Contextualist treatments of clashes of intuitions can allow that two apparently conflicting claims can both be true. But making true claims is far from the only thing that matters—there are often substantive normative questions about what contextual parameters are appropriate to a given conversational situation. This paper foregrounds the importance of the social power to set contextual standards and how it relates to injustice and oppression, introducing a phenomenon I call “contextual injustice,” which has to do with the unjust manipulation of conversational parameters in context-sensitive discourse. My central example applies contextualism about knowledge ascriptions to questions about knowledge regarding sexual assault allegations, but I will also discuss parallel dynamics in other examples of context-sensitive language involving politically significant terms, including gender terms. The discussion further illustrates some of the deep connections between language, epistemology, and social justice.
Language, as we know, carries power. In this paper I’ll buttress and apply this familiar point by looking to cases concerning the manipulation of conversational standards that affect the truth conditions of context-sensitive sentences. I’ll identify, explain, and apply a notion of contextual injustice—the unjust setting of disadvantageous contextual parameters. There are many kinds of contextual injustices; my primary focus will have to do with contextualism about knowledge ascriptions. I’ll also discuss contextualist approaches to gender terms like ‘woman.’
I’ll begin by laying out two philosophical disputes—one about knowledge and one about gender—in a way that will emphasize the political relevance of many philosophical questions. In §§2–4 I articulate a contextualist response to these disputes, and I formulate a triviality worry for such responses. According to this worry, contextualism implausibly [End Page 1] implies that one’s choice of what to say is too arbitrary—since one could speak truly in any of a variety of ways, we lack substantive grounds on which one might choose between them. I reject this worry. In §5 I argue that some contextual parameters are better than others, and §§6–7 explain a way in which unduly high epistemic standards can serve oppressive social structures. §8 articulates the central notion of contextual injustice. §9 returns focus to gender terms and relates contextual injustice to the use/mention distinction. This will help illustrate and precisify how the power to affect contextual parameters can have practical effects.
Let’s begin with a thought experiment.1
Professor Vine, a prominent philosopher, has sexually harassed many of his students and employees over his career. Several of his victims have recently publicly accused Vine of serious sexual misconduct. Various people now face choices about how to react. For example:
• students who might otherwise have taken Vine’s courses might worry whether this would put them at risk.
• colleagues who would otherwise have engaged with Vine as a member of the profession in good standing (e.g., by teaching his work, inviting him to conferences, suggesting that their students go work with him, etc.) might now worry that these actions would put people at risk and/or contribute to a culture that normalizes misconduct.
• administrators at Vine’s university may consider whether to reprimand him, suspend his courses, or terminate his employment.
For each of these choices, some of the central questions are epistemological: what should one believe about Vine? For each contemplated action, it is easy to imagine skeptical arguments against it: these are unproven accusations; one is innocent until proven guilty; since you don’t really know what happened, it would be unfair of you to __________. Such arguments will feel more compelling for more momentous fillings in of the blank, like terminating Vine’s employment, but one also hears protestations like this for much smaller contemplated actions, like declining to submit a paper to a conference in Vine’s honor.2
Like skeptical arguments generally, this kind of skeptical argument involves a clash of intuitions. It is intuitive—at least to many—that under circumstances like these, without special firsthand knowledge or the results of a careful objective public investigation, one can’t know what really happened. On the other hand, it is also very intuitive that quite often, we can know what happened simply on the basis of being told so—and claims to the contrary have unacceptably skeptical implications. It is also intuitive, [End Page 2] at least to me and to many, that in cases like Vine’s, the publicly available evidence is sufficient for knowledge of wrongdoing, justifying action.3
Here is a second thought experiment, illustrating an unrelated philosophical dispute.
Dora is a trans woman. She was designated as male at birth and raised as a boy as a young child, but later came to identify as a girl, then a woman. She has undergone gender-affirming hormone therapy. She has a penis. She has now lived as a woman for many years. Here are some questions about Dora:
• should Dora be welcome into professional women’s networks in her field?
• would a sexual relationship with Dora be consistent with one’s lesbian identity?
• should Dora’s physician rely primarily on the medical guidelines for men, or for women, in giving Dora medical advice?
At least some of these questions are obviously political. They also seem importantly connected to a metaphysical question about gender: is Dora a woman? As in the Professor Vine case, these are philosophical questions that are also deeply personal ones. One expects to find divided intuitions here too.
In one respect, such clashes of intuitions are typical of philosophical theorizing. One way this can play out is for the theorist to side with the stronger intuition—yes, q does seem false, but my intuition that p is true is just so much stronger, so I’ll accept the counterintuitive q. This, perhaps, is a way of describing G. E. Moore’s way with skepticism: I’m much more sure that I know many ordinary things than I am that I can only know things that are entailed by my sense data (1959, 216; 225–6). Or compare Talia Mae Bettcher’s remarks about accepting trans people’s gender self-identification as a methodological starting point (2007, 59).
There are other resources available for navigating conflicting intuitions, too. For one thing, we can consider likely sources of particular intuitions, in a way relevant to evaluating their trustworthiness. If we recognize an intuition as the sort that is likely to derive from particular known biases, for instance, that’s reason to mistrust it. This is the project of ‘explaining away intuitions.’ I have written in some detail about this kind of dynamic in past work (2009 (2013)4; I’ll touch on some of it again below.
A different kind of strategy for resolving apparently inconsistent patterns of intuitions finds ways to treat each intuition as true, while denying that [End Page 3] they are genuinely inconsistent. Contextualism is the most prominent strategy of this kind. Contextualists about knowledge ascription treat the language we use to discuss knowledge—e.g. the word ‘knows’—as context-sensitive, holding that this word can require the satisfaction of different epistemic standards, when used in different conversational contexts. Contextualists about gender language hold something similar of terms like ‘woman,’ saying that what it takes to count as a ‘woman’ depends in part on the conversational context in which the word is used.
Contextualism about ‘knows’ has it that when we say something like, ‘we can’t know whether we’re brains in vats,’ or ‘there’s no way to know whether Professor Vine groped that student,’ we’re speaking in high-standards conversational contexts where those sentences are true; when we say ‘we often know things by relying on others’ testimony,’ or ‘we’ve all known for months that Vine has a long history of sexual assault,’ we’re speaking in lower-standards contexts where those sentences are true. Since these are different standards, the contextualist says the apparent inconsistency of these competing assertions is illusory; neither intuition need be denied.5
Recall that the presentation of the case stipulates that the allegations are true; if not, they cannot be ‘known’ relative to any context. Moreover, I am focusing on cases where it is plausible that, at least by low standards, it is possible to know that they are true. (Many of Professor Vine’s real-world counterparts have this status.)6
Contextualism about ‘woman’ would suggest that in some contexts the extension of that word depends on, say, biological features like genital shape, and in others, it depends on self-identification. (There are many options besides these two; I focus on these for illustrative simplicity.) According to this view, whether the word ‘woman’ includes trans women who have penises in its extension will depend on the conversational context of the speaker. If one is speaking in a self-identification-emphasizing context, they will; if one is speaking in a genital-emphasizing context, they won’t.
I have defended contextualism about ‘knows’ in past work,7 and I continue to consider it an attractive and plausible view. While it is not my project here to convince anyone to become an epistemic contextualist, I do write in the spirit of attempting to further develop a plausible view. By contrast, I have no particular commitment with respect to the plausibility of contextualism about ‘woman.’8 I discuss it alongside ‘knows’ for two reasons. [End Page 4]
One is that one of my broader aims is to highlight the political significance of discourse about knowledge. The application of gender terms is more obviously political, so the significance of the possibility of contextual injustices about ‘woman’ provides particularly stark illustrations of this kind of injustice. And some of the parallels between gender debates and epistemic puzzles are striking; my hope is that this may help us better recognize the political nature of many disputes about knowledge.9
My second reason is that I think that some of the particular objections to contextualism about ‘woman’ in the literature have compelling answers that are of broader interest, and they are parallel to similar dialectics about ‘knows.’ So although I remain neutral on contextualism about ‘woman,’ I will argue below that certain objections to the view miss their mark.
In each case contextualism may seem to offer a way around intuitive conflicts. I will suggest that the idea that contextualism lets us avoid these disputes is only partly right. This is because the question of which intuitions are true is only one of many important questions we face. This point can be easier to recognize if one focuses especially on realistic cases with obvious practical import; as my central cases should make clear, the issues I am interested in have their home in our contingent, non-ideal social reality. Sometimes abstract and artificial cases deserve central roles in our philosophical theorizing, but sometimes, ‘messier,’ more realistic cases help us think more clearly.10
3. the triviality worry
An implication of contextualism is that you may speak truly, whether you ascribe a word (in my central cases, ‘knows’ or ‘woman’) or deny that it applies. Indeed, this is one of contextualism’s central selling points—it allows that both competing intuitions may be true. But one might worry about this implication. Does this mean there’s no substantive choice in the matter? Don’t we want to hold on to the idea that there might be something you should say? Call this the triviality worry.
In epistemology, the triviality worry would be that there is no substantive answer to the question of whether one should say that, e.g., ‘I know that Professor Vine has sexually harassed some of his students.’ Although it feels as if this is a substantive question, if one is a contextualist, one may think one will speak truly whether one says or denies that, so it doesn’t really matter which one picks. Given the strong apparent connections between questions about knowledge and the practical questions gestured at in the thought experiment, this might seem an implausible result.11 [End Page 5]
With regard to gender terms, the triviality worry is the worry that contextualism obscures, or even eliminates, what is manifestly an important moral and political question. While it allows trans-inclusive sentences (like ‘some women have penises’) to come out as true, because spoken in self-identification-emphasizing contexts, it also allows trans-exclusionary sentences (like ‘no one with a penis is a woman’) to be true, because spoken in contexts that essentialize genitals for gender terms.12 And this implication, while more trans-inclusive than an approach to ‘woman’ that characterizes the genital-emphasizing language as univocally correct, might still be thought to be insufficiently trans-inclusive, since it fails to condemn trans-exclusionary sentences as false.13
To feel the force of these worries, focus on conversational situations where there’s not already a determinate fact about which standards are in play. If one is in a conversation where the standards have already been set to make a given knowledge ascription false, for example, then one will probably speak falsely if one makes that ascription. But often, the relevant contextual parameters will not yet have been set; one can set lower epistemic standards by applying ‘knows,’ or higher standards by denying it. It is cases like these where the triviality worry is especially pressing. The same goes, mutatis mutandis, for gender language.
Note that the triviality worry isn’t that there’s no way to be wrong about the relevant ascriptions (even if the context is indeterminate in the relevant ways). For example, if, unbeknownst to you, you are a brain in a vat, then you don’t know that you have hands, even by conversationally accepted low standards; if Professor Vine actually is the innocent victim of an elaborate conspiracy, then you don’t know that he has sexually assaulted his students. (And if that’s really little Johnny, a cis boy, disguised in women’s clothing, ‘woman’ will not apply to him, regardless of the context.) But you are not a brain in a vat, and Professor Vine did harass those students. (And that’s not Johnny, it’s Dora, a trans woman.) If we assume these truths and speak in low-standards contexts, our knowledge ascriptions will be true; if we question them and speak in high-standards contexts, the corresponding knowledge ascriptions will be false. The worry is that the choice is arbitrary.
I think the triviality worry is deeply mistaken, and diagnosing its error can lead us to important insights about language and social power, as well as about knowledge and action. In particular, it will illustrate the significance of contextual injustice, which occurs when contextual parameters are fixed in unjust ways. First, I need to go into a bit more detail about conversational dynamics. [End Page 6]
4. single-scoreboard semantics
I said above that according to contextualism, an utterance like ‘we know that Professor Vine has sexually harassed some of his students’ may be true in some contexts (ones involving relatively lax epistemic standards), and false in others (ones involving relatively stringent epistemic standards). But one reason this doesn’t straightforwardly entail that a speaker will always speak truly, in affirming or denying that sentence, is that speakers typically lack the conversational authority to decide unilaterally what standards govern their sentences. Conversational contexts, and the standards they include, are socially shared things.
Keith DeRose’s explication of a “single-scoreboard” treatment of contextualism is helpful here. DeRose is engaging with what he takes to be a widespread misconception about contextualism, namely, that it implies that in a paradigmatic epistemological dispute—between a skeptic and a Moorean, say—both parties to the disagreement are speaking truly. This thought is presumably motivated by the idea that, since there is an available context that makes the skeptic’s claim true, and there is also an available context that makes the Moorean’s claim true, each participant will be speaking in a context suitable for the truth of their utterance.
I agree with DeRose that this is a mistake; conversations involving ‘knows’ are governed by a single-scoreboard semantics, according to which “the truth-conditional content of both our speakers’ uses of ‘know(s)’ is given by the score registered on this single scoreboard” (DeRose 2009, 137). A dispute like this will typically be in a single conversation involving multiple people. So there typically won’t be multiple contextual parameters at play. There is the epistemic standard operative in the shared conversation, and it isn’t one that will make both utterances true. (DeRose focuses on ‘knows,’ but intends the framework to apply to context-sensitive language more generally.)
DeRose doesn’t quite put things in this way. Instead of talking about the conversational standards directly, he theorizes in terms of what he calls the “personally indicated content” of an individual’s use of ‘knows.’ He characterizes this latter in terms of “the content that speaker’s conversational maneuvers have a tendency to put in place for that term” (2009, 133). When the “personally indicated” standards of each conversational participant agree that a given knowledge ascription is true, it is true in the shared context; when they agree that it is false, it is false in the shared context. (When the personally indicated contents disagree, DeRose characterizes the conversation as a defective one, in which the knowledge ascription is neither true nor false.) [End Page 7]
I agree with DeRose here more than I disagree, but I do wish to highlight two respects in which my treatment is different from his. First, I do not wish to articulate the conversational dynamics of situations like these in terms of individualistic “personally indicated” standards. Language is a thoroughly social phenomenon, and I am not sure it makes sense to discuss an individual’s standards in a way independently of the conversation in which they are speaking. So while I share DeRose’s ambition to understand how individuals’ assertions affect the collective contextual standards, I do not do so in terms of individual standards.
Second, more importantly, DeRose assumes a kind of even linguistic playing field. The key questions for DeRose are ones about whether there is unanimity among the speakers’ personally indicated standards; the framework is indifferent to social asymmetries: by definition, everyone has an equal say in whether there is unanimity on a given question. But conversations are not always, or even typically, between individuals with symmetrical social positions; conversations often have participants in more dominant and more submissive roles. This may be due to very contingent features of the conversation—perhaps its primary role is for one person to inform another about an area in which they are much more familiar, for example. Or it may be part of a more systematic social domination, as in the case of members of one social caste systematically expecting, and receiving, deference from members of another. In such cases, the more dominant speaker has more influence in setting the shared conversational context; the assertions of someone in authority come with more pressure to accommodate, than do those of others in a conversation. Because DeRose’s approach looks only at whether all participants’ standards agree about a given knowledge ascription, it cannot reflect these kinds of authoritative asymmetries. I shall say more about these social dynamics in §6 below.
Let me return now to the point of the previous section, to place my discussion of DeRose in context. In §3 I introduced the triviality worry (which I reject): since, given contextualism, one will speak truly whichever way one comes down on the question of whether one knows that Vine has harassed some of his students, contextualism implies that there’s really nothing to it—you can answer either way, and no one can complain that you’re speaking falsely.
Given a single-scoreboard semantics, the key premise of this worry isn’t quite right. Just because you personally say “many people know that Vine has harassed students” doesn’t mean that your conversation is in a context where that sentence is true. Maybe you’re speaking to people [End Page 8] who are more skeptical, with the effect that your shared conversation is one that makes your sentence false or truth-valueless. There is still going to be a difference between getting things right and getting things wrong; you’re not just going to end up speaking truly, no matter what you say. This complicates the triviality worry.
It complicates it, but it does not obviate it. The individualistic view and the single-scoreboard view give different verdicts only when there is no unanimity among conversational participants. But just as there can be substantive questions about what standard an individual should use, so too can there be substantive questions about what standards a group of conversational participants should use. Given the kind of normativity the triviality worry is in tension with, someone’s adopting a standard doesn’t make it the right one; but for exactly the same reason, a pair of people, or a larger group, all adopting a given standard within a conversation doesn’t necessarily make it the right standard either.
5. appropriate contextual parameters
Here is a banal example. Consider the question, ‘are there any doctors here?,’ asked by a flight attendant, when a passenger needs emergency medical attention. Focus on the context-sensitivity of the indexical ‘here.’ It is doubly context-sensitive: its “pure” component is approximately just a matter of the location of the speaker, but there is also contextual variability with respect to how large a region of space is covered. It might refer to a specific row on the airplane (‘the emergency exits are here’), or it might refer to the geographic region being flown over (‘there are beautiful mountains here’). In the case of the flight attendant’s question about a doctor, we’d expect ‘here’ to refer to the plane. A conversational context that restricted the relevant location much more narrowly—to one seat, say—would be disadvantageous. Asking ‘is there a doctor here?,’ meaning here in this particular seat, is a terrible response to a medical emergency of the kind we’re contemplating.14
Something very similar goes for knowledge ascriptions, in a contextualist framework. Maybe skeptics speak truly. It doesn’t follow, from the fact that their utterances are true, that they are proceeding appropriately. Refusing to accept what is manifest to one’s senses is a lousy way to get around the world. The radical skeptic invites us to make use of inappropriately high standards.
Evelyn Brister has emphasized the connection between skepticism, contextualism, and oppression. She puts it as follows: [End Page 9]
Recognizing the power dynamics of skeptical arguments allows us to examine the interests that philosophical skepticism serves. . . . [S]kepticism distrusts all knowledge claims and avoids falsehood at the cost of dismissing possible truths: this represents a deeply conservative attitude toward the production of knowledge. For this reason, skepticism poses a special danger to novel claims that may already face a higher burden of proof and where skeptical objections may then prove.(2009, 682)15
Since there are normative questions about what standards to employ, the contextualist “dissolution” of the skeptical paradox is only partial. It tells us that each party—the skeptic and the person affirming ordinary knowledge—may be speaking truly. But that doesn’t answer every question that needs answering. In particular, it doesn’t tell us which way we ought to speak. This is part of why the bare claim of contextualism can often feel dissatisfying to those engaged in traditional epistemic questions.16
This kind of normative question may also underwrite the persistent sense that there is a genuine disagreement between skeptics and non-skeptics, even though the contextualist says that each may be speaking truly. This has historically been one of the central motivations against contextualism about ‘knows.’17 But not all disagreements are disagreements about the truth of a given proposition; some are practical—they are disagreements about what to do.18 In this case, they may be disagreements about what epistemic standards to employ.19
This point is perhaps more obvious in the case of ‘woman,’ where the political import is more conspicuous. Holding that people’s gender should be classified on the basis of their self-identification, rather than their genital shape, is part of what it is to be trans-inclusive. Those who oppose classifying trans women as ‘women’ do not merely happen to be speaking in contexts that emphasize other criteria; they think those criteria are the important ones for the purpose of discourse about ‘women.’ Just how these arguments will play out depends on the role of this kind of gendered language—a question well beyond my scope in this paper. (I will discuss the parallel question about knowledge in §7 below.)
Assume that the trans-inclusive stance is correct in a given situation; consider again Dora, a trans woman, who falls under the extension of ‘woman’ in the context most appropriate to the situation. She ought to be categorized along with cis women as a ‘woman’ for the purpose at hand. (Perhaps the best policy is one in which she is welcome to use the women’s locker room, for instance.) In such a situation, the contextualist approach does imply that there is a possible (trans-exclusionary) context in which [End Page 10] the sentence ‘Dora is not a woman’ is true. But this does not imply, as Saul suggests it does, that it is just as acceptable as a trans-inclusionary utterance would have been.20 For one thing, the trans-inclusionary context is (I have stipulated) the better one.21
But the utterance isn’t merely better for taking place in a better context— it serves to create and maintain that context. Contexts don’t just appear ex nihilo or change randomly; the context in a conversation is the one that the accepted utterances require for their truth.22 So an utterance of ‘Dora is not a woman’ actively serves to create an inappropriately trans-exclusionary context. It does so even if it is true. Indeed, it especially does so when it is true—it is true when the relevant trans-exclusionary commitments are taken up in the conversation, leading to discriminatory harms toward Dora and other trans women. So, contra the triviality worry, I do not agree that the contextualist approach to ‘women’ implies that trans-exclusionary contexts are equally legitimate. It only implies that they are possible. There are many more, and more important, questions to answer than questions about which sentences are true; there are deep questions about which kinds of contexts are appropriate. Given a single-scoreboard semantics, disagreements within a single context between those engaging in trans-inclusive language and those engaging in trans-exclusive language will literally be disputes about what kind of context to be speaking in. These are deep practical and moral questions, which may well have an objectively correct answer.23 Like Saul and Bettcher, I consider trans-inclusivity to be a legitimate desideratum in semantic theorizing, and I agree that we need grounds on which to side with the trans-inclusive language and against trans-exclusive language. But I reject the idea that any of this is a strong objection to the contextualist approach to ‘woman.’ I can, and do, agree with Saul that contextualists need a way to treat the trans-exclusionary language as unacceptable, without going on to suppose, as Saul does, that the way to do this must be to say that it is false. One of the central ideas of this paper is that, when it comes to context-sensitive language, the truth of an utterance does not exhaust its normative profile.
6. knowledge, contextual parameters, and social power
Considering broader disagreements of this kind rebuts a popular argument against contextualism.24 But it also illustrates interesting interactions between the epistemic, the semantic, and the political.25 Our choice of epistemic standards has a lot to do with which sources we treat as [End Page 11] authoritative, which in turn will inform and influence our decisions. To see this, let’s return to Professor Vine.
In the Professor Vine case, there are possible contexts relative to which the sentence ‘we know that Vine has sexually assaulted some of his students’ expresses a truth, and there are possible contexts relative to which it doesn’t. Assume that your conversation is so far indeterminate as to which epistemic standards are in play, such that you could affirm or deny that sentence, and you could thereby express a truth either way. It does not follow that each utterance would be equally appropriate; as in the airplane case, there are normative constraints on conversational standards. Just as one should not use ‘here’ in a way that is specific to a particular seat when looking for a doctor, there may well be normative constraints about what kinds of standards ought to be operative in responding to sexual assault allegations.
Adjudicating such questions cannot happen in an ideological vacuum. The considerations that bear on what kinds of standards to use are wrapped deeply in socially and politically controversial stances. (It does not follow that there are not objectively preferable standards, since we shouldn’t assume that there aren’t objectively preferable politically controversial stances.) There is a possible conversational practice of raising epistemic standards whenever considering sexual assault allegations; this would amount to a kind of systematic mistrust of such allegations. (Unfortunately, this practice seems actual in many communities.) Phrases like ‘innocent until proven guilty’ and labels like a ‘he said–she said situation’ tend to raise epistemic standards. In criminal contexts, this is appropriate; we have high bars for knowledge in these situations for good reasons. But using language of this sort in less formal settings where a less momentous sanction is being contemplated is much less obviously appropriate.26 It will, for example, undermine particular contemplated actions like declining to invite Vine to conferences. It also makes the possibility that the reports are false more salient, which in turn impugns the reporters.
One of the ways we influence standards is by making utterances that require particular contextual parameters to be in place for them to be true. If I say, ‘we know that Professor Vine sexually harassed several of his students,’ I am expressing a particular low-standards knowledge proposition, but I am also making a conversational bid for a low-enough conversational standard such that this sentence can come out true. Mary Kate McGowan, following Lewis (1979), describes moves of this latter sort as “conversational exercitives”—speech acts that alter what is [End Page 12] conversationally permissible, without expressing anything metalinguistic about the conversation (2004, 93–94). If I make the utterance above, and it is accepted, then my conversation may become one in which we can take the complainants’ testimony for granted.
One’s ability to affect conversational standards is a kind of social power; like social power generally, it can be wielded for evil or for good. It is also, like other kinds of social power, distributed unevenly in society. Socially powerful individuals, and those serving their interests, will have an easier time setting conversational parameters than more marginalized people will. In the present case, this can be manifest in, for example, the intuitive perception of the situation as one in which the most salient worries concern being unfair to Vine—one is tempted to suppose that, since the stakes are high (It is very threatening to be accused of sexual assault!) we should have a high bar for acceptance. The salience of the potential effects of one’s epistemic judgments on Vine, as opposed to the potential effects on the students who are not being believed, or on other potential victims, is a part of a widespread patriarchal structure—what Kate Manne has called “himpathy” (2018, 196–201). It is also both a product of, and a contribution to, rape culture. ‘Rape culture’ is the label for the features of society that make it conducive to sexual assault. A systematic tendency to disbelieve assault allegations—or even to suspend judgment about them—certainly qualifies.
This kind of power differential affects the ability to set contextual parameters in multiple ways. For one thing, participants who might wish to advocate for a counternormative epistemic standard—those in a rape culture wishing to affirm testimonial knowledge in cases like Vine’s, for example—may, in recognition of the challenges to uptake they are likely to face, decline to even offer the kind of assertion that could shift the standard in that way. This would be a particular form of what Kristie Dotson (2011) has called testimonial smothering. And if they do choose to make such a conversational attempt, they are likelier than their opponents to face resistance to uptake. There is a rhetorical advantage to those who contribute to rape culture by raising epistemic standards to exclude ‘knowledge’ of sexual misconduct. There is intuitive force behind the following argument: you can’t say you know that these allegations are true unless they’ve been proven beyond a reasonable doubt! If such a statement is accepted—and in a rape culture it often will be—it will create a context in which it is literally true. Anti-rape activists bidding for lower standards that would—that should—have made their opposing claims true will be thwarted. [End Page 13]
7. skepticism and conservativism
In addition to the particular social forces that work against victims of sexual assault, there are also more general epistemic tendencies that pull in the same direction. I have in mind the general emphasis on negative epistemic norms, as opposed to positive ones. We tend to emphasize the epistemic error of mistaken or unjustified beliefs far more than that of unwarranted agnosticism. This skepticism-emphasis is baked deep. Descartes’s Meditations begins with the meditator’s observation that some of his past beliefs have been mistaken, which motivates him to reject all beliefs until they are certified as in good standing. The converse worry, that one may mistakenly suspend judgment on questions for which there is sufficient evidence, does not even register.
This kind of emphasis on the skeptical also seems to be present in discussions of contextualism. Many contextualists have observed that it seems easier to raise standards than to lower them. On David Lewis’s version of contextualism, for example, merely thinking about a skeptical possibility brings it into epistemological relevance, where it will remain until it is forgotten about some time later (1996, 559–60). I am not convinced that this represents a linguistic rule so much as a contingent fact about psychology and human interaction. Maybe many humans have a tendency to be more deferential to standards-raisers than lowerers, but if this tendency were resisted, the asymmetry would disappear. (I also suspect the deference to the skeptical may be more prevalent among philosophers than among speakers at large.) This would seem to be an implication of standard approaches (including Lewis’s!) to conversational accommodation: one can lower the standards by making a knowledge ascription that requires that some previously considered possibilities be ignored; if it is accepted unchallenged, one has lowered the standards.27 Still, Lewis seems right that there seems to be at least some tendency to be deferential to higher standards.
Given the connections between epistemology and action, the bias towards the skeptical amounts to a bias towards the conservative.28 How to make the connections between epistemology and action precise is controversial, but the idea that there is such a connection is not. What you do depends in important ways on your epistemic situation. The idea that you have insufficient epistemic access to a given proposition is closely tied with the idea that you should not rely on it in your decision making. One natural precisification is a knowledge norm for practical reasoning.29 For concreteness, I will assume such a connection here. I think most or [End Page 14] all of the foregoing points will apply, whatever one’s preferred way of explaining the role of epistemology in appropriate decision making, but I won’t try to make that case here.
On my approach, if you know something, you can count on it. Or to put it another way, if you can’t count on it, you don’t really know it. This is why skepticism is so easy to use as pressure against action. You have not seen the product of an independent investigation into Vine’s conduct, so you can’t know what happened, so it would be reckless to act as if Vine has a history of harassing his students. Since you have to know something to act on it, skeptical arguments can easily be used to undermine action.
The intuitive attraction of the negative bias in epistemology, then, is connected to a kind of status quo bias—the idea that things are basically okay as they are, and what it means to be careful is to minimize the risk of messing things up. On this kind of approach, it is acting that is dangerous; the potential risk of inaction—the harm not avoided—is not modelled. I agree with Hundleby (2016) that the status quo bias is a fallacy—it is helpfully theorized as such alongside more canonical rational errors like affirming the consequent, base rate fallacies, or asymmetric loss aversion.30
Moreover, it is a bias that impacts different people differentially. In particular, a bias in favor of the status quo will disadvantage those who are less well-off under the status quo. Given the connection between skepticism and the status quo, a bias towards the skeptical will impede efforts at reform. Sextus Empiricus described the Pryrrhonian skeptics as “living in accordance with the normal rules of life,” including “the tradition of laws and customs” (1933 I. xi. ¶23). Descartes similarly embraced his local norms as part of the provisional moral code he would follow while engaging in his method of doubt.31 David Hume famously suggested that, since we can never use reason to justify our fundamental epistemic assumptions, we must simply resort to habit and form beliefs in the way that feels most natural to us (1739/2000, 144, §1.4.2, ¶57).32 None of this is any way to motivate reform.
8. contextual injustice
A central upshot of what I have said is that contextualism points to an important kind of social power that is easily overlooked. Indeed, because contextualism itself is a controversial thesis, not universally recognized, the social power it implies is one that will typically operate tacitly. The unreflective wielding of power is particularly likely to be used in ways that further oppressive norms. For example, if like most people, my intuitions [End Page 15] are influenced by an unreflective tendency to place more trust in socially powerful people than in marginalized ones, I might tacitly invoke less skeptical standards in engaging with the former’s testimony than the latter’s. And I might do so without any particular oppressive intent. It will feel to me like objectivity.33
I would like to suggest that this is one natural way of implementing some of Miranda Fricker’s ideas about epistemic injustice. As I’ll explain below, contextual injustices with respect to knowledge ascriptions amount to a more specific category within Fricker’s framework of testimonial injustice, but the concept also has broader applicability outside of epistemology.
Fricker’s Epistemic Injustice articulates and explores two distinctive kinds of “wrong done to someone specifically in their capacity as a knower” (2007, 1). The first is testimonial injustice. Testimonial injustices are cases of credibility deficits due to an identity prejudice. A credibility deficit is defined as a situation in which a speaker is “receiving less credibility than she otherwise would have,” if not for the operation of a prejudice (2007, 17).
One way one’s identity prejudice might result in credibility deficits would be that one gives a lower credence to the content of the speaker’s testimony than one should. (I think this is the way most people tend to think about Fricker’s cases.) But one could also think about credibility assignments in terms of whether one treats testimony as knowledge-conferring; unfairly raising the standards for ‘knows’ could constitute an unjust credibility deficit. If it met Fricker’s other criteria, it would also constitute a testimonial injustice. For example, if my prejudice against women leads me to raise the standards for ‘knows’ when it comes to women’s testimony, I unfairly allow men to settle what I count as “knowledge” to a greater degree than I do women. My sentences denying women’s “knowledge” may be true, but they constitute testimonial injustice nonetheless.34
A very similar injustice occurs when prejudicial standard-raising attaches to particular kinds of contents, as opposed to particular kinds of identities. For example, one might employ unusually high standards when assessing testimony describing sexual misconduct (regardless of the identity categories of the testifier). This would obviously further rape culture, and it should arguably count as a testimonial injustice, even if it doesn’t depend on a prejudice against a certain kind of speaker.35
The unjust raising of epistemic standards, serving to diminish a speaker’s testimony by denying “knowledge” status in a conversation, is a kind of testimonial injustice.36 But it is also an example of a broader phenomenon [End Page 16] of independent interest. More generally, a subject suffers a contextual injustice when speakers unfairly employ contextual parameters that are disadvantageous to the subject. The assertions characteristic of contextual injustice will typically express truths—this is why they can be so insidious. One might think of contextual injustice as a kind of metasemantic reimagining of Sally Haslanger’s observation that an inappropriate ideology tends to “make itself true” (2017a, 15). I suffer a contextual injustice if you unjustly employ high epistemic standards that result in my say-so being treated as inconclusive; you say that one can’t “know” on the basis of my testimony, thus wronging me, even while creating a conversational context where what you say is accurate.37
Contextual injustices regarding gender language are also easy to understand, given the contextualist approach to gender terms I have been discussing. Supposing again, as we did in §5, that Dora deserves to be treated as a woman for the purpose at hand, creating a trans-exclusionary conversational context does her an injustice.
Andrew Peet (2017) has recently introduced a category of “interpretive injustice,” which is similar in some interesting ways to contextual injustice, but it is importantly different in others. Interpretive injustices are cases in which one misinterprets the intended communicative content of a speaker’s testimony, due to a prejudice. For example, in one of the cases Peet discusses, female entrepreneurs who included, as part of an investment pitch, sentences like ‘we aim to achieve x by 2018,’ were interpreted by would-be investors as making a rather timid claim, describing only their goals which might be somewhat vaguely held. By contrast, when male entrepreneurs produced the same sentences, they were interpreted as something more like a firm commitment to achieving x by 2018 (2017, 3431).38
Like contextual injustices, interpretive injustices exploit the gap between the words a speaker produces and the contents speakers attach to them. They are both ways in which power and language can interact to oppressive ends. But they are also importantly different categories. For one thing, an interpretive injustice is always committed against a given speaker—it is an unjust interpretation of what they said. By contrast, a contextual injustice might be an injustice that serves to harm someone uninvolved in the conversation.39 In the sexual harassment case described in the introduction, for example, speakers may commit contextual injustices collectively against Professor Vine’s victims, without committing any kind of contextual injustice against one another. (If the speakers are professors, [End Page 17] the inappropriate standards they use may actually further advantage and privilege one another; it is students who might complain about sexual harassment who are unfairly disadvantaged.)
Note also that interpretive injustices by definition are incorrect ideas about a speaker’s intended communication, so they constitute false beliefs. By contrast, contextual injustices serve to make inappropriate sentences literally true. Thus, an interpretive injustice can, in theory, be avoided by scrupulous attention to truth. (I do not mean to suggest this is likely to be easy, or even practically possible—only that it belongs to a traditional domain of accuracy-emphasizing philosophy.) One of the reasons contextual injustices can be so insidious is that their harmful sentences are true. Because of this, one cannot overcome them by showing that they are mistaken.
I have focused so far on contextual injustices about ‘knows’ and ‘woman,’ but contextual injustice is a much broader phenomenon. Here is a simple example. Gradable adjectives in general are context-sensitive. Gradable adjectives express properties that come in degrees, where the conversational context sets the threshold for the term to apply. For example, ‘tall’ is a gradable adjective; some people are taller than others, and just how tall one must be to count as ‘tall’ in a particular conversation depends on the context.
Consider a gradable adjective that also attaches to a prejudicial stereotype. For example, consider the stereotype that girls are not good at math. ‘Good at math’ is a context-sensitive predicate; one will use different standards for this predicate depending on one’s conversational interests. I myself would ace a junior high algebra exam, but I would quickly confuse myself if I tried to work through some differential equations. I count as ‘good at math’ in some contexts (my teenage niece would do fine in coming to me for advice on her math homework), but not in others (my mathematician colleagues should not seek my help grading their students’ homework). Such is the flexibility of the phrase ‘good at math,’ and that flexibility is there for good reasons.
Still, it is prone to misuse; one can generate contextual injustices with it. For example, suppose one set out to investigate the stereotype that girls tend not to be good at math. One could, if allowed a relatively free intuitive categorization of children into the ‘good at math’ and ‘not good at math’ categories, shift the standards in a way disadvantageous to girls. A relatively crude way one could do this would be to simply employ higher standards for girls than for boys. Suppose that when assessing a boy, one [End Page 18] counts them as ‘good’ if and only if they score above the 50th percentile for junior-high students, but when assessing a girl, one counts them as ‘good’ only if they score above the 90th percentile. This is an unjust categorization procedure, and one likely to entrench harmful stereotypes. But the problem is not that the adjudicator is speaking falsely in any given judgment about whether a student is ‘good at math.’ Like contextual injustices generally, this procedure is illegitimate and harmful, even though it involves saying true things with context-sensitive language. Indeed, part of the reason it can be so insidious and harmful is precisely that its inappropriate verdicts are literally true.
This is a stark example, and an admittedly artificial one, but I hope it illustrates the space for contextual injustices to work outside epistemology.40 The cases of ‘knows’ and ‘woman’ are only two examples of the way contextualism can help make sense of linguistic disagreements that have political import. In the final section, I return to ‘woman,’ to illustrate how ideas about contextual injustice interact with the distinction between use and mention.
9. use and mention
In addition to a version of the triviality worry, Talia Mae Bettcher presents another worry about the kind of contextualism about ‘woman’ considered in this paper. Someone’s status as a woman, Bettcher says, should not be a political matter; it should be settled by metaphysics, not by political considerations (2013, 239).41
This is directly analogous to the worry for epistemic contextualists that conversational maneuvers having nothing to do with epistemic factors should not establish or diminish a subject’s knowledge. To at least some degree, the answer is the same: if we are careful with the use–mention distinction, the apparent dependence of knowledge, or womanhood, on linguistic matters will disappear.
The use–mention distinction is a distinction between different things we do with words. We use words to express ideas, which often have nothing in particular to do with language. When we mention words, we are taking those words as subject matters in their own right. (A standard convention, employed in this paper, is to designate the mention, as opposed to the use, of a word by enclosing it in quotation marks.) The word ‘woman’ contains two vowels. But a woman would not typically contain any vowels; women are made of flesh and blood, not letters. Contextualism about a given word is a linguistic claim about that word; a canonical statement of [End Page 19] contextualism about ‘knows’ or ‘woman’ will mention that word; it will not use it. According to the kind of contextualism about ‘woman’ under consideration, the word ‘woman’ is sometimes used to pick out adults who self-identify as female and sometimes to pick out adults who have genitals of a certain shape. This is very different from saying that sometimes self-identification determines whether one is a woman, and other times one’s genitals do. (The former statement mentions ‘woman’; the latter uses it.)
Similarly, according to epistemic contextualism, satisfying the word ‘knows’ sometimes requires meeting a rather strict epistemic condition, and sometimes it requires meeting only a more modest one. For example, in skeptical conversational contexts, ‘knows’ may express quite a strong relation, such that almost all ascriptions of ‘knowledge’ are false, even though in non-skeptical contexts, many sentences containing ‘knows’ are true. But this is not at all to say that whether one knows depends on the conversational context or that talking about skeptical scenarios and raising the standards has the effect of diminishing one’s knowledge. In the same way, contextualism does not imply that whether Jane knows that Vine harasses his students depends on how skeptical my conversational context is. I am in fact writing in a relatively low-standards context, and Jane does know that Vine harasses his students; my switching to a more skeptical context would have no effect whatsoever on what Jane knows.42
Thus, read literally, Bettcher’s description of contextualism about ‘woman’ as a view according to which trans women’s “status as women is . . . decided by the political” is incorrect. Bettcher is writing in a conversational context that treats trans women as “women”; given contextualism, in the context of her paper, someone will count as a “woman” just in case (to an approximation) they are an adult who self-identifies as a woman. And whether any given individual is an adult who self-identifies as a woman is not up for political resolution, in the kinds of debates Bettcher is imagining.
We can recover a political dependency in the neighborhood of the one Bettcher describes, if we are careful about use and mention. We can not say that whether Dora is a woman depends on whose gender standards prevail in a given conversation, but we can say this: whether the asserted sentence ‘Dora is a woman’ is true depends on whose gender standards prevail in the conversation in which it is uttered. It is true that whose standards will prevail is in significant part a political question. Is this a problematic or implausible result? It does not say that anyone’s status as a woman depends on whose standards prevail in a conversation, but it [End Page 20] does say that which sentences are true so depends. But this must be true; it is a consequence of the fact that all language is socially constructed. It manifestly is possible to use language in trans-exclusionary ways; doing so is not a mere linguistic error.
If what I have been suggesting is right, this is exactly why it can be so harmful.43
Jonathan Jenkins Ichikawa, PhD, is an Associate Professor of Philosophy at the University of British Columbia in Vancouver. He works primarily on topics in epistemology and issues connected to rape culture and sexual violence. He is the author of Contextualising Knowledge (Oxford University Press, 2017). He is currently working on a research project on Rape Culture and Epistemology, funded by a Canadian SSHRC Insight Grant.
2. This is a literal example I observed myself. One philosopher suggested, about an upcoming conference in honor of a real-world “Professor Vine,” that this would be a good conference to decline to submit a paper to; I watched another philosopher censure that stance on grounds of fairness and presumption of innocence.
3. I characterize the conflict in terms of a clash of intuitions to emphasize its continuity with more traditional philosophical puzzles; I do not assume that the “intuitions” in question are held pre-theoretically. They may even be the result of theoretical arguments. (I think the same of more paradigmatic intuitions about thought experiments, but this is controversial. See Ichikawa and Jarvis (2009).)
6. I do not say—as a referee once misinterpreted me as saying—that this is true of all examples of sexual harassment allegations; I’m merely focusing on examples that are like this.
8. Jennifer Saul (2012) discusses this kind of contextualist view, but she does not endorse it. She describes it as one with which she previously had sympathy, but she raises what she takes to be serious concerns about it. Many of these concerns are closely related to those I’ll go on to discuss below.
9. Thanks to Hannah Bondurant for first bringing this parallel to my attention.
11. The triviality worry is related, but distinct, from the worry that contextualists inappropriately shift the focus from epistemic questions to linguistic ones. This latter worry has been widely discussed; Sosa (2000) and Kornblith (2000) give its canonical presentations. Montgomery (2017) gives a response.
12. As Saul points, out, this isn’t the same as saying that trans-exclusionary denials of trans women’s self-ascription of womanhood can be true. On this view, there is no content that someone in a trans-inclusionary context affirms with the sentence ‘I am a woman,’ that is also denied by the trans-exclusionary language. See Saul (2012, 209).
I think it may be right to say that there is something trivializing about the way that the contextualist grants the truth of trans women’s claims. The reason the trans woman’s claims are true, on the contextualist view, is simply that there is a huge range of acceptable ways to use the term “woman” and the trans woman’s way of using “woman” isn’t ruled out. The trans woman’s use of “woman” is perfectly acceptable—just as acceptable as her opponent’s. In effect, I am saying, “Yes, your claim to be a woman is true—because ‘woman’ can mean so many things.” And this, I can’t help but think, would be deeply unsatisfying to the trans woman who wants to be recognized as a woman simply because she is a woman rather than because “woman” is such a flexible term. What the trans woman needs to do justice to her claim is surely not just the acknowledgement that her claim is true but also the acknowledgement that her opponent’s claim is false. And the contextualist view does not offer that.
See also the similar worries in Bettcher (2013, 239). Bettcher worries that given contextualism about ‘woman,’ “in questions concerning whether a trans woman is a woman, there does not seem to be room for metaphysical disagreement,” because the question “comes down to which standards are applicable in a given context.”
14. Some contextualists hold the metasemantic thesis that the appropriate contextual parameters are thereby the parameters actually employed. See e.g. the discussion of ‘domain goals’ and metasemantics in Dobler (2019). Lewis (2019) argues on similar grounds that intentions play a limited role in setting contextual parameters, as compared to objective features of the speakers’ situation. I disagree; the possibility of using inappropriate parameters is part of the explanation for why they are harmful. Thanks to Karen Lewis for clarifying discussion here.
19. Cf. Esa Díaz-León’s suggestion that “Moore and the skeptic (might) express compatible propositions at the semantic level, but they still express a genuine disagreement . . . at the pragmatic level” (2017, 75). She describes this as a disagreement about which kinds of standards one should use. Díaz-León is building on the ‘metalinguistic negotiation’ framework of Plunkett and Sundell (2013).
21. A contextualist approach of this sort might allow that in certain kinds of contexts—some medical ones for instance—different, trans-exclusionary standards might be appropriate. Contextualism is also consistent with the denial of this idea.
22. See again Lewis (1979), cited in the parallel discussion in §7. Compare also Lynne Tirrell’s remark that “[o]ur speech acts also undertake a meta-level expressive commitment about the very saying of what is said. Expressive commitments are commitments to the viability and value of particular ways of talking, modes of discourse” (2017, 144).
23. What I say here has many points in common with Esa Díaz-León’s (2016) discussion of contextualism about ‘woman.’ We are both motivated by the worries Saul and Bettcher raise for the contextualist approach. Like me, Díaz-León emphasizes the role of questions about which standards are appropriate to a given conversational context. As she puts it, which standards are “relevant” in a context isn’t merely a matter of the speakers’ ideas, but they can depend on broader social and political factors. But unlike me, Díaz-León builds the idea that the appropriate standards are operative into the semantics; in other words, she holds roughly that an utterance of the form ‘S is a woman’ is true if and only if S satisfies the standard for women that should be operative, given S’s situation. So in the case above, the trans-exclusionary speakers would be speaking falsely (2016, 249). Although Díaz-León describes her view as a defense of contextualism, her version of the view is not really a contextualist one as I’d describe it, as she holds that the relevant standards are determined entirely by the subject’s circumstances, not the speaker’s. See her n9. So her view is really more similar to what epistemologists call “subject-sensitive invariantism” or “interest-relative invariantism.” See Weatherson (2017).
25. Compare the discussion of ‘serious verbal disputes’ in Jenkins (2019). Jenkins engages with questions about divergent uses of the word ‘love’; her approach has some significant parallels with my own treatment of ‘knows.’ In particular, she says, as I do, that divergent uses sometimes correspond to semantic differences among the speakers’ intended senses, but emphasizes, as I also do, that it would be a mistake to trivialize the dispute; disputes about whether people count as in ‘love,’ like those about whether someone counts as ‘knowing,’ can be politically important. Unlike me, however, Jenkins does not explain the semantic variation in terms of context-sensitivity, instead positing different individually indicated concepts that are expressed by their use of the key word. Consequently, Jenkins does not emphasize, as I do, normative constraints on appropriate contextual parameters, instead focusing on norms about which concepts are the appropriate ones. This latter question, she treats with reference to the greater accuracy of some concepts over others. (This notion is spelled out more fully in Jenkins (2008).) My treatment does not depend on the idea that a concept (as opposed to a proposition) can be more or less accurate.
26. Many epistemologists have posited a connection between stakes and standards for satisfying knowledge ascriptions; see e.g. Stanley (2005), Fantl and McGrath (2009), or Weatherson (2017). See also the parallel literature on inductive risk, e.g. Douglas (2000) and Biddle and Kukla (2017).
27. Lewis (1979). McGowan (2004, 107–9) observes that this kind of dynamic represents one way that language can be used for harm. I am emphasizing the flip side of McGowan’s observation: it can also be used to improve the practical situation by lowering inappropriately high standards.
31. “I formed a provisory code of morals. . . . The first was to obey the laws and customs of my country, adhering firmly to the faith in which, by the grace of God, I had been educated from my childhood, and regulating my conduct in every other matter according to the most moderate opinions, and the farthest removed from extremes, which should happen to be adopted in practice with general consent of the most judicious of those among whom I might be living. For, as I had from that time begun to hold my own opinions for nought because I wished to subject them all to examination, I was convinced that I could not do better than follow in the meantime the opinions of the most judicious; and although there are some perhaps among the Persians and Chinese as judicious as among ourselves, expediency seemed to dictate that I should regulate my practice comfortably to the opinions of those with whom I should have to live. . . .” Descartes (1637/1912 Part III, p. 19)
34. Not all testimonial injustices are contextual injustices. For instance, one might give someone a credibility deficit even if they were not authoritative enough to confer knowledge (even by low standards).
36. At least it will often be one. In a more recent discussion, Fricker suggests that it should be partly definitive of epistemic injustice that it occur nondeliberately, due to the tacit operation of a prejudice Fricker (2017, 54). As I pointed out at the top of this section, this will typically be the case for the kinds of contextual injustices discussed here. But it is also possible that one might raise the standards knowingly and deliberately in order to prevent the acceptance of testimony as “knowledge”; this would not count as an epistemic injustice, under Fricker’s (2017) definition.
37. Contextual injustices may be examples of a broader phenomenon about meaning fixation generally; even when it comes to context-insensitive discourse, there can be substantive questions about what given words ought to mean, which may be negotiated at a broader metasemantic level, raising similar normative questions. Thanks to Derek Ball for helpful discussion here.
39. This is also a point of contrast with Fricker’s testimonial injustice, which is always committed against a speaker in the conversation. Thanks to Cam Gilbert here.
40. There could also be contexts that are unjust for reasons separate from making certain sentences true. Davies (2014), for example, identifies a class of cases where inappropriate contextual parameters unjustly limit which things can be said, given conversational pragmatics. These might be rather different kinds of contextual injustices.
41. It is by no means uncontroversial that womanhood is not ultimately a political matter; social constructionists about gender might well think it is. See e.g. Haslanger (2012), Dembroff (forthcoming), or Barnes (2019). Since I do not propose getting into the metaphysics of gender here, I assume for the sake of engagement with Bettcher that the idea that womanhood depends on politics is a consequence to be avoided.
42. This is widely recognized in the literature, but some philosophers are confused and/or write sloppily about it. Cf. e.g. David Lewis’s (1996) discussion of knowledge being “elusive.” See DeRose (2009, 212–15), Ichikawa (2017a, 1–3), or Ichikawa (2017b, 3–4) for further discussion.
43. This paper developed out of an idea first introduced in §6 of Crewe and Ichikawa (forthcoming). I presented a version of it at a 2018 UNC Philosophy Department colloquium, a 2018 UBC epistemology working group meeting, a Society for Women in Philosophy symposium at the 2019 Canadian Philosophical Association Meeting, a 2019 Arché reunion conference, a 2019 conference at Peking University on Knowledge, Context, and Responsibility, and a 2019 conference at Glasgow University of Epistemic Injustice, and Blame, where I received a lot of useful feedback from many people. I owe particular thanks to Derek Ball, Aliosha Barranco Lopez, Sam Berstler, Christopher Blake-Turner, Hannah Bondurant, Jessica Brown, Roger Clarke, Kristin Conrad, Bianca Crewe, Alex Davies, Alexander Dinges, Davide Fassio, Jennifer Foster, Miranda Fricker, David Friedell, Jie Gao, Georgi Gardiner, Joaquim Giannotti, Cam Gilbert, Sandy Goldberg, Jasper Heaton, Cassie Herbert, Torfinn Huvenes, Carrie Jenkins, Ira Kiourti, Alison Duncan Kerr, Os Keyes, Marc Lange, Karen Lewis, Qilin Li, Yong Li, Stefan Lukits, Federico Luzzi, Mauricio Maluff Masi, Elinor Mason, Matt McGrath, Robin McKenna, Michaela McSweeney, Shoshana Messinger, Chris Mole, Graham Moore, Ram Neta, Holly Onclin, Molly O’Rourke-Friel, Phyllis Pearson, Kathryn Pogin, Geoff Pynn, John T. Roberts, Gillian Russell, Geoff Sayre-McCord, Dan Shahar, Kevin Scharp, Jonathan Shaheen, Mona Simion, Zach Thornton, Emily Tilton, Lauren Townsend, Kelsey Vicars, Sam Wakil, Yanjing Wang, Jenna Woodrow, Alex Worsnip, Audrey Yap, Ru Ye, Yiwen Zhan, Weiping Zheng, and others, for helpful discussions of these ideas and early drafts of this material. Thanks to referees and editors for Ergo, the Canadian Journal of Philosophy, and the Kennedy Institute of Ethics Journal for critical feedback. Thanks to Cam Gilbert for proofreading. Work for this paper was funded in part by a Canadian SSHRC Insight Grant for a project on rape culture, testimony, and epistemology.