations. It is common among those who agree that knowledge excludes too much veritic luck to formulate antiluck conditions that are designed to rule out such luck. Take the safety condition, which has become one of the more popular accounts of the antiluck condition (e.g., Pritchard Reference Pritchard2005).Footnote ⁶

Safety states, roughly, that it is a necessary condition of S knowing that p that, given how the belief was formed, it could not easily have been false. Crucially, safety must be relativized to methods or bases. For instance, safety requires that if the subject were to form their belief in nearby worlds on the same basis as they do in the actual world, the belief would still be true.Footnote ⁷

Applying this to Version 2 of Harman’s assassination case, it is important to consider on what basis Jill forms her belief that the president was assassinated. If the method is construed narrowly as the belief being held on the basis of the newspaper report, as I suggested above, then the belief is safe. If, on the other hand, one takes the method she is using as relying upon trustworthy news sources generally,Footnote ⁸ then her belief is unsafe. This is because that method is unreliable given the nearby worlds where she forms her false belief about the president’s fate on the basis of the false TV news reports.

Notice that if this is the correct way of specifying the belief-forming method Jill is relying upon, as some will maintain, her belief that the president was assassinated is infected by too much veritic luck and so plausibly does not count as knowledge. But there is nothing inherently social here. There is simply too much veritic luck in the environment such that Jill’s belief is not safe, and so it does not amount to knowledge.

The upshot is this: we have the resources to explain why Harman’s assassination cases are ones of knowledge, but even if they are ones where knowledge is lacking, it is not genuine social defeat. The social sources are causally relevant in explaining why, for example, Jill forms the beliefs she does, but the social dimension is not itself epistemically salient.

In looking for a form of specifically social defeat, I suggest that we follow Sandy Goldberg’s recent proposal that social defeat, if there is such a thing, is a kind of normative defeat. Goldberg understands the notion most generally as follows: “doctrine of normative defeat: if S should have known that p, then the proposition that p is a defeater for S” (Reference Goldberg2017, 2893). While I agree that this is a central kind of normative defeat, this construal is too narrow, as there are surely kinds of normative defeat that need not involve knowledge (in the sense that the subject should-have-known that p). For instance, if you have justification to believe that p, but you believe that not-p, then p is a normative defeater for you (and presumably the same is true for other epistemic statuses).Footnote ⁹ So generalizing somewhat, if S (epistemically) ought to believe that p, then p is a normative defeater for S.

So how, or under what conditions, might there be a kind of normative defeat that is inherently social? After all, clearly not all normative defeat is social defeat, for example, in cases involving only oneself. For instance, to return to an earlier example, if one cannot bring oneself to accept overwhelming evidence one has that counts against what one believes, this evidence will constitute a normative defeater that rebuts or undermines one’s justification. No other agents need be involved. Goldberg’s suggestion is that social defeat is a species of normative defeat because it arises from an essentially social should-have-known phenomenon that is grounded in social expectations. It is to that proposal I now turn.

3. Goldberg’s analysis of social defeat: should-have-known

Goldberg is not primarily interested in cases where someone should have known that p given the evidence they have. That would be a straightforward case of either mental state or normative defeat and need not in any way be social. Rather, Goldberg’s primary focus is, very provocatively, on when one should have known even when the subject is not in a position to know. (Reference Goldberg2017, 2864) At first that claim might have a paradoxical ring: How could it be true that someone ought to know something even if there is no way that they could know it? Does “ought” not imply “can” after all?

To begin to motivate the most controversial and interesting versions of this claim, Goldberg begins with the least controversial examples. (Reference Graham, Lyons, Brown and Simion2017, 2865–67). In these examples, the subjects have professional responsibilities and duties to know certain things. If they fail to have the relevant knowledge, it is both true that they should have known, given their roles, and we are entitled to expect that they should have known. For example, a day-care worker, given their job, ought to know where the children under their care are located. If they do not know where one of the children they are responsible for is, it is both true that they should have known and that we are entitled to expect that they have this knowledge. And this is true even if the day-care worker is in no position to know where the child is (for instance, the child might have become lost on an excursion and, given that, the day-care worker now has no way of knowing where they are).

Goldberg also offers Hilary Kornblith’s case of epistemically culpable ignorance to motivate what he claims is the social dimension of when subjects should have known:

Here again, Jones should have known that his paper faces damning criticism and this, Goldberg contends, is because it is part of the physicist’s job to weigh all the relevant evidence for and against hypotheses—they have a professional responsibility to know if the hypotheses they accept are true or well supported. Goldberg’s central contention is that it is in virtue of legitimate social practices that we are entitled to certain epistemic expectations, and that these legitimate expectations ground the should-have-known phenomenon.

From examples of professional and other practice-generated entitlements, Goldberg generalizes as follows:

Goldberg offers a key illustration of this principle, one that is supposed to be a case of should-have-known that is rooted inherently only in social practices, and not in professional responsibilities or other institutions. Call this the case of Rodriguez/Yang:

Let us grant for the sake of argument that our basic epistemic expectations capture our basic epistemic standards. Grant also that each of us is entitled to have these expectations when we engage with each other. Even grant that these expectations warrant should-have-known allegations. Does all that entail that it is true that the subject should have known when Goldberg’s SHK is met? If so, is that because it was socially expected of them? I shall argue for negative answers to both questions.

I agree that Rodriguez should have known what was on the grocery list, but I submit that this has nothing to do with social expectations. It matters why he fails to know, and this, I suggest, is because of his inattentiveness. A grocery list is normally written to remind one to buy all of the items listed on it. The goal in writing the list is to end up with (at least) everything on it. If one reads only part of the list, given one’s aim, one is very unlikely to meet this goal (but it is not impossible to meet this goal of course: one could randomly buy what just so happened to be on the list, or someone might add items to your shopping cart that unbeknownst to you were on the list, and so forth). Given the point of writing grocery lists, Rodriguez’s inattentiveness explains why he should have known what was on it.

To be clear, I am not contending that his inattentiveness is a defeater. Footnote ¹⁰ It is not that Rodriguez has a belief that Rodriguez and Yang need milk, eggs, and sugar, but that this belief is unjustified, or fails to amount to knowledge due to inattentiveness. Rather, his inattentiveness results in him not forming the target belief at all. This is a case of ignorance and one where Rodriguez should have known, but it is not one of defeat (normative or otherwise).

To see even more clearly why there is no relevant social dimension in play here, notice that if Rodriguez wrote himself a grocery list, the above points equally apply, which suggests it is not the social aspect of the Rodriguez/Yang case that makes a difference. If due to inattentiveness Rodriguez failed to read all the items on the grocery list he wrote himself, it is still true that he should have known all the items on it. Here again, the inattentiveness is not a defeater of justification or knowledge, but it explains his ignorance. Given Rodriguez’s goals and purposes in writing himself a grocery list, he should have known everything on it. In short, the Rodriguez/Yang case does not motivate the position that there is a variety of should-have-known, or normative defeat, that is essentially grounded in social practices.

Aside from the question of whether Goldberg’s central case motivates the role of social considerations in normative defeat, it is also important to note that there seems to be a gap in his overall approach between a principle like SHK and the sort of question we are meant to be asking. What we want to know is when is it true that someone should have known that p, so that if they fail to, p amounts to a normative defeater for them? But SHK is couched in terms of entitlement to expectations that someone knows. The conclusion of the Yang/Rodriguez argument is also explicitly framed in terms of what Yang is warranted in saying about whether Rodriguez should have known.

In the absence of further argument, it seems that when it is true that someone should have known can come apart from whether others are entitled to expect someone to know. Perhaps one is entitled to expect someone to know when one has justification to believe that they know. But evidence and appearances can be misleading. It might be perfectly reasonable to expect someone to know something without this entailing that they ought to know it.

By way of illustration, suppose that you and I are both new in town, but I do not know this about you. Instead, I am told by someone I know to be a local that you were born and raised here, and I have no reason to distrust them. On this basis I might then be perfectly well entitled to expect that you know lots of things about the city: its geography, its best restaurants, and the like. Despite this entitled expectation, surely it does not follow that it is true that you ought to know these things. Like me, you too are new in town; since it is not true that I ought to know these things given that I just arrived, neither should you.

Likewise, granting that Yang might be warranted in saying that Rodriguez should have known is not equivalent to saying that it is true that Rodriguez should have known. Here too one’s evidence might overwhelmingly warrant one in believing that the subject should have known without it being true that they should have known. Again, it is possible that the evidence is undetectably misleading or incomplete in various ways. In the above example, I might be warranted in saying that you should know the city’s landmarks, given that I justifiedly believe you are a local, without it being true that you ought to have this knowledge.

In response, one might claim that there is no room for the distinction I want to draw between when it is true that someone should have known and when others are entitled to expect them to have known since being entitled to expect S to X entails that S ought to X. Footnote ¹¹ Whether this objection is sound will, of course, turn on what it is to be entitled to believe or expect something.

Goldberg says that the practice-generated entitlements he is interested in must be distinguished from epistemic entitlements (Reference Goldberg2017, 2867; 2018, 166). And that is all for the good because on all dominant accounts of epistemic entitlement (e.g., Dretske Reference Dretske2000; Burge Reference Burge2003; Wright Reference Wright2004; Peacocke Reference Peacocke2004), it is not factive: being entitled to hold a belief (and presumably by extension an expectation) does not entail that the belief is true, and so presumably it would not follow from these accounts of entitlement alone that being entitled to expect S to X entails that S ought to X. Footnote ¹² Entitlements in this regard are more like justification and less like knowledge (while entitlement is nevertheless a notion distinct from justification, according to Dretske, Burge, and Wright; Peacocke, on the other hand, seems to use “entitlement” and “justification” interchangeably).

So if this objection stands, another notion of entitlement is needed. Goldberg does offer a proposal, but it cannot simply be assumed or stipulated that being entitled to expect S to X entails that S ought to X. The onus is on Goldberg to offer such an account that implies this entailment, but this is not something that he has done in this connection. The account of practice-generated entitlement that Goldberg sketches does not have the result that being entitled to expect S to X entails that S ought to X.

Goldberg offers this general analysis of the relevant sense of when a subject has a practice-generated entitlement to an expectation:

PGE is supposed to be wholly general, with the entitled expectations regarding when others should have known only a specific instance of this more general principle. Goldberg motivates and illustrates PGE with the following (nonepistemic) example:

Here PGE is satisfied: Melissa is entitled to expect her bins to be emptied since she participates in the practice of garbage collection, the practice is legitimate, and calling into question the legitimacy of her expectation would call into question the legitimacy of the practice of garbage collection or the standard that it is collected on Wednesdays. I grant that Melissa is entitled to expect her garbage to be collected. What I contest is that this entails that the sanitation workers ought to have collected her garbage.

If the sanitation workers saw her bins but did not empty them, then it is true that they ought to have done so. If they did not see the bins because they were playing on their phones as they passed by Melissa’s house, rather than doing their jobs, then I also grant that they should have emptied her bins. But suppose instead that through no fault of their own, no sanitation workers saw the bins.Footnote ¹³ Perhaps the bins were incredibly well camouflaged by local pranksters such that they were invisible to passersby. In this case, I submit, it is not true that the sanitation workers ought to have collected the garbage.

Or take a more dramatic example: suppose that the garbage truck exploded on the way to Melissa’s house after picking up a box of discarded landmines, tragically killing all the sanitation workers.Footnote ¹⁴ Again, I submit it is not true that the sanitation workers ought to have collected the garbage. Assuming that ought implies can, then given that there is no way that they could have collected the garbage (in view of their being dead), it is hard to see how it could be true that they ought to have collected it. None of this detracts from the fact that Melissa would be entitled to expect that her garbage bins would be emptied.

It is not merely that the sanitation workers cannot be blamed in these cases for failing to collect the garbage. Blamelessness is not equivalent to justification.Footnote ¹⁵ Rather, the workers cannot be properly criticized in any sense for failing to meet the norm of collecting all the garbage in Melissa’s neighborhood on Wednesdays, which is good evidence that it is not true that they ought to have collected it.Footnote ¹⁶ So even nonepistemic cases of practice-generated entitlements appear to admit to possible gaps between someone being entitled to expect S to X and when S ought to X. Further argument is needed to close the gap. In the absence of such an account and given the examples I offered above illustrating how it might be perfectly reasonable to expect someone to know something without this entailing that they ought to know it, my argument against Goldberg’s account of the should-have-known phenomena stands.

What is the cost of rejecting Goldberg’s should-have-known analysis? Goldberg says that if we reject his SHK, then we need to reject this plausible principle:

I grant, however, that this principle is true. But notice that it only says when one is warranted in criticizing S for not knowing. It does not say anything about when it is true that S should have known. If you are entitled to expect someone to know something but they do not, then their ignorance warrants criticism. But I have been arguing that it does not follow from the fact that one is entitled to expect a particular person to know, that they should know (so that if they do not know, then they should have known).

Goldberg thinks that there is another cost in rejecting his analysis. If we reject the analysis, then what should we say about Kornblith’s objection-ignoring scientist, the doctor who failed to keep up with the journals she ought to have kept up with, the day-care worked who failed to read the allergy report of the new child in his care, etc.? Do we have to count them as having justified beliefs if we stipulate that their beliefs conform to the evidence that they currently possess? While the cases differ, I agree that in all of the cases Goldberg discusses the subjects lack justification. But this is for various reasons, none of which are social.

In some cases, the subject mishandles the evidence they have. In others, like the arrogant scientist or inattentive Rodriguez, the subjects do not have the evidence they ought to have, or sometimes they do not even have the beliefs they ought to have, due to intellectual vice (e.g., arrogance, closed mindedness, inattentiveness, etc.). If the doctor failed to keep up with the journals that they ought to have, it is relevant why they did not. If it was due to arrogance, intellectual laziness, or insouciance, then their failure to know will also have an explanation stemming from their epistemic vices.Footnote ¹⁷

In some cases, like the doctor or day-care worker, while the subjects are blameworthy and rightly subject to criticism in some sense, it is not obvious that their failing, or their only failings, are purely epistemic. Ignorance is not always in itself an epistemic failing. Doctors and day-care workers have professional and moral obligations to know various things. Accordingly, their failings to have the relevant knowledge in these cases may be professional and moral failings.

In light of the fact that there appears to be a gap between when it is true that someone should have known and whether others are entitled to expect someone to know, or when one might be warranted in saying that someone should have known, let us explicitly focus on the former issue. We can now ask: Is social expectation itself sufficient to ground the truth of when someone should have known?

4. Social expectations

Let us consider another well-known case Harman used to motivate the thesis that social considerations can make a difference to whether or not a subject has knowledge.

Harman’s own analysis of this kind of case is that the reason that you do not know that Donald is in Italy is because of the ease of obtaining the postal evidence. Harman writes of the same sort of case, but with different character names:

Commenting on Harman’s postal cases and his analysis of them, John Pollock notes, rightly, that it is not the mere easy possibility of obtaining the postal evidence that makes it a defeater since if the trick letter from San Francisco had been hidden under the doormat, the subject would still know that Donald is in Italy, despite the letter still being easily available (Reference Pollock1986, 192). Looking under one’s doormat is easy enough to do and so is readily available. But that is not enough to make information hidden under one’s doormat a defeater.

Instead, Pollock suggests that a defeater arises in Harman’s postal case because of social expectations. Pollock writes,

Using the notion of social expectations, Pollock goes on to propose an account of social defeat. Pollock holds that a proposition is a social defeater for S if and only if it is of a sort S is expected to believe when true (192). Similar to Goldberg’s account which follows it, it is the violation of a relevant social expectation that is thought to defeat knowledge.

On Pollock’s analysis of his variation of the postal case where the trick letter is hidden under the doormat, the hidden letter does not defeat in our social environment despite being easily available. This is because checking under our doormats is not something we are socially expected to do. On the other hand, in a society where everyone is expected to check under their doormats when they return home, the hidden letter would defeat your knowledge that Donald is in Italy (Reference Pollock1986, 192).

But do we need a notion of social defeat to explain why knowledge is absent in Harman’s postal case? If we cannot explain the case without it, that would motivate postulating the existence of social defeat. But if we can explain the case without introducing a new kind of defeater or defeat phenomenon, then considerations of parsimony dictate that we should not. I suggest that we do not need a notion of social defeat to explain Harman’s postal case. I propose the same unified treatment here as the assassination cases above: either you do know that Donald is in Italy and it is a matter of evidential luck; or you do not know that Donald is in Italy and that is because there is too much veritic luck present. Either way, nothing inherently social is involved.

For my own part, I think a stronger case can be made for interpreting this example as an instance of knowledge. Like Harman’s assassination cases, it is important to consider on what basis you form the belief that Donald is in Italy. On the assumption that the belief is held on the basis of Donald’s testimony alone, the belief is safe. Your belief is in a sense lucky since you could have easily opened the misleading letters. But this appears to be a case of benign evidential luck: you are lucky to have the actual evidence you do since in very nearby worlds you have different total evidence which includes the misleading letters. But as noted above, evidential luck is compatible with knowledge.

On the other hand, suppose one takes one’s cue from Harman. Then the method used could be construed as including all easily available evidence, and so includes not only Donald’s testimonial evidence, but also the easily available misleading postal evidence. In that case, your belief is unsafe. This is because the method used is unreliable given the nearby worlds where you form a false belief about Donald’s whereabouts on the basis of the trick letters. If this is the correct way of specifying the belief-forming method you are relying upon, your belief that Donald is in Italy is infected by too much veritic luck and so plausibly does not count as knowledge. But here again, just as on one interpretation of the assassination cases, there is nothing inherently social. Your belief does not amount to knowledge since there is simply too much veritic luck in the environment such that your belief is not safe. So here, too, we have the resources to explain why Harman’s postal case is one of knowledge, but even if it is one of lacking knowledge, it is not because of any social factors.

Suppose one rejects my disjunctive analysis of the postal case and, following Pollock, persists in holding that it is due to social expectations operative in the case that knowledge is lacking. Even then, we should still consider: Are mere social expectations sufficient to defeat knowledge? I have already argued that it does not follow from the fact that one is socially expected to be aware of X that one ought to be aware of X. Furthermore, there is no straightforward connection, if any, between a mere social expectation and what one ought to believe. To illustrate that social expectations are insufficient to defeat knowledge, consider this case from Kevin Meeker:

Meeker’s AMANDA case is a counterexample to several of the accounts of social defeat we have considered so far. It is a counterexample to Harman’s account, since no post under Amanda’s doormat would constitute a defeater for her despite being easily available to her. It is a counterexample to Pollock’s account since no post under Amanda’s doormat would constitute a defeater for her despite the societal expectation that she check under her doormat for mail and other information. This case is also a counterexample to Goldberg’s SHK principle: it is not the case that Amanda should have known what was under her doormat even though the society she finds herself in is entitled to expect that she knows what is under the mat. The case helpfully illustrates the gap between when it is true that Amanda should have known and when others might be entitled to make the charge that she should have known.

In response to his AMANDA case, Meeker introduces the notion of a legitimate social expectation (e.g., Reference Meeker2004, 164–65; Goldberg also makes use of the notion of legitimate social practices, e.g., Reference Goldberg2018, 170). Meeker’s proposal is that defeaters can arise from legitimate social expectations in contrast to any old social expectations. A natural question, of course, is what makes a social practice and its associated expectations legitimate?

5. Legitimate social expectations

The case of Amanda illustrates that mere social expectations are insufficient to defeat knowledge. In light of this, the suggestion to consider is that only legitimate social expectations can have this effect. In light of the Amanda case, Meeker refines his account of social defeat thus:

Again, the pressing question is: What makes a social practice and its associated expectations legitimate? Given that the notion of legitimacy is the crux of his proposal, Meeker, somewhat unsatisfyingly, does not attempt to provide an account.

Instead, Meeker claims that recognizing the legitimacy of social expectations is on par with recognizing instances of knowledge (165). To adopt such an approach is to endorse a kind of Chisholmian particularism: just as we can intuitively recognize instances of knowledge when we are presented with them, the suggestion is that we can know legitimate social expectations when we see them too.Footnote ¹⁸

I am sympathetic to particularism as a methodological approach in exploring the nature of knowledge and epistemic justification. But consistent with this kind of particularism is the thesis that we can attempt to formulate, at least partially, illuminating conditions for the instantiation of certain epistemic properties. I propose the same for elucidating the notion of legitimate social expectations. That is, given an intuitive idea of which social expectations are legitimate, we can seek to give an account of what all the intuitive particular cases have in common that make them instances of legitimate social expectations, rather than something else.

According to Meeker, it is intuitively obvious that the social expectation operative in the Amanda case is not legitimate. I agree. While he does not attempt to offer an account of legitimacy, his remarks on the case are suggestive. Of Amanda’s situation, Meeker writes:

Generalizing from Meeker’s comments, my suggestion is that the relevant kind of social expectation is legitimate only if you have, or ought to have, evidence for it. If Amanda had been informed of the custom of checking under doormats, she would then have evidence that this was expected of her. As Meeker rightly notes, if having been so informed, and so having evidence of the expectation, she failed to check under her doormat, she could be rightly criticized.

We can see that it is evidence doing the epistemic work, and nothing inherently social, by considering a case which combines Harman’s Post Case with Meeker’s AMANDA example, but with a further twist. Suppose that Amanda had a justified true belief that Donald is in Italy for the whole summer in the way Harman describes in his original case. Meanwhile, as the result of an emergency landing, Amanda finds herself in a hotel in a certain culture that expects people to check under their doormats for mail and other information, as in Meeker’s original case. Now for the twist: learning of Amanda’s location, Donald couriers the trick letters saying that he is in San Francisco to the hotel to be placed under Amanda’s doormat. Unlike in Meeker’s original case where the expectation was illegitimate, however, suppose that Amanda has been informed of the expectation to check under her doormat for post, but she has not done so.

As Amanda has been informed of the practice, she has evidence that she should check under her doormat for possible information. If she did check under her doormat, she would then have found the letter saying that Donald is in San Francisco for the summer. So the contents of the letter serve as a normative defeater for her belief that Donald is in Italy. That is, the contents of the letter amount to a belief that Amanda ought to have that indicates that her belief about Donald’s whereabouts is false or unreliably formed or sustained. And why ought Amanda have this postal evidence which would amount to a normative defeater? Not because it is merely socially expected of her to have looked under her doormat, as the original AMANDA case shows, but rather, because she has been given evidence that she ought to check for post and other information under her doormat.

But what is epistemically salient here is the evidence, not the social expectation; the evidence is the thing that makes the difference to what Amanda ought to believe. If this is correct, it is not the social expectation that makes it true that the subject should have known. Rather, it is simply that given the evidence the subject has, she should have known what was under her doormat. In addition, this would not be the kind of case Goldberg is interested in since it is just a straightforward case of the subject not responding properly to their evidence, not one of someone who should have known even when the subject is in no position to know.

Suppose instead that Amanda had been informed of the custom of checking under doormats, but she was so incredulous of what she took to be an outlandish practice, that she simply did not believe what she was told, despite being sincerely told of the practice by trustworthy locals. Still, she would have evidence of what was expected of her, so if she failed to look and find information under her doormat, it would have been true that she should have known. Given the evidence Amanda should have had had she not been so incredulous and mistrusting, she should have known. But then again this is just a regular (nonsocial) normative defeat case: one ought to have some evidence but does not.

Crucially, nothing about either of these two cases where Amanda is given evidence of the practice is inherently social. They simply involve evidence about something which happens to be social, namely a practice involving the transmission of information under doormats peculiar to this community. But having evidence about something social does not by itself make the social somehow epistemically important. Compare: if one had evidence of something political or aesthetic that was inconsistent with one’s political or aesthetic beliefs, we should not conclude that there is a special class of political or aesthetic defeaters, or kinds of uniquely political or aesthetic defeat phenomena. Any defeat, and the sense in which it is true that Amanda should have known in these cases, can be fully explained by the existence of mental state and normative defeaters. There is no need to appeal to a special class of social defeater, nor any essentially social defeat phenomenon.

Before concluding, it is worth pausing to examine Goldberg’s account of what makes a social practice legitimate. While Goldberg does not provide a full-fledged analysis, he does say more than Meeker does on this score. In the course of describing practice-generated entitlements, which I discussed above, Goldberg tends to say things like this about the legitimacy of social practices:

More formally, Goldberg defines the general requirements on the legitimacy of a practice (and so also a social practice) as follows:

What are we to make of requiring that it is a necessary condition of a practice being legitimate that it must be recognized and acknowledged by those within the practice?Footnote ¹⁹ My suggestion is that Goldberg’s concern is the same as Meeker’s when the latter claims that what made the expectation of Amanda to look under the doormat illegitimate was that she was not properly informed of it. Goldberg’s remarks on the importance of recognition and acknowledgment fit with my suggestion that the kind of social expectation at issue is legitimate only if you have, or ought to have, evidence for it. If Amanda had been informed of the custom of checking under doormats and acknowledged and recognized the practice, she would then have evidence that this was expected of her.

But what is epistemically salient here is the recognition and acknowledgment. Here again it seems that what is important is evidence and not the social expectation per se. If this is correct, it remains the case that it is not the social expectation that makes it true that the subject should have known. Rather, it is simply that given the evidence the subject has, they should have known. But then again, this is simply a case of a regular mental-state defeater.

On the other hand, if the social practice was widely acknowledged and recognized but Amanda just failed to give it any proper credence, she would still have evidence of what was expected of her. If she then failed to look and find information under her doormat, it would have been true that she should have known. But then this too is just a regular (nonsocial) normative defeat case: she ought to have some evidence but does not. Any defeat, and the sense in which it is true that Amanda should have known in these cases, can be fully explained by the existence of mental state and normative defeaters without the appeal to anything social.

In sum, my aim in this article has been to explore the questions of whether there are social defeaters, or else a kind of uniquely social defeat phenomenon that falls under one of the three general kinds of defeaters that epistemologists tend to admit. I argued that despite initial appearances to the contrary motivated by a number of interesting and influential cases, we have been given no reason to accept a special class of social defeater, nor any essentially social defeat phenomenon. Rather, we can explain putative cases of social defeat with our existing epistemological machinery. So while justification and knowledge undoubtedly have social dimensions, we have yet to see that there is an inherently social form of defeat.