1.1 Testimonial Resources, Literary and Social Psychological

One source of evidence is testimony from those with lived experience of unintentional racism. One recent source of testimonial evidence is Claudia Rankine's (Reference Rankine2014) Citizen: An American Lyric, which provides a rich and bleak compendium of the everyday slights and insults experienced by Black Americans. Rankine presents testimonies, based on her own and others’ experiences, articulated in the second-person ‘you’ form, in order to position the reader as ‘an apparent part of the encounter.’ (Rankine explains two other reasons for this form of address: first, that since the experiences were not all her own, she did not want to ‘own’ them with the first person pronoun, and second, that addressing the readers as ‘you’ forces them to make assumptions about race: ‘you're race-ing these people in order to understand this dynamic. I wanted that positioning to happen for readers’ [Rankine, quoted in Sharma Reference Sharma2014].) For example:

In these passages Rankine explores the way neighbors, strangers, and friends are implicated in—perpetrators of or party to—a daily grind of racist slights: assuming as other, as dangerous, intrusive, out of place.

The following features of Rankine's examples are especially important to the current analysis. First, these insults are presented as slips or unintentional or unthinking. This is not to say they are driven by implicit bias, but rather that perpetrators of them may not be driven by ill will (though this is not precluded). They would fit Yamato's (Reference Yamato, Anderson and Collins2004) category of ‘unintentional racism’. Second, prominent alongside these descriptions is the grappling with ambivalence: ‘what did he just say? Did she really just say that? Did I hear what I think I heard? Did that just come out of my mouth, his mouth, your mouth?’ (2014: 9). Third, the deniability of what happened hinders the ability to definitively label any one instance racism: couldn't the incident be rationalized some other way (an honest mistake)? Compare also the title of Yamato's piece: ‘Something about the Subject Makes it Hard to Name’. Fourth, dealing with the ambivalence that results, in part from the deniability, is part of the burden that Rankine describes. But, fifth, the cumulative effect reveals—to the protagonist and to the reader—a pattern of treatment underwritten by a set of associations that others apply to her and that she carries with her: Black, dangerous, intrusive, other, trespassing in white space.

Testimonies can also be found in the literature on experiences of microaggressions. In a series of focus groups, Sue and colleagues (Reference Sue, Capodilupo and Holder2008) gathered testimonies of people of colors’ experiences of microaggressions and identified commonalities in participants’ experiences. One of the prominent themes—one that resonates with Rankine's writings—is the deniability of the behavior and the possibility of rationalizing it away as something other than racism, in spite of how it was experienced. Sue et al. write of ‘Another participant['s] . . . struggle’ when a white woman changed seats on the train from sitting next to her:

Notwithstanding the ambiguity, the understanding of how one might be perceived—dangerous, out of place, incompetent—is reported as ever present and shaping interactions:

Others reported strategies for dealing with this, notably, identifying what has gone wrong with the interaction in terms of unintentional or unconscious white racism:

However, the participants emphasized the pervasiveness of the slights and the burden of interpreting or challenging every one of them: ‘If you were to address every microaggression, it's like all “Oh there you go again, you people”’ (2008: 333).

At issue here again is a failure to be able to communicate effectively due to the incompetence of one's audience, what Dotson has called ‘testimonial smothering’ (2011: 244). What one wants to express is that harmful discrimination has occurred, but the behavior is not recognized as such by the perpetrators.

These examples show how testimony from lived experience can provide evidence about unintentional racism. Note also that again, there is emphasis on the deniability of the discriminatory treatment; there is attention to the difficulty of identifying it as an instance of discrimination; and the difficulty of confronting the perpetrator who (it is anticipated) shares a different way of understanding the incident. As we can see, testimonial evidence—whether from literature, qualitative research in social psychology, or elsewhere—has the potential to be a powerful tool wherever evidence of unintentional racism can be used to motivate and shape institutional change.

1.2 Evidence from the Social Psychology Research Program on Implicit Bias

A second source of evidence about some instances of unintentional racism is psychological research on implicit bias. Many philosophers—and, indeed, many people outside the academy—are by now familiar with the notion of implicit bias. Implicit biases are the mental constructs that lead people automatically to associate members of social groups with certain characteristics (though there are competing ways of understanding what these mental constructs are, see, e.g., Holroyd, Scaife, and Stafford Reference Holroyd, Scaife and Stafford2017). These automatic thought processes may encode stereotypes or negative evaluations contrary to the thinker's professed values. For example, studies show that people who report egalitarian principles can harbor implicit biases that encode racist stereotypes and evaluations. Implicit biases are hypothesized to be at work in studies that reveal discriminatory behaviors. Implicit biases are recognized to be pervasive but difficult to detect and to control (see Jost et al. Reference Jost, Rudman, Blair, Carney, Dasgupta, Glaser and Hardin2009 for an overview).

Psychological studies purport to measure such biases using indirect measures. These are experimental techniques that bypass participants’ reflective self-reports of their beliefs or attitudes. For example, the Implicit Association Test (IAT) measures how quickly and accurately participants respond to stimuli representing a social group (e.g., white or Black) and stimuli representing concepts such as loyal, honest or abuse, murder (on an evaluative IAT), or brainy, smart vs. athletic, rhythmic on a stereotype IAT (Amodio and Devine Reference Amodio and Devine2006). The speed and accuracy of participants’ responses are used to generate a score that reflects the difference in response times with respect to each social group (e.g., how much faster at categorizing white with positive notions or stereotypes than Black with positive notions or stereotypes). It is inferred that this indicates how strongly participants associate the respective social groups with the target concepts (the evaluations, positive or negative, or stereotypes). Typically, these studies indicate that participants hold biases against Black people.

Studies also look for the impact of implicit bias on discriminatory behaviors. For example, studies that focused on employment contexts found that participants made differential evaluations of the same CV and differential hiring recommendations based on interview materials depending on the race of the candidate (Dovidio and Gaertner Reference Dovidio and Gaertner2000). In medical contexts, practitioners who participated in studies made differential treatment prescriptions to white or Black patients based on the same presentation of symptoms, and their tendency to do so correlated with the extent to which they showed implicit racial bias on indirect measures, that is, in the IAT (Green et al. Reference Green, Carney, Pallin, Ngo, Raymond, Iezzoni and Banaji2007). In studies that looked at behaviors that may manifest in everyday interactions, white participants were found to sit further away from Black individuals, smile less, converse less, and manifest more errors of speech (McConnell and Leibold Reference McConnell and Leibold2001). If such behaviors issue from implicit bias, then they may be instances of unintentional racism, perpetrated without ill will or explicit prejudice. Researchers on implicit bias are at pains to emphasize the pervasiveness of the phenomena; the majority of white participants and perhaps half of Black participants show patterns of anti-Black bias. And it is hypothesized that the occasions in which bias might be most likely to influence behavior are those in which one's behavior can be rationalized by some other, non-race-related considerations (Dovidio and Gaertner Reference Dovidio and Gaertner2000).

Drawing on such findings has been a catalyst for change, for example:

1. 1. Enabling individuals (who may be in a position to enact institutional change) to gain a greater awareness of the possibility that they are perpetrators (albeit unintentionally and despite goodwill) of discrimination;

2. 2. Opening up conversations about discrimination and inclusion, which may prove easier if they do not involve attributions of ill will or explicit prejudice (however, see Eddo-Lodge [Reference Eddo-Lodge2017: xii] on the problems with tailoring discussions of race to accommodate white people's discomfort);

3. 3. Motivating institutional reforms that combat unintentional racism, such as: reform of reading lists, hiring procedures, workplace climate policies, and so on (see, e.g., Saul Reference Saul2018; Holroyd and Saul Reference Holroyd and Saul2018 for an overview of these reforms in academia).

Recent impetus to make institutional changes to combat unintentional racism has been generated in large part due to discussions drawing on the implicit bias research program. But despite these gains, there are problems with drawing on this evidence alone—it is these that we aim to highlight.

2.1 The Vindication Dynamic

Our concern in this section is an epistemic dynamic that is described by Patricia Hill Collins in Black Feminist Thought (2002). Collins provides the following example:

Testimony provided by Hemings and her descendants was denied credibility until scientific tests supported the claims found therein. Faced with conflicting testimonial evidence, racist norms of credibility ensured that testimony from white speakers was afforded credibility that the testimony of Hemings and her descendants was not. Yet Hemings's testimony was ultimately vindicated by DNA evidence (Gordon-Reed and Takagi Reference Gordon-Reed and Takagi1997). While this vindication is important, there is something problematic in the idea that certain kinds of testimony—testimony from people of color, especially when it challenges existing power structures and hierarchies—stand in need of vindication. The assumption that Hemings's testimony alone cannot provide sufficient reason for belief, absent verification from other sources of evidence, is part of a long history of epistemic exclusion.

The vindication dynamic may be thought of as an instance of what Kristie Dotson and Marita Gilbert (Reference Dotson and Gilbert2014) have recently labeled ‘disappearances’: as they put it, ‘the lives and experiences of Black women are rendered invisible in public narratives’ (Dotson and Gilbert Reference Dotson and Gilbert2014: 873), pointing to the work of Fannie Barrier Williams (Reference Williams1905), Zora Neale Hurston (Reference Hurston1950), Audre Lorde (Reference Lorde, Braxton and McLaughlin1990), and Rebecca Wanzo (Reference Wanzo2009). For example, Zora Neale Hurston (Reference Hurston1950) described the one-dimensional stereotypes that obscure the intellectual and emotional lives of Black Americans. Rebecca Wanzo (Reference Wanzo2009), writing more than half a century later, argues that while narratives of Black women's lives are present in public discourse, those narratives often homogenize experiences and focus on personal, rather than social, transformation as a way of overcoming oppression, thus obscuring the reality and experience of systemic racism. Our focus in this section is on one epistemic dynamic that can render narrative and testimonial evidence of racism invisible, ensuring that it is often not present in contemporary discourses about discrimination (cf. Davidson Reference Davidson, Sherman and Goguen2019, who raises similar concerns about the use of implicit bias evidence in academic philosophy; Davidson focuses on how the use of the empirical evidence prevents widespread [‘third-order'] change to epistemic systems).

Collins takes the treatment of Hemings to illustrate the ways in which ‘power relations shape who is believed and why’ (2002: 270). In this case, power dynamics of race and gender result in the testimony of a Black woman being treated as unreliable. The contributions made by her testimony are therefore marginalized and undermined as a source of knowledge, and moreover, the subsequent exclusion of such knowledge conveniently serves to sustain certain myths about race and racial purity (e.g., that whites were superior, that whites and Blacks should not and did not miscegenate, and that slave owners’ sexual abuse of enslaved women was not one of the ways in which the latter were exploited), the conduct of whites in power (as morally impeccable, rather than as exploitative, morally corrupt, and as perpetrators of abuses with impunity), the place of Black people in America (as passively, indeed naturally, subservient). Collins explores the ways that such epistemic dynamics marginalize important understanding of experiences of oppression, and Black women are further harmed by their exclusion from the epistemic community and the barriers to getting the nature of their social experiences on the record.

This is just one form of what Dotson calls epistemic oppression: ‘the epistemic exclusions afforded positions and communities that produce deficiencies in social knowledge’, whereby an exclusion concerns infringements that reduce knowers’ ability to participate in epistemic communities (2012: 24). Our suggestion here is that such a dynamic is risked in discussions of unintentional racism focusing solely on quantitative psychological research on implicit bias. To understand this point, let us probe the example of Hemings further. In this case there are multiple epistemic wrongs. In the first instance, there is the wrong of the initial epistemic exclusion of Hemings (and indeed, slave women in general). To the extent that Hemings's testimony is later vindicated by the emergence of DNA evidence, it is important to note that it is the proposition at issue that is vindicated. Her testimony is regarded as something that stood in need of post hoc vindication. Her status as a knower and her testimony alone was and remains insufficient to provide the evidential support for the proposition that Jefferson fathered her children (despite her being, one might think, uniquely placed to determine this and despite the consistency of her account with what is widely known about the sexual exploitation of enslaved women by white owners).

Note that while the later means of vindication might itself be epistemically respectable—endorsing the findings of well-conducted DNA studies is a rational response to the evidence—it nonetheless contributes to an oppressive epistemic dynamic. Even though the moment of vindication is not itself one in which we see an exclusion, it puts in stark relief the epistemic norms that cast some knowers as unreliable sources of knowledge. The vindication dynamic contrasts with cases in which some initial evidence, taken to have probative value, is later given further support by being confirmed or corroborated. In cases of the vindication dynamic, the initial testimony is not considered to have probative value, which is precisely why the later evidence is vindicatory. To the extent that this vindication dynamic entrenches norms that shape who is and is not reliable or which testimony does or does not stand in need of vindication, it contributes to patterns of epistemic oppression.

What is the relevance of this dynamic to our practical question of to what evidence we might appeal in supporting the proposition that discrimination can be unintentional? Focusing solely on quantitative evidence from the research program on implicit bias sidelines preexisting testimonial evidence about racism, foregrounding evidence emerging from academic studies, generated in laboratory conditions. It also marginalizes the work of people of color who have articulated these aspects of their experience, thereby marginalizing evidence based in lived experience. This has particularly consequential racial dynamics, insofar as, as Collins (Reference Collins2002) writes, such evidence has particular importance in Black women's epistemic communities. Conversely, evidence emanating from the predominantly white academy is privileged. Collins's diagnosis focuses in particular on hierarchies of knowledge (academic research vs. testimonial), which are themselves racialized. This explanatory framework is more plausible than an explanation that appeals just to the race of those producing the research because in this case much prominent research on implicit bias is conducted by scholars of color (e.g., J. Eberhardt, M. Banaji, C. Lai). As Collins argues, testimonial evidence is not valued in this hierarchy, but in particular the testimony of women of color is devalued.

While there is something very important about providing evidence for the proposition at issue—namely, that racism can be perpetrated unintentionally —there are risks involved with doing so by appeal only to the quantitative research program on implicit bias. While the proposition for which there is testimonial evidence—that there exists unintentional racism—is confirmed by appeal to evidence of implicit bias, the emphasis on that evidence risks entrenching certain oppressive epistemic dynamics. In particular, it risks entrenching the message that a certain kind of testimony—qualitative testimonial evidence based in experience, specifically the experiences of people of color—is insufficiently credible and stands in need of vindication. It risks entrenching the notion that this, rather than that evidence suffices to warrant belief. (One might think that evidence from research on implicit bias is helpful in that it increases the credibility of future testifiers about racism. This does not mitigate the problem for past and present and future testifiers, whose words apparently need vindication.)

Focusing only on the psychological research on implicit bias also risks conveying the message that the kind of evidence that is important is that which affirms the existence of (the possibility of) unintentional racism by focusing on the cognitions of the perpetrators. The ‘mere’ experience of those targeted by the discrimination, according to these norms, is insufficient to warrant belief. To prioritize psychological research on implicit bias is to suggest not only that testimony alone is insufficient, but that a certain kind of evidence is needed to vindicate these ‘perceptions’—evidence concerning the cognitions of the perpetrators.

In drawing attention to the vindication dynamic, we do not need to establish that (as is plausibly the case with Hemings) the disregarding of testimonial evidence is in each instance a case of testimonial injustice in the sense that has become familiar through the work of Miranda Fricker (Reference Fricker2007), namely, involving prejudicial judgments of deflated credibility due to one's social identity. We do not need to maintain that all, or even the majority, of cases in which people appeal solely to quantitative psychological findings on implicit bias when discussing unintentional racism are cases in which those people are displaying prejudice or making judgments of reduced credibility. Nor do we want to maintain this, because we lack the data about the cognitive states of those who have appealed to implicit bias evidence. What is important to the current discussion, however, is that focusing solely on psychological research on implicit bias serves to entrench certain hierarchical norms and to perpetuate epistemic exclusions that produce deficiencies in social knowledge (Dotson Reference Dotson2012: 24). Failing to heed someone's testimonial evidence is to proceed as if it lacks credibility even if such judgments have not been made. The dynamic that we describe thus instantiates a form of epistemic oppression.

In selecting one's evidence, then, we must be alert to the risk that attention only to psychological research on implicit bias risks entrenching a set of epistemic norms whereby the status of testimony of lived experience of racism is devalued. Moreover, focusing solely on the psychological research on implicit bias risks entrenching oppressive norms about the kind of evidence that is reliable: the lived experience of people of color is devalued and stands in need of vindication from more ‘respectable’ sources of evidence.Footnote ¹

2.2 Contributory Injustice

We now consider a second risk faced by focusing on the research program on implicit bias rather than testimonial evidence: the risk of perpetrating a specific form of injustice identified by Kristie Dotson, i.e. contributory injustice.

Contributory injustice is defined as the circumstance where:

Hermeneutical resources are our resources for making sense of the world—concepts, theoretical frameworks, narratives, practical starting points or assumptions, and so forth. They are structurally prejudiced if the ability of people in different social locations to contribute to the shared hermeneutical resources is asymmetrical. For example, if white people are able to participate in the shaping of the resources available to make sense of the social world to a greater extent than Black people, those resources are structurally prejudiced. Such a pattern of participation will produce hermeneutical resources that are unduly influenced by those hermeneutically advantaged people who have contributed—usually, those in positions of power and influence—and insufficiently influenced by members of the marginalized groups themselves. This asymmetrical influence is likely to result in a set of resources that is a better ‘fit’ for the experiences of those in a hermeneutically advantaged position and a worse ‘fit’ for the experiences of those in hermeneutically disadvantaged positions in that it lacks the conceptual resources or interpretative meanings that are needed to capture the experiences of the latter group(s) adequately. Marginalized groups may have developed their own hermeneutical resources: ‘alternative epistemologies, countermythologies, and hidden transcripts’ (Dotson Reference Dotson2012: 31). These alternative sets of hermeneutical resources are available for marginalized individuals to understand their social experiences, but are not part of shared collective resources.

Contributory injustice occurs, then, when the resources that marginalized people use to interpret and articulate their experiences are not acknowledged or taken seriously in an epistemic community and when the dominant hermeneutical resources used in that community thwart those individuals’ ability to contribute to the shared resources (the alternative resources they do have available are not widely understood or regarded as useful conceptualizations). An example of contributory injustice is provided by Reni Eddo-Lodge, who describes how the concept of ‘intersectionality’—developed by Black feminist scholars to articulate how gender and racial oppressions intersect to disadvantage Black women—has been refused uptake by writers in the mainstream media in the United Kingdom, ostensibly for being ‘utter jargon’ (Reference Eddo-Lodge2017: 140).

On Dotson's account of contributory injustice, these structurally prejudiced hermeneutical resources are utilized by an agent or agents due to their willful ignorance. Willful ignorance, for Dotson, ‘maps onto [an] epistemically culpable failure’ (2012: 33); thus, at stake here are cases of ignorance where one should have known better. On Dotson's account, it is easy to fall short of the relevant epistemic standards. Her claim is that contributory injustice is very easy to commit and very difficult to avoid (2012: 38–40). Dotson argues that people can perpetrate contributory injustice without being aware of doing so, without thinking that they are putting a foot wrong. Accordingly, the degree to which the failure is epistemically culpable could vary considerably. Because contributory injustice is difficult to avoid, one might think that the degree of culpability may, in some instances, be minimal.

With this exposition of contributory injustice in place, we now consider how one's selection of evidence—in particular, a selection that focuses solely on research on implicit bias—could perpetrate contributory injustice. Earlier, we presented what can be thought of as two sets of hermeneutical resources that can be used to frame one's understanding of unintentional racism and exclusion. One set of hermeneutical resources is grounded in the lived experiences of individuals who have experienced and witnessed discrimination. The other set of hermeneutical resources has been developed in social psychology and is concerned with the notion of implicit bias and its role in unintentional racism. Of course, these two sets of resources need not be exclusive; they may be partially overlapping, as we saw above, when some participants in Sue et al.'s (Reference Sue, Capodilupo and Holder2008) studies refer to the notion of unconscious bias to explain their experiences.

Here, we want to argue that in selecting one's evidence and choosing to focus solely on the resources from the quantitative research program on implicit bias, one may perpetrate contributory injustice. To show this, we need to show that such resources are structurally prejudicial, that use of them thwarts other knowers from contributing to the epistemic community, and that the use of them can be understood as an instance of willful ignorance.

Recall that hermeneutical resources are structurally prejudiced if the ability to contribute to those resources of people in different social locations is asymmetrical, with the likely outcome that those resources differentially reflect the social experiences of the contributing groups. To the extent that the collective resources utilized in discussions of implicit bias are technical tools developed within academia (psychology, philosophy), the exclusion of Black and minority ethnicity individuals from academia is one way in which differently racialized groups are asymmetrically positioned to contribute to those resources. In the United States, the National Centre for Education Statistics (2020) found that in 2018, 75 percent of full-time faculty were white, and among professors, 80 percent were white, with Black men and women comprising 4 percent of all professors. In the United Kingdom, a recent study indicates that ‘black academics constitute 1.54% of the total UK academic population a significant underrepresentation in terms of the broader resident black population being 3.3%’ (Shilliam Reference Shilliam2015: 32). At professorial level fewer than 1 percent are Black, 25 (raw figure) of whom are women (Rollock Reference Rollock2019). Insofar as psychology is not an outlier in terms of representation, the context in which the research on implicit bias is being done is one where the ability of Black and minority ethnicity individuals to contribute to those resources is asymmetric, simply insofar as these individuals are drastically underrepresented. This will be exacerbated in conditions, such as those reported on by Bhopal and Jackson (Reference Bhopal and Jackson2013), in which Black and minority ethnicity academics feel excluded, stereotyped, and overscrutinized. And, as Romdenh-Romluc argues, hermeneutical marginalization can remain even when individuals have access to hermeneutical power and are in positions to shape meanings—even then, their hermeneutical marginalization is likely to show up in the social meanings available for use (Reference Romdenh-Romluc2016: 596).

Alongside the asymmetries in who contributes to the shared resources, we see differences in what kinds of evidence are afforded significance. To the extent that testimonies grounded in lived experience of racism are devalued, individuals offering such understanding for incorporation into the collective hermeneutical resources are thwarted in their attempts to do so. The case to be made here draws in part on the claims set out in our discussion of the vindication dynamic above. There, we drew upon Patricia Hill Collins's claims that Black women's knowledge has been marginalized, and in particular knowledge that is developed through and communicated in testimonies grounded in lived experiences of oppression is devalued. Already marginalized, these testimonies would be further disregarded if we were to focus solely on evidence pertaining to implicit bias.

To a certain extent, this is a contingent matter: given limited air time or journal space, if that is given to quantitative psychological research on implicit bias rather than research that attends to lived experiences of racism, the latter are marginalized. However, this pattern is one that is entrenched in a set of epistemic dynamics that devalues the perspectives and experiences of people of color, in particular women of color. Indeed, Dotson draws on Audre Lorde (Reference Lorde, Braxton and McLaughlin1990) to tease out different mechanisms through which Black women's experiences (‘words, work and lives’) have been excluded: scrambling or distorting their words (e.g., by dissembling that one simply cannot understand what is being said), erasure, and disregard (2012: 33). On the one hand, as mentioned above, psychological research on implicit bias may in some sense serve to counteract the tendencies to scramble or erase experiences of racism—indeed, to focus on implicit bias is to point to evidence of its pervasiveness and to affirm experiences of unintentional racism. However, to draw only on such evidence is nonetheless to disregard the contributions to addressing racism that have been made by those who have experienced and witnessed it and to focus instead on the experimentally revealed dispositions of the perpetrators. This is to perpetuate a certain form of exclusion—both of certain knowers and of a certain kind of knowledge—namely, that grounded in experiences of the oppressive structure of the social world and presented in the hermeneutical resources that capture those experiences.

So, the hermeneutical resources are structurally prejudiced in terms of who contributes to them, and this corresponds to the kinds of knowledge and understandings that are given prominence and marginalized, respectively, in those shared resources. In particular, testimony offers up a different aspect of the phenomena at issue: what it is like to experience persistent, everyday racism. The resources needed to describe this aspect of the social world will, of course, be different from those required to describe the cognitive mechanisms that might underpin such patterns of discrimination. Our resources will be structurally prejudiced if the interpretative resources we have available to understand unintentional racism become dominated by or excessively focused on the cognitions of the perpetrators. As such, the shared or dominant resources may be asymmetrical in having rich resources to capture the phenomenon of perpetrating unintentional racism, but fewer resources—or a gap in the resources—to understand what it is like to experience such discrimination. This is to the disadvantage of anyone trying to register in the shared collective resources such experiences and the concepts needed to make good sense of them.

This exclusion thwarts the abilities of knowers to contribute further to the shared hermeneutical resources. One way these contributions are thwarted is simply the lack of uptake they receive. One cannot contribute to the shared hermeneutical resources if audiences do not hear or attend to the testimonies available. A second way in which such contributions are thwarted is a matter of emphasis; to the extent that evidence about experiences of discrimination is available, if it is not regarded as important or useful, it will be not be fully incorporated into the shared conceptual resources. The conceptual space for such meanings and understandings is there, but relegated to an unimportant or peripheral aspect of the phenomenon that does not significantly advance collective understanding. A third way in which contributions may be thwarted is if the failure to get any uptake or to be regarded as contributing valuable concepts for understanding the phenomenon results in downstream ‘testimonial smothering’ (Dotson Reference Dotson2011: 244). Testimonial smothering occurs when knowers fail to speak in a particular context, having learned that they will not be heard. While the harms of hermeneutical injustice and contributory injustice have already been articulated (see, e.g., Fricker Reference Fricker2007; Dotson Reference Dotson2011), it is important in this context to emphasize that the collective resources of an epistemic community are weakened overall if the evidence from testimony about lived experiences of racism is not included—there are epistemic costs all round in addition to the harm that accrues to individuals who are unable to contribute. We elaborate on this in the following section.

Finally, we contend that the use of these structurally prejudiced resources is (often) due to willful ignorance. Recall that on Dotson's view, it is very easy to commit contributory injustice and very difficult to avoid it. Overlooking evidence that one really should be aware of can suffice to make this a case of willful ignorance. An overemphasis on the quantitative psychological literature on implicit bias when one is aware of other sources of evidence would constitute willful ignorance on this construal.

Reflection on discourse about implicit bias, however, brings to light that there may be other ways of perpetrating contributory injustice that are not grounded in willful ignorance. For example, as mentioned above, there have been some gains from the attention to quantitative evidence of implicit bias, and one might make a strategic choice to focus on this evidence rather than on testimony because, for example, one thinks there will be greater uptake among one's peers of the psychological research program on implicit bias. There can be benefits to appearing to conform to oppressive epistemic norms of this type—norms such as ‘assign comparatively higher value to (quantitative) evidence generated by (white) academic experts’—under specific, localized circumstances. For example, Michael Doan (Reference Doan2018) describes how activists and residents fighting for access to clean affordable water in Michigan (USA) benefited from conforming to epistemic norms giving preference to expert testimony from academic researchers. They (appeared to) conform to these norms by getting such experts engaged in their cause while also pushing back against the comparative value judgment expressed in the norm by highlighting the contribution that could be made by the stories of residents (cf. Medina Reference Medina2013).

We do not wish to underplay the benefits of this type of strategic conformity to problematic norms, but we do wish to highlight the associated risks. While not setting aside testimony due to willful ignorance, this kind of strategic choice may still use and maintain structurally prejudiced hermeneutical resources that thwart others’ abilities to contribute to the shared epistemic resources in the ways we described above—especially if it does not at the same time challenge the oppressive epistemic norms. That is, such a strategic choice in this context may still place focus on quantitative research about the cognitions of discriminators at the cost of attention to testimony of those who have experienced discrimination. (Our claim is not that these strategic choices could never be justified but that they come with certain risks, consideration of which is needed in a full assessment of the strategic choice.) Thus, we propose an expansion of the notion of contributory injustice, to extend not only to exclusions that result from willful ignorance but also to exclusions based in strategic choices. However, if one is concerned to preserve the notion of contributory injustice only for those cases grounded in willful ignorance, then it suffices for our purposes to claim that something very much like it—the use of structurally prejudiced hermeneutical resources that thwarts knowers’ abilities to contribute to shared epistemic resources, thus compromising their epistemic agency—can result from strategic choices as well as from willful ignorance.