Testimonial Injustice from Countervailing Prejudices

ABSTRACT

In this paper I argue that Fricker’s influential account of testimonial injustice (hereafter ‘TI’) should be expanded to include cases of TI from mutually neutralising countervailing prejudices. In this kind of case, the hearer is given due credibility by the speaker. I describe a relevant case, defend it from objections, highlight how it differs from extant cases of due-credibility TI and describe its distinctive features. This case demonstrates how paying attention to the way multiple prejudices operate in concert leads to expanding the repertoire of cases of TI and casts light on the harms of TI.

KEYWORDS:

• Testimonial injustice
• credibility
• prejudice
• countervailing prejudices

1. Introduction

Several critical expansions to the category of testimonial injustice (hereafter `TI’) have been suggested since Fricker’s (2007) definition of TI as the injustice done to a speaker when a hearer deflates their credibility for identity-prejudicial reasons.¹ For example, it has been argued that TI should not be restricted to assertion, since speech acts such as asking questions and putting forward hypotheses can be the proper targets of TI (Hookway 2010); that TI can occur when even when no words are uttered at all, as when a speaker self-censors in the justified expectation that they will not be taken seriously (Dotson 2011), or when a would-be speaker withholds important information from someone, owing to prejudice (Carmona 2021); that TI does not require identity prejudice, since structural features like domination are better suited to understand and explain the workings of TI (Bohman 2012), or that it does not require any kind of prejudice at all, since unduly speaking on behalf of others counts as TI, and this need not occur for prejudicial reasons (Steers-McCrum 2020). These constitute valuable and interesting suggested modifications to Fricker’s account, but I will largely set them aside in this paper to discuss the expansion of Fricker’s account most relevant for our purposes – that TI need not involve credibility deficit.

Fricker’s account explicitly endorses the view that credibility deficit is a necessary condition on TI. In recent years, however, some critics have denied this by proposing alleged cases of TI of two broad kinds: cases in which the hearer grants the speaker too much credibility and cases in which the hearer grants the speaker the right amount of credibility. Cases of the first kind include a speaker being called on tokenistically and against their will to speak for the social group they belong to (Davis 2016) and cases in which the credibility excess afforded to one social group is accompanied by an associated deficit inflicted on another social group (Medina 2011, 2013). While these excess-based cases place independently interesting pressure on Fricker’s account, in this paper our focus will primarily be on cases of TI in which the hearer grants the speaker due credibility.

Two authors have discussed cases of this kind in the literature: Lackey (2018) and myself (Luzzi 2016). My aim here is to highlight a new kind of case of TI that Fricker has not considered but, I argue, ought to accept. These are cases of TI in which due credibility is granted, but they differ significantly from extant cases of due-credibility TI. Because of these differences, my argument does not strictly depend on accepting Lackey’s and my prior cases, even though I am in agreement with the broader claim towards which our cases point, namely, that Fricker should accept due-credibility TI in her account.

Importantly, the new kind of case I will describe arises when we pay attention to how an individual’s belonging to multiple social groups affects, both positively and negatively, the credibility they are given. Of course, it is widely acknowledged that a variety of prejudices may individually give rise to TI (those concerning the speaker’s sex, gender, race, ethnicity, trans status, disability, religion, social class, etc.), and it is a short step from here to the plausible point that speakers who simultaneously belong to multiple groups targeted by prejudice can suffer compounded, distinctive and distinctively severe forms of TI. What is less evident is how negative and positive prejudices can act in tandem to produce TI. For example, a gay man may find that in certain contexts his word is given less-than-due credibility by a prejudiced audience, due to his sexual orientation. In certain distinct contexts, he may find his word given more-than-due credibility by a prejudiced audience that takes men more seriously than people of other genders. Crucially, in some contexts his audience may hold both kinds of prejudices concomitantly and to similar degrees, and he may as a result be given the right amount of credibility. Similarly, the key case I present below, Countervailing Biases, relies essentially on a speaker’s simultaneously belonging to two social groups, where one group has their credibility routinely and prejudicially deflated, while the other group has their credibility routinely and prejudicially inflated. The speaker is Black and recognisably upper-class, and the negative credibility deflation they would receive due to their audience’s racial prejudice is neutralised by the audience’s positive view of upper-class people’s epistemic competence. The net effect is that the speaker is granted the right amount of credibility mandated by the hearer’s evidence; thus, the hearer fulfills what Fricker takes to be their ethico-epistemic obligation. Despite this, the speaker suffers TI, or so I will argue, so this is a kind of due-credibility TI. This demonstrates that Fricker’s characterisation of the hearer’s ethico-epistemic obligation stands in need of correction and that the repertoire of cases of TI should be expanded in a novel way. More specifically, the case I discuss suggests that whether TI occurs does not depend on the speaker being prejudicially denied the right level of credibility, but rather on prejudice being operative in the formation of the hearer’s credibility judgment about the speaker.

This kind of due-credibility TI has two interesting features that distinguish it from extant cases of due-credibility TI: it is especially difficult to detect, because in its presence knowledge is seamlessly transmitted (unlike in Lackey’s case), and relatedly, it does not involve the hearer denying that the speaker speaks from knowledge (unlike both Lackey’s and my prior cases). This latter observation demonstrates that my previous suggested fix for immunising Fricker’s account from due-credibility TI is mistaken. After presenting my original cases of due-credibility TI, I suggested on Fricker’s behalf that her account could avoid countenancing this kind of TI if her conception of granting due credibility were strengthened so that it required the hearer to correctly ascribe knowledge to the speaker rather than mere true belief. But the case I will describe shows how even with this strengthened conception of credibility in place, Fricker would still have to admit due-credibility TI into her view. Additionally, the case I will discuss also prompts a hypothesis for what, minimally, the harm of TI amounts to. My main aim, then, is to cast light on TI arising from differently valenced prejudices in order to enhance our understanding of the variety of kinds of TI and of the problematic feature that unifies diverse cases.

I will start by sketching the relevant elements of Fricker’s account of TI (section 2). In section 3, I describe the key case (Countervailing Biases) and defend it from objections. I then demonstrate how it differs from extant cases of TI with due credibility discussed by Lackey and myself (section 4); finally, I will draw some lessons concerning Fricker’s conception of the harm of TI (section 5).

2. Testimonial Injustice: A Concise Characterisation

Fricker (2007) understands TI as the injustice done to a speaker when, owing to prejudice, their audience denies them the credibility they deserve. Fricker characterises the relevant prejudice harboured by the audience as identity prejudice – a prejudice about the speaker’s social identity – and focuses on so-called systematic cases, where the prejudice in question ‘“tracks” the subject through different dimensions of social activity – economic, educational, sexual, legal, political, religious, and so on’ (Fricker 2007, 27). Of this central type of case, Fricker states:

With the concepts of identity prejudice and systematicity in place, we are now in a position to propose a refined characterization of the central case of testimonial injustice—the systematic case. The speaker sustains such a testimonial injustice if and only if she receives a credibility deficit owing to identity prejudice in the hearer. (28)

As this quote makes clear, Fricker’s account posits that granting credibility deficit driven by identity prejudice is both necessary and sufficient for TI.

To attribute to a speaker a credibility deficit of this kind is for Fricker a violation of one’s duties, understood as follows:

Epistemological nuance aside, the hearer’s obligation is obvious: she must match the level of credibility she attributes to her interlocutor to the evidence that he is offering the truth. (19)

Thus, in order to fulfill one’s ethico-epistemic obligations and avoid committing TI towards the speaker, the hearer must grant them the level of credibility mandated by the evidence that they speak truly; in particular, the hearer must not allow any identity prejudice to negatively distort the level of credibility granted to the speaker.

It is important to note that while Fricker speaks explicitly of ‘matching the evidence’ here, in other passages Fricker’s discussion suggests that the way we ought to form judgments about the credibility of others is not a matter amenable to precise formalisation and may admit of imprecision (Fricker 2007, 75–76).² This point will be important later. For the time being, we should simply bear in mind that for Fricker, giving due credibility to a speaker involves the hearer’s ensuring that their judgment of the speaker’s credibility reflects the available evidence that what the speaker says is true.

We can see TI at work by considering two central illustrations proposed by Fricker. In To Kill a Mockingbird, Tom Robinson is a disabled Black man accused of the rape of Mayella Ewell in Alabama in the 1930s. During the trial, the white jury’s racist prejudice leads them to disbelieve Tom’s truthful version of events. Consequently, even though the evidence is clearly inconsistent with his guilt, Tom is convicted. In The Talented Mr Ripley, Marge Sherwood’s justified suspicions that her fiancé Dickie Greenleaf has been murdered are abruptly dismissed by Dickie’s father owing to his sexist prejudice that women are overly emotional and irrational. In each case, TI occurs when the audience (the jury and Dickie’s father) fail to give proper weight to the speaker’s word (Tom’s and Marge’s) on account of the prejudice they harbour (against Black people and women, respectively).

3. Countervailing Biases

I now move on to propose a case of due-credibility TI, one in which – unlike other proposed cases of due-credibility TI and unlike most other cases of TI discussed in the literature – multiple group membership plays an essential role:

Countervailing Biases: Kelly was educated at Eton and Oxford and now works as a lawyer. They speak with a recognisably upper-class accent. This makes Kelly appear more epistemically competent than they in fact are, and exerts upward pressure on the amount of credibility granted to Kelly by many interlocutors.

Kelly is also Black, so they are often the target of prejudicial treatment on the basis of their race. Consequently, to racially prejudiced audiences Kelly appears less epistemically competent than they are, and this exerts downward pressure on the credibility judgments about Kelly formed by many interlocutors.

One day, in a meeting with colleagues about an important case, Kelly says (before laying out their justification): ‘the defendant must be withholding important information’.

Suppose Kelly’s assertion is true, well-justified and that they speak from knowledge. Upon hearing Kelly’s statement, Kelly’s colleague Alex deems Kelly to be speaking truly and knowledgeably, as the evidence at Alex’s disposal indicates. So Alex’s credibility judgment about Kelly corresponds to the judgment mandated by the evidence; moreover, this credibility judgment is correct, since the evidence in this case is not misleading.

But in the formation of Alex’s judgment about Kelly, two kinds of prejudice are at work. Firstly, Kelly’s accent exerts upward pressure on this credibility judgment: other things being equal, this positive prejudice would have led Alex to form a more positive assessment of Kelly’s credibility than the evidence warranted. Secondly, Kelly’s race exerts downward pressure on Alex’s credibility judgment about Kelly: other things being equal, this negative prejudice would have led Alex to form a more negative assessment of Kelly’s credibility than the evidence warranted. As it happens, the two prejudices of opposite valence are of equal strength. So while in counterfactual circumstances, i.e. absent the operation of the other prejudice, each would have led Alex either to over- or under-estimate Kelly’s credibility, in the actual world Alex’s credibility judgment is the one Alex ought to hold, given the evidence. Therefore Alex comes to believe that the defendant is withholding important information and takes Kelly to know this.

My contention is that Fricker should accept that Countervailing Biases instantiates TI with due credibility. Even though Kelly is, in line with the evidence, deemed by Alex to speak truly, Alex still commits TI against Kelly. To make good on this claim, we must defend the view that the following two theses should be accepted by Fricker: (i) Alex grants Kelly due credibility, and (ii) Alex commits TI against Kelly. Let’s tackle these in order.

Regarding thesis (i), it is clear that Alex’s judgment about Kelly’s credibility is the right one to hold: Alex’s judgment that Kelly’s assertion is true (and knowledgeable) is stipulated to be the judgment warranted by Alex’s evidence. So Alex grants Kelly due credibility.

Thesis (ii) requires more substantial argumentation. To see why Alex commits TI against Kelly, let’s start by imagining that a second colleague of Kelly’s, call them Parker, also hears Kelly’s assertion and gives Kelly due credibility. Parker is not prejudiced against Black people or in favour of upper-class people, and the formation of Parker’s correct credibility assessment is uninfluenced by any further bias. Thus, while both Alex and Parker wind up with the same credibility judgment, Alex’s behaviour is morally worse than Parker’s: in a just world, people would assess credibility like Parker does, not like Alex. This indicates that Alex’s behaviour is at the very least non-ideal. I argue in the following that it is also unjust.

In what might the injustice consist? There is more than one way the injustice can be cashed out, but one perspicuous way is this: Alex behaves unjustly towards Kelly because they judge Kelly’s credibility in a way that is influenced by bias. That Alex ultimately reaches the credibility judgment warranted by the evidence is not sufficient reason to deny that this injustice occurs. Consider that if Alex had only harboured the negative bias against Black people’s epistemic competence (so that the credibility they ascribed to Kelly was lower than what the evidence warranted), Alex would very clearly have committed TI against Kelly. But to claim that in Countervailing Biases Alex does not commit TI is to say that TI can be averted by the operation of bias, if it is of equal strength and opposite valence to a negative bias, even if this bias is itself morally problematic.³ This is a version of the implausible ‘two wrongs make a right’ thought: in judging Kelly’s credibility, Alex is relying on not one but two morally problematic prejudices. The moral wrongness of Alex’s behaviour consists in their relying on these prejudices when forming their credibility judgment; that Alex ultimately winds up with the appropriate credibility judgment does not detract from the fact that they did so essentially by relying on their biases, and thus did so unjustly.

More support for the view that Alex should be deemed to have committed TI can be drawn from the fact that Countervailing Biases meets Fricker’s characterisation of the specific harm of TI. Fricker understands it as a form of objectification:

testimonial injustice demotes the speaker from informant to source of information, from subject to object. This reveals the intrinsic harm of testimonial injustice as epistemic objectification: when a hearer undermines a speaker in her capacity as a giver of knowledge, the speaker is epistemically objectified. (2007, 133)

This feature of Fricker’s view lends further support to my argument that Countervailing Biases should be accepted by Fricker as a case of TI, for it is plausible to think that prejudice-influenced assessment of a speaker’s credibility involves objectifying the speaker. A prejudicial assessment of a speaker’s credibility adverts to morally problematic and epistemically irrelevant factors, namely the speaker’s social group membership, in a way that denies them the individualism and subjectivity they deserve.⁴

Let me now address three objections, two swiftly and one more carefully, in order to further buttress the claim that Countervailing Biases instantiates TI. The first is that Kelly doesn’t suffer any kind of injustice – and a fortiori no TI – because Kelly is unable to detect that they are the subjects of biased judgment. After all, the workings of Alex’s credibility judgment are inscrutable to Kelly (and we can suppose that even Alex is not aware that their biases are influencing their assessment of Kelly’s credibility). Kelly’s life does not proceed any differently from how it proceeds in the counterfactual circumstance in which Alex judges Kelly’s credibility without any bias.

In reply, consider that there are independently plausible cases of injustice that carry no detectable negative practical consequence for the victim to serve as precedent. For example, if you steal the passwords to several of my online accounts and offer them for sale on the dark web, yet nobody buys them, and so no intruder ever logs into my accounts, then my life remains unchanged. But you have arguably still treated me unjustly by stealing my passwords and offering them up for sale. A more relevant example: unconscious bias may influence behaviour in ways that are difficult to detect, both for the victim and for the perpetrator. Its undetectability does not make the behaviour any less unjust.

Secondly, and relatedly, one might critically ask what makes the injustice suffered by Kelly distinctively epistemic – and indeed, testimonial – in nature. After all, Alex accepts Kelly’s assertion and grants Kelly the right amount of credibility. Knowledge is smoothly transmitted. If this is an injustice, how is it testimonial and epistemic? Is it not better assimilated to myriad other non-epistemic biased behaviours that target members of disadvantaged social groups, such as, for example, the prejudicial denial of mortgages to Black families?⁵

The appropriate response here is that the injustice suffered by Kelly is distinctively epistemic for the same reason that the central cases of TI advanced by Fricker are: because the object of the hearer’s biased judgment is an epistemic characteristic of the speaker’s.⁶ Since credibility is a distinctively epistemic notion that for Fricker is centrally involved in the transmission of knowledge by testimony, the injustice that Alex commits by having their judgment of Kelly’s credibility be influenced by bias is testimonial and epistemic in nature. Relatedly, the norm it violates – to judge speakers’ credibility in an unbiased manner – has a clear epistemic dimension. The objection that the case involves a non-epistemic harm would only be plausible if other, non-epistemic aspects – e.g. Kelly’s overall likeability – had been judged by Alex in a biased manner.⁷

The third objection requires more thorough discussion:⁸ Fricker’s overall account is that many of us harbour prejudices of one sort or another. Fricker maintains that TI can nonetheless be avoided if we learn to exercise testimonial virtue, which neutralises the effects of these prejudices, effectively transforming what would have been an unduly deflated credibility judgment into what is in fact a correct credibility judgment (Fricker 2007, 86–108).⁹ So we have from Fricker a paradigmatic model of how TI is avoided that has at its heart a negative prejudice exerting downward pressure on the hearer’s credibility judgment, which is neutralised by a compensatory force of equal measure. One might be tempted, then, to think of Countervailing Biases along similar lines: as a case in which TI is averted, where the compensatory force that neutralises the negative prejudice is a kind of luck, rather than testimonial virtue. (This diagnosis might be accompanied by the judgment that Alex does not deserve any moral credit or praise for averting TI, precisely because TI is avoided through luck rather than testimonial virtue.) If this alternative description of Countervailing Biases is plausible, then the argument that Countervailing Biases is an instance of TI does not succeed.

But this description is not plausible, for the simple reason that the compensatory force that neutralises the downward pressure on credibility is not luck, but rather Alex’s prejudice in favour of upper-class people. In other words, the compensatory force is a moral flaw of the hearer, not a feature external to the hearer suitably described as ‘luck’.

This independently plausible point finds support from Fricker, who explicitly contrasts prejudice and luck. This contrast appears in Fricker’s distinction between genuine cases of TI and so-called ‘bad luck’ cases, which for Fricker do not count as TI. In both TI and ‘bad luck’ cases, the hearer deflates the credibility of the speaker. The difference for Fricker is that in ‘bad luck’ cases, the deflation does not occur through any ethical fault on the hearer’s part, but through ‘ethically innocent but epistemically culpable error’ (Fricker 2007, 22).

For example, Fricker considers a case in which someone conducts a hasty web search that yields the incorrect result that a particular academic is a medical scientist, instead of the correct result that they are a philosopher specialising in medical ethics. When this academic makes a forthright statement about moral fictionalism, the hearer, believing them to be a medical scientist, gives less credibility to the academic than is in fact warranted by their expertise. About this hearer, Fricker writes (2007, 23):

While her error is epistemically culpable, its ethical non‐culpability still seems to prevent the resultant credibility deficit from constituting a testimonial injustice: an ethically non‐culpable mistake cannot undermine or otherwise wrong the speaker. It seems that the ethical poison of testimonial injustice must derive from some ethical poison in the judgement of the hearer, and there is none such wherever the hearer’s error is ethically non‐culpable.

Fricker is here endorsing the plausible contrast between prejudice and luck. When the former generates a credibility deflation, we have TI; when mere bad luck (‘an ethically non-culpable mistake’) generates a credibility deflation, we have no TI.

In Countervailing Biases, unlike in ‘bad luck’ cases, the factor that counterbalances the deflationary effect of Kelly’s negative prejudice against Black people is the prejudice in favour of upper-class people – it is not luck. And prejudices clearly count as the ‘ethical poison’ at the root of TI.¹⁰ Granted, its effect on the speaker’s credibility is positive rather than negative, but it would be implausible and highly ad hoc for Fricker to think both that (i) when they deflate credibility, prejudices do not count as mere luck because they are moral flaws, but that (ii) when they upgrade credibility, they do count as mere luck despite being moral flaws. Luck and prejudice in the present context, then, should be kept distinct regardless of whether the luck is good (credibility-restoring) or bad (credibility-decreasing).

In addition to the independently plausible point that the credibility-upgrading force in Countervailing Biases is prejudice and not mere luck, then, we have textual evidence that Fricker herself would understand the case in this way. Since the compensatory force that neutralises the credibility-downgrading effect of the prejudice against Black people is itself a prejudice, we have in Countervailing Biases not one, but two doses of the ‘ethical poison’ at the root of TI, and thus no good reason to deny, on the basis of any considerations based on luck, that TI takes place.

4. Countervailing Biases and Extant Cases of Due-Credibility TI

Having argued that Countervailing Biases should be accepted by Fricker as a case of due-credibility TI, in this section I explain how this case differs from extant cases discussed in the literature, namely Lackey’s and those I previously proposed. I should emphasise that the claim that Countervailing Biases is a case of due-credibility doesn’t strictly depend on accepting that the putative cases proposed by Lackey and myself should also be countenanced by Fricker as cases of TI; even though all cases argue for the same broad claim that Fricker should accept due-credibility TI, they do so in different ways, as we will see. Appreciating the differences between Countervailing Biases and extant cases is useful to gain a more precise taxonomy of due-credibility TI. My discussion also serves to highlight an error in the lesson I previously drew from my original cases regarding how Fricker could modify her account to exclude due-credibility TI.

4.1. Lackey’s Case of Due-Credibility TI

Fricker understands credibility as a non-distributive good: giving more to someone does not mean giving less to someone else. On this view, credibility is not finite, and everyone can in principle receive due credibility. Against this view, Lackey (2018) argues extensively that in many significant circumstances credibility is distributive: in a court case where the prosecution claims p and the defendant claims not-p, the credibility granted to one party will necessarily affect the credibility granted to the other. For our purposes, the important point that Lackey makes is that TI is not necessarily a case of granting less-than-due credibility in an absolute sense but can arise through granting relatively less credibility to one group than to another.

Lackey’s relevant case (2018, 149) involves a research team in which women researchers are given due credibility, but where the individual men researchers give themselves and other men more credibility than they are due. In these circumstances, the due credibility given to women does not immunize them from TI, since a woman who disagrees with a man sees her view unjustly dismissed by the men in virtue of the relative difference in credibility that the men afford to women and to themselves. Thus, she suffers TI at the hands of the men despite receiving due credibility.

Countervailing Biases differs from Lackey’s case of due-credibility in three respects. Firstly, we need not posit more than one bias in Lackey’s type of case: a single prejudice harboured by men, which conceives of women as intellectually inferior, can drive the case; secondly, while in cases of disagreement the men grant women due credibility, they don’t grant women knowledge: women’s views are dismissed as inaccurate, because they conflict with the view of a man; thirdly, and relatedly, the TI is detectable: women who suffer this kind of TI will notice the injustice and consequently feel legitimately slighted by the evident dismissal of their views.

By contrast, as we have seen, Countervailing Biases requires two prejudices to neutralise each other; the hearer grants the speaker knowledge, and this makes the kind of due-credibility TI undetectable. So Countervailing Biases differs from Lackey’s case in crucial ways.

4.2. My Previous Cases of Due-Credibility TI

In Luzzi (2016), I observed that a hearer may take towards a speaker the following stance: they may believe that what the speaker says is true, and thus accept the relevant belief into their body of beliefs, yet deny that the speaker knows what they assert. For example, the hearer may think that the speaker has happened to latch on to a true belief, but that the speaker lacks good reason to believe the asserted proposition or that the speaker lacks a proper grasp of the concepts it involves, and so does not know it. If this attitude is held for prejudicial reasons, I argued, then Fricker should countenance this as a case TI even though the speaker is granted due credibility. Recall that for Fricker, granting due credibility means that the hearer’s judgment matches the evidence that what the speaker says is true, so in such a case Fricker’s understanding of the hearer’s obligation is satisfied, but the crucial feature of my prior cases is that the hearer fails to have their credibility judgment match the evidence that the speaker knows the content of their assertion. Intuitively, this latter failure is epistemically disrespectful enough to trigger TI, and thus, I claimed, shows that Fricker’s understanding of the hearer’s ethic-epistemic obligation towards the speaker is too narrow.

I provided a rather abstract example, with respect to a group G that is the object of this prejudicial attitude and the non-Gs who hold this attitude towards Gs:

[N]on-Gs regard the beliefs held by Gs as true but not doxastically justified. Perhaps, for instance, non-Gs believe that Gs have generally truth-conducive belief-forming mechanisms, but also believe that once Gs form their beliefs, a cognitive malfunction leads Gs to sustain their beliefs with completely irrelevant considerations, so that Gs’ beliefs are true but never appropriately epistemically based. In other words, non-Gs falsely and dogmatically believe that the reasons Gs have in support of their (i.e., Gs’) beliefs are inadequate, and truly believe that Gs’ beliefs are true. In this way, non-Gs believe that Gs don’t know the propositions they assert, even though they (i.e., non-Gs) routinely accept them (quite correctly, in fact) as true. (207)

A concrete example of this kind of case might be the attitude historically taken by many men who considered women to be intellectually inferior, particularly in intellectual domains such as scientific ones. On occasion, a man who held this attitude might have accepted that a woman was capable enough to have succeeded in picking up a true belief in the relevant domain but not capable enough to have good reasons for her belief or proper understanding of the concepts involved. He thus would not have taken her to know what she asserted, though he would have taken her to speak truly. Another concrete example is the phenomenon of `mansplaining’, where a man explains to a woman an assertion she has made, prejudicially taking her not to have proper knowledge or adequate understanding.¹¹ Men who hold such attitudes understand women merely as potential conduits for knowledge rather than as holders of knowledge.

Countervailing Biases differs from my prior cases in important respects. First, those cases need not be driven by multiple prejudices. Secondly, those cases necessarily involve the hearer failing to grant knowledge to the speaker, while in Countervailing Biases such knowledge is ascribed. These differences suffice to show that the cases are of different kinds. There is a third potential difference to do with detectability, although this will depend on how some of the details are spelled out. While in Countervailing Biases the TI is not detectable, in my prior cases of due-credibility TI it is possible for the hearer to deem the speaker’s belief to be true-but-not-known in a way that does not suggest to the speaker that they are being dismissed, but it is also possible for this epistemic disregard to be evident. So this third point of difference may be less clear. Nonetheless, the first two differences noted suffice to show that Countervailing Biases is of a distinct nature.

The second difference I mentioned – which depends on whether the hearer attributes knowledge to the speaker – turns out to be theoretically important. To see why, note that on the basis of my previous cases I proposed a modification on Fricker’s behalf aimed at ruling out the possibility of due-credibility TI. Recall that Fricker’s conception of due credibility requires that the hearer respect the evidence that the speaker’s assertion is true. My previous cases can be seen to arise because one can fulfil this obligation while prejudicially denying that the speaker speaks from knowledge. This denial is a form of objectification and so seems to constitute TI.

The theoretical amendment that I considered on Fricker’s behalf, then, was to ensure that credibility is understood fundamentally in terms of knowledge (2016, 209–210). On this revised account, to grant due credibility to a speaker is not merely to have one’s judgment match the evidence that what the speaker says is true, but, more strongly, to have one’s judgment match the evidence that the speaker knows what they assert. Once credibility is understood in this more demanding form, my prior cases clearly no longer count as TI-with-due-credibility. This is because when the hearer prejudicially deems the speaker to have beliefs that are true-but-not-justified, they are prejudicially deeming the speaker not to know what they assert. Thus, they are not granting the speaker due credibility in this enhanced sense, for the prejudice causes the hearer to fail to align their judgment that the speaker knows the content of their assertion with the evidence that the speaker indeed knows what they assert. This strengthening of the notion of credibility in terms of knowledge accommodates my prior cases. I suggested on this basis that it allows Fricker to retain the key claim that TI requires credibility deficit, where credibility is understood in terms of appropriate knowledge ascription.

However, Countervailing Biases demonstrates that this amendment cannot work, for it is a case of due-credibility TI in which the hearer not only matches their credibility judgment to the evidence that what the speaker says is true, but they also match it to the evidence that the speaker speaks from knowledge. Unlike in my prior cases, then, in Countervailing Biases the hearer not only believes what the speaker says, but they also deem the speaker to know what they assert – a judgment that corresponds with what the evidence mandates. Thus, the upshot is that even if the amendment I previously suggested were accepted by Fricker, Fricker ought still to acknowledge TI with due credibility.¹²

5. The Significance of Countervailing Biases

Having argued that Countervailing Biases is a case of TI, let me explain why cases like it ought to be expected. Just like my previous cases exploited the epistemic gap between regarding someone as telling the truth and regarding someone as knowledgeable, Countervailing Biases exploits a specific epistemic gap: the gap between holding the belief warranted by the evidence and holding the belief warranted by the evidence through an appropriate weighing of that evidence. Both Alex (the perpetrator of TI) and Parker (the hearer who harbours neither of Alex’s two prejudices) wind up holding the correct credibility judgment about Kelly. But the genesis of this belief is relevantly different for the two of them: while Parker holds this belief through properly weighing the evidence, Alex allows non-evidential and morally problematic prejudices to interfere in the formation of their judgment. Thus their judgment, while corresponding to the one warranted by the evidence, is not appropriately based on that evidence.¹³ That the interfering biases at work in the genesis of Alex’s belief reciprocally neutralise does not detract from the point that the process by which Alex forms their judgment is neither epistemically nor morally appropriate. The moral badness of such behaviour can be supported by appeal to the morally bad negative identity prejudices that influence the formation of Alex’s judgment. We can also make salient the epistemic badness of Alex’s behaviour by observing that even though Alex’s prejudices have yielded a credibility judgment that corresponds in its content with an unbiased credibility judgment, in many other contexts (where the speaker is Black but not upper-class or where the speaker is upper-class but not Black) such prejudices will lead Alex to form credibility judgments that are epistemically inaccurate, and Alex will consequently unjustly deflate or inflate the speaker’s credibility. This result follows from the point made above, that in the formation of Alex’s judgment the evidence is not appropriately weighed, since its assessment is influenced by prejudice.

As noted above, Countervailing Biases is distinct from extant cases of due-credibility TI. It is also different from paradigmatic cases of TI, such as those regarding Tom Robinson and Marge Sherwood. What is the import of recognising this kind of case as TI?

Firstly, recognising Countervailing Biases as a case of TI means expanding the repertoire of cases of TI. Understanding accurately the various ways TI can manifest is of independent philosophical interest, given the substantial attention this concept has received from philosophers in recent years and continues to receive.

Secondly, doing so entails acknowledging that TI can be less detectable than we might assume. Recognising real-world instances of TI as TI is of course no easy business: even in many cases where it is clear that one’s epistemic contribution is not being considered seriously, it can be difficult to ascertain that this is due to prejudice rather than to other factors. (It is telling in this respect that Fricker’s central cases are drawn from fiction, where the fact that it is prejudice specifically that causes an undue dismissal can be stipulated.) However, cases like Countervailing Biases demonstrate that even when one’s epistemic contribution is taken seriously – when one is deemed by one’s interlocutors to be speaking truly and from knowledge – TI might still be at work. Thus, we gain from Countervailing Biases a distinctive reason not to rule out the possibility of TI even when knowledge is transmitted and the testimonial exchange therefore appears epistemically and ethically appropriate.

Nonetheless, two thoughts might drive scepticism about the significance of cases like Countervailing Biases: first, two prejudices exerting exactly the same opposite pressure on credibility judgments, in the way required by Countervailing Biases, is rare; and second, the harm accrued to the speaker is not particularly important, precisely because such cases necessarily involve the speaker not feeling epistemically diminished or dismissed and not feeling like any injustice has occurred.

To the former point, it’s worth bearing in mind Fricker’s observation noted in section 2 that granting the right amount of credibility to a speaker is not a matter that is amenable to precise codification and may admit of imprecision (2007, 75–76). If so, then the right amount of credibility may be afforded even in situations when oppositely valenced prejudices are roughly of equal strength. Moreover, cases of due-credibility of the same kind as Countervailing Biases can arise through the joint interaction of a greater number of prejudices, if the influences they exert on the hearer’s credibility judgment ultimately mutually neutralise. Given the multiple identities many people have, this is likely a more realistic and more frequently instantiated scenario than the artificiality of Countervailing Biases at first suggests.

To the latter point, the considerations I adduced in section 3 in defense of the claim that Countervailing Biases is a case in which Kelly suffers an injustice are relevant and demonstrate that just because one cannot detect an injustice does not mean that no injustice is done: ethical and epistemic disrespect can fly below the radar, so to speak. But granting, as I have argued Fricker should, that Countervailing Biases instantiates a genuine kind of TI also suggests a hypothesis for making more precise what the problematic feature of TI is. At first pass, Fricker deemed the problematic feature of TI to be its prejudice-driven deflation of credibility. However, in the light of arguments in Medina (2011, 2013), Davis (2016), Luzzi (2016), Lackey (2018), this conception seems insufficient to capture the full range of cases of TI, since such cases do not exhibit credibility deflation. As we have seen, my previous suggestion that the problematic feature of TI is that the speaker prejudicially fails to ascribe knowledge when they ought to is also insufficient, since in Countervailing Biases Alex ascribes Kelly knowledge, as they should. What, then, is the common denominator of all these cases of TI, the problematic feature shared by cases of deflated-credibility TI, excess-credibility TI and due-credibility TI?

Countervailing Biases suggests that merely to have credibility judgments about the speaker (or – in situations of `zero-sum credibility’ where the speaker’s and the interlocutor’s credibility are pitted against one another – about one’s interlocutor) be influenced in their genesis by prejudicial considerations suffices to trigger TI, even if this influence does not ultimately lead to credibility deflation. All Fricker’s critics who take issue with her deflated-credibility requirement present cases that exhibit this feature: in Lackey’s case of the research team, the excessive credibility prejudicially granted to the men in this zero-sum credibility situation triggers TI against the women; in my previous cases, the hearer prejudicially assesses the speaker’s credibility, even if due credibility is granted. In Medina’s cases, the prejudicially inflated credibility of dominant groups triggers TI against disadvantaged groups; in Davis’s cases, where the speaker is compelled against their wishes to speak for their social group, their credibility is inflated owing to the hearer’s prejudicial belief that all members of that group share the same view.¹⁴

In conclusion, Countervailing Biases instantiates a novel kind of case that puts independent pressure on Fricker’s thesis that TI requires credibility deflation; it highlights a new way in which TI involves treating the speaker unjustly, despite the injustice being difficult to detect; it helps provide a more accurate understanding of the full taxonomy of cases of TI – a project of independent philosophical interest, and it suggests a novel account of the problematic feature of cases of TI.¹⁵

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Federico Luzzi

Federico Luzzi is Senior Lecturer in Philosophy at the University of Aberdeen. His research interests include epistemology, as well as feminist issues in philosophy of sport and ethics.

Notes

1. Fricker has further defended and discussed her account of TI in Fricker (2010, 2013, 2015, 2016, 2017).

2. Thanks to an anonymous reviewer here.

3. The case that ascriptions of credibility excess are morally problematic is well-made by Medina (2011, 2013) who argues that these excesses are usually ascribed to members of socially advantaged groups and are typically accompanied by corresponding deficits towards members of disadvantaged social groups. Thus, they constitute unfair patterns of credibility distribution and serve to reinforce oppression. See also Davis (2016) for further argument and examples of how positive prejudices can underpin TI.

4. Fricker’s conception of the harm of TI as denying subjectivity has been criticised on the grounds that it cannot handle cases in which the speaker is prejudicially deemed to be insincere, such as the central case of Tom Robinson: in such cases, it seems the hearer must grant substantial epistemic agency to the speaker, since taking the speaker to be insincere involves attributing to them a view of the relevant facts as well as the ability to fabricate and communicate an in-principle credible story (Pohlhaus 2014, Congdon 2017, McGlynn 2020, 2021). However, Countervailing Biases trades on the speaker’s competence, not their sincerity, and Fricker’s notion of objectification remains plausible for cases that trade on competence. Additionally, on a promising alternative proposal to Fricker’s that better handles cases of sincerity, Alex harms Kelly in the relevant way: McGlynn argues plausibly that objectification for TI is a broad notion (cf. Nussbaum 1995) that can be satisfied in many ways, including treating the speaker as fungible ((McGlynn 2020, 2021), cf. Davis (2016)). In relying on their prejudice, Alex treats Kelly in the relevant sense as interchangeable with other members of the social groups that are salient to the two prejudices. Therefore, despite the prima facie concerns one may nurture regarding Fricker’s understanding of objectification, it remains nonetheless plausible that Alex objectifies Kelly and that they commit TI.

5. See e.g. Coates (2014).

6. Note in this respect that Fricker (2013, 1319) explicitly allows that in some cases of TI, namely when the credibility deficit is sufficiently small, transmission of knowledge from speaker to hearer will nonetheless take place. So from the point that knowledge has successfully been transmitted through testimony we cannot infer that no TI has occurred.

7. A referee asks whether the injustice might be considered epistemic but non-testimonial. This seems implausible for the reasons just mentioned. Consider, moreover, that in all respects other than the presence of a countervailing bias, Countervailing Biases is akin to ‘standard’ cases of TI, such as Tom Robinson’s and Marge Sherwood’s.

8. I owe this objection to an anonymous reviewer.

9. For criticism of testimonial-justice virtue correcting via reinflation, see Alfano (2015), Sherman (2016a, 2016b), Washington (2016).

10. One might deny that the prejudice is problematic because it is positive. But again, see fn 3 above and Medina (2011, 2013) and Davis (2016) for compelling reasons why positive prejudices in favour of dominant groups are morally problematic and can underpin testimonial injustice.

11. Credit goes to Bernhard Salow for this example.

12. In passing, Marušić and White (2018, 106) briefly describe (but do not discuss in any detail) an example in which a hearer would deflate the speaker’s credibility on account of the prejudicial belief that the speaker’s ethnic group is usually insincere, yet they grant the speaker due credibility, since they have reason to think that in this particular case it is in the speaker’s interest to tell the truth. This might also count as a case of TI with due credibility. Countervailing Biases differs from this case because in the latter case the potential effects of the negative bias are neutralised by a second problematic prejudice, rather than by background information. So we lose the ‘two wrongs don’t make a right’ reply to critics who would not deem it a case of TI, which instead can be used to buttress the case for deeming Countervailing Biases a case of TI in Fricker’s framework.

13. One consequence of this is that Parker knows that Kelly is credible, but Alex plausibly does not know this. A possible remedy for Fricker’s account might thus be to require of TI that the hearer’s credibility judgment itself amount to knowledge. However, I do not wish to commit to this account, for it depends on the claim that in all cases in which one’s belief-formation is influenced by bias, knowledge is compromised, and this doesn’t seem like an exceptionless claim.

14. The case briefly described by Marušić and White (2018) and described in footnote 12 also exhibits this feature: the hearer’s credibility judgment about the speaker is affected in its genesis by their prejudicial belief that the speaker might be lying.

15. Thanks to Paula Sweeney for helpful comments. I am also grateful to three anonymous referees, including two for this journal.