Introduction
In How Stereotypes Deceive Us, Katherine Puddifoot has produced a convincing non-normative account of what stereotypes are, and of the complex relationship between epistemic and ethical goals, and the conditions under which we appropriately rely on stereotypes in achieving these ends. I found myself agreeing with much of the book. What follows are therefore not substantial worries for Puddifoot’s account, but rather just a few points to consider with regard to how Puddifoot’s analysis and cases fit with other cases in the literature on epistemic ethics, specifically, with cases of testimonial injustice. I hope this will serve to build on the invaluable work that Puddifoot has contributed to our understanding, and perhaps offer some alternative ways to think about cases of testimonial injustice more generally.
Testimonial injustice, reliability and faulty stereotypes
A person who is marginalised in a society can fail to be treated fairly as an epistemic agent, through the unjust epistemic behaviours of those who she interacts with. One way in which this can happen is, as Puddifoot identifies, through testimonial injustice (Fricker 2007). Someone who experiences testimonial injustice is “wronged specifically in her capacity as a knower” (Fricker 2007, 20) when she attempts to take part in a testimonial exchange, or when she offers up her own knowledge. In central cases, on Fricker’s account, the hearer fails to take up, believe, or otherwise appropriately epistemically process the testimony of the speaker, when she otherwise ought to, because of her (implicit or otherwise) acceptance of an identity-prejudicial stereotype which “embodies a prejudice that works against the speaker” (Fricker 2007, 17). The speaker’s testimony is treated as less credible than it ought to be by the hearer. The hearer may also distort the speaker’s meaning, attach unintended meaning to the speaker’s words, or even attach meaning to the fact of that person speaking at all, which is not justified by the actual content or context of the testimony provided.
Fricker provides a case from literary fiction (the screenplay of “The Talented Mr Ripley”), which she takes to be a paradigm example of systematic, prejudicial identity-based testimonial injustice. A woman (Marge Sherwood) has her credible—and importantly valuable—evidence downplayed as “feminine intuition”, by a male acquaintance (Herbert Greenleaf), who otherwise holds her in warm personal regard (Fricker 2007, 14). Sherwood experiences a kind of subconscious downgrading of her testimony, and her attempts to provide it to Greenleaf, due to prejudices held towards people with her social identity as a woman within that society, at that point in time. This can be observed in the following interaction, suggests Fricker:
GREENLEAF, speaking to Tom Ridley: “This theory, the letter he left for you, the police think that’s a clear indication he was planning on doing something (…) to himself.”
SHERWOOD: “I just don’t believe that!”
GREENLEAF: “You don’t want to, dear. I’d like to talk to Tom alone—perhaps this afternoon? Would you mind? Marge, what a man may say to his sweetheart and what he’ll admit to another fellow-”
(Minghella, quoted in Fricker 2007, 87).
Despite his desperate need for information relating to his missing son, Dickie, Greenleaf “fails to see Marge as the source of knowledge about Dickie that she manifestly is” (Fricker 2007, 88). Greenleaf’s uptake of negatively-valenced prejudicial stereotypes about the rational capabilities of women cause him to downgrade Sherwood’s testimony to that of a hysterical woman, too delicate to understand her fiancé’s infidelities and mind, whose testimony can be dismissed as that of someone who simply wants to believe the best of her partner. This is summed up in the line: “Marge, there’s female intuition, and then there are facts” (Minghella, quoted in Fricker 2007, 14). In the novel and screenplay, however, we discover that Sherwood is correct and that Ripley has murdered Marge’s fiancé.
There are a number of key features in this account of paradigm cases of testimonial injustice:1
(i) The speaker is a member of a group about which there are widely held stereotypes, associations between that group and particular attributes (Fricker 2007, 30), in her society.
(ii) These stereotypes are negatively valenced: they ascribe traits to the group which are derogatory within the specific context.
(iii) These stereotypes “embody” prejudice: they are resistant “to counter-evidence owing to some affective investment on the part of the subject” (Fricker 2007, 35, author’s italics).
(iv) The hearer is influenced, knowingly or otherwise, by these stereotypes.2
(v) This leads to the hearer’s epistemically culpable mistreatment of the speaker’s testimony in a specific interaction.3
All of this suggests that there is something wrong with the stereotypes in play in cases of testimonial injustice. The reason that these stereotypes ought not to influence the hearer’s assessment of the speaker is that they are faulty or inaccurate; they are prejudicial and “resistant to evidence”. The use of these faulty stereotypes therefore makes the hearer epistemically and ethically culpable for their failure to treat the speaker appropriately in a testimonial interaction.4
The problem of accurate stereotypes
An important insight
However, Puddifoot draws our attention to very important cases which appear to mirror the epistemic and ethical harms of testimonial injustice, despite the hearer acting in ways which do not, on the face of it, appear to be epistemically or ethically suspect in the way that paradigm cases of testimonial injustice suggest. These are cases in which accurate stereotypes are in play but someone is still treated unjustly in terms of their testimony.
For example, it may be the case that in wider society, women are less likely to have scientific expertise than men are. This may be due to any number of reasons, including a lack of access to opportunities, or the discriminatory treatment of women in educational settings within that society. Nevertheless, the stereotype that “women are less likely to have scientific expertise than men”, is accurate for that society. When a woman who works in STEM then attempts to provide her expert insight, her credibility might be unfairly downgraded because of the holding and application of this stereotype by her audience. Viewed as a woman, and therefore as someone who is less likely to have scientific expertise than her male colleagues, she might be ignored and overlooked in discussions with her colleagues, or in providing information to a wider public.
In this case, the stereotype that “women are less likely to have scientific expertise than men” is one which (a) accurately “reflects reality”, and (b) appears to have been gained in an epistemically appropriate way, given that it is responsive to evidence about the number of women vs men who have scientific expertise: for the audience, the stereotype is “a fitting response to facts found in their social environment” (Puddifoot 2021, 68). Puddifoot therefore suggests, on my reading of her account, that the use of the stereotype does not embody prejudice of the kind found in Frickerian testimonial injustice (or perhaps that the use of the stereotype is not prejudicially motivated).
Puddifoot’s key claim then, is that the application of a stereotype which accurately reflects “some aspect of social reality” “can (…) lead to the undue dismissal of the testimony of members of some social groups” (Puddifoot 2021, 190). The non-prejudicial use of accurate stereotypes mean that these types do not fit well under Fricker’s account of paradigm cases of testimonial injustice, in which our prejudicial attitudes are at the root of the wrongdoing, and in which we can be held epistemically culpable and blameworthy for the outcomes because of this. The speaker experiences the epistemic harm of not being treated fairly as a knower (with the risk of subsequent harms to her career, and perhaps even to how she herself views her own epistemic capabilities, with further knock-on effects to her attempts to contribute to science). But despite mirroring the epistemic harms found in Frickerian cases (and perhaps the sense that something has still gone “wrong” in how the individual has been epistemically treated), prejudice does not seem to be at the root of the problem.
Puddifoot’s identification of cases such as this is a hugely important contribution to understanding the landscape of epistemic harm. It seems vital, for anyone concerned with epistemic justice or the avoidance of epistemic harm in structural oppression, that we recognise that in many ordinary cases of testimonial interaction the mechanism by which our epistemic judgements can cause harm is not immediately obvious through the presence of false (and either prejudicial or prejudicially motivated) identity-based beliefs.
A worry for ethical assessment
One upshot of identifying these types of cases, however, is that it becomes harder to recognise whether the use of stereotypes as heuristics will lead, or have led, to epistemic harm in our testimonial interactions. This creates a challenge for evaluating the ethical behaviour of hearers who rely on (accurate) stereotypes in situations which result in epistemic harm for the speaker. If it is indeed the case that accurate and fairly formed stereotypes can cause us to epistemically harm speakers on some occasions, while not doing so on others, it becomes difficult to argue that an ethical harm has been perpetrated by the hearer, despite there being someone who appears to be on the receiving end of identity-based, systematic mistreatment of her testimony.
Puddifoot proposes that we rely on evaluative dispositionalism to determine whether or not a person ought to adopt and use a heuristic or not. This is a process in which we need to understand and evaluate the complicated and wide-ranging dispositions of the actors involved taken as a whole, in order to determine whether a person has acted, or will act, rightly or wrongly in relying on a particular stereotype in any given context or situation.
There may be worries around how demanding this process would be. However, a more pressing worry for many might be that, even when applied well, this process will lead to counter-intuitive results. Specifically: that the weighing up of the balance of all possible upstream and downstream dispositions of the speaker—the dispositions displayed in forming beliefs, and the dispositions possessed as a result of holding those beliefs—might in some instances suggest that the hearer did act in a reasonable or fair way in adopting and relying on a stereotype, even though the speaker experienced epistemic harm in her testimony not being taken up when it was in fact credible.
And if this is the case, then this would leave no real recourse for identifying an ethical, epistemically culpable, wrongdoing or wrongdoer, despite the presence of an epistemically unfair outcome for the speaker. This might appear to let those who downgrade credibility in this way “off the hook” for what, intuitively, still seems in some way to be bad behaviour, in and of itself. Or at the very least, to suggest that there is no ethical injustice for someone who experiences testimonial credibility downgrading under these circumstances.
We could, of course, bite the bullet on this, and accept that there was no ethical wrong. Alternatively, it does in principle remain possible that an exhaustive and detailed analysis of the dispositions involved in every case we can think of might in fact turn up that there are not cases in which there is no residual ethical wrong in the adoption and application of an accurate stereotype in scenarios which lead to testimonial harm.
However, biting the bullet requires giving up an important ethical stance many might want to take, and completing a detailed analysis of every possible case that initially seems problematic is something that arguably falls beyond the bounds of a reasonable expectation for our ethical judgements (and interpretation of right action for ourselves and others in practice).
I wonder, therefore, whether the worry around ethical culpability for epistemic harms in some of the cases Puddifoot identifies might justify revisiting whether or not these types of cases are distinct from “ordinary” testimonial injustice—in which we can identify a wrongdoer and wrong action—after all.
A first thought in this direction might question whether or not it is right to say that Puddifoot-style cases do not, in fact, involve the operation of prejudice, construed in some suitable manner. If this is the case, then this alone would be enough to categorise them as “ordinary” testimonial injustices involving clear wrong action (albeit not necessarily ones that Fricker’s original 2007 account accommodates).
A second thought relates to the question of what makes a stereotype accurate or otherwise in the first place. Is it in fact the case that a stereotype reflecting “some aspect of social reality” and being mis-applied in a context, is in itself enough to demonstrate that testimonial injustice comes about as a result of an accurate, yet irrelevant, stereotype? It is the fact that these stereotypes appear to be accurate that creates the challenge of the missing ethically wrong action. If the stereotypes turn out to not reasonably be considered accurate after all then, again, this problem goes away, and these apparently different cases may not actually differ from what we might think of as central cases of Frickerian testimonial injustice, where the wrongdoing is connected to the use of faulty stereotypes, ones which are resistant to counter-evidence that the hearer is epistemically responsible for not recognising.
I expand on these thoughts in turn in sections 4 and 5 below.
Prejudice: “Resistance to counter evidence” in stereotype content vs context
When credibility deficits of this type occur, there will at least sometimes be an absence of prejudice, so the effect is not a case of testimonial injustice, as defined by Fricker, because this occurs due to the operation of prejudice. (Puddifoot 2021, 190, my italics)
Testimonial injustice relies on the use of stereotypes which embody prejudice. As such, the first thing we might question is whether, in cases such as the woman in STEM who is not treated with due credibility by her colleagues or the public, there really is no prejudicial content either in the stereotype, or its use.
Applying a stereotype inappropriately without appropriate contextual boundaries, as a result of some “ethically bad affective investment” on the part of the hearer (Fricker 2007, 35), involves prejudicial treatment. For Fricker, it seems that the specific stereotype must itself embody a negative association with members of the group it targets, which is resistant to counter-evidence (ibid., 35). As such, for any case of testimonial harm in which the stereotype is accurate, given the available evidence, and in which the individual hearer would update their stereotype should the underlying facts change, it seems prejudice has not played a part in the hearer’s actions, where their downgrading of a speaker’s credibility is due to their belief in an accurate stereotype.
However, it is not clear to me that testimonial injustice ought to only involve instances in which the resistance to counter evidence is within the stereotype itself, rather than within the context in which one uses the stereotype. Prejudice can enter into our actions at alternative points, related to the use of the stereotype within a given context, rather than the content of the stereotype itself.
Puddifoot highlights that stereotypes are (regularly) applied in contexts to which they are not relevant. As she notes: “A stereotype can reflect some aspect of social reality but influence judgements about other aspects of social reality” (Puddifoot 2021, 62). In this situation, a person acquires a stereotype which is responsive to evidence in their wider environment. However, they go on to act in ways which generate testimonial harm through applying that stereotype in irrelevant contexts. In the case of women in STEM being downgraded in testimonial interactions with their colleagues:
Women who enter universities and workplaces to work in STEM subjects are stereotyped as lacking expertise. But the stereotype is not relevant in these cases, in which women have (…) gained scientific expertise. (Puddifoot 2021, 63)
In this case we might say then that someone has made a prejudgement without proper regard to the evidence about a more relevant group: that of “women who work in STEM”.
The question we should ask is whether this prejudgement itself was due to prejudice. If prejudice is resistant to counter-evidence, due to “affective investments”, then in the case of women in STEM, it may well be the case that while the initial stereotype is accurate, the failure to consider further relevant readily-available contextual evidence (that the specific woman is a STEM professional and must therefore have scientific expertise), comes about due to the same prejudicial views about women that make it the case that in the society one lives in, fewer women than men have received a STEM education or range of career opportunities. The very fact that the stereotype is true is indicative of prejudice in the background conditions of society. The willingness to ignore further evidence (including evidence that the stereotype may not apply in this case) in the presence of that stereotype might very likely be due to the same prejudice towards that group of persons.
Consider the following two scenarios by way of illustration of the difference between prejudicial and non-prejudicial use of accurate stereotypes:
CROWDSOURCING
Jonno the journalist needs to quickly get a few fairly innocuous anonymous reactions, regarding Manchester City’s victory at the Etihad Stadium earlier that day, to include in a quick summary in a local news report going out live in fifteen minute’s time. There is a good chance that many City fans are still in the city centre, celebrating the win. Jonno has limited time so, relying on the stereotype that “men are more likely than women to attend football matches”, they approach a group of men sat outside the pub, rather than a group of women who are sat outside a different pub over the road, to try and find a few supporters to speak to. Had a woman approached Jonno proactively, they would have happily asked her for a quote. But given time, Jonno’s strategy is to just maximise their chance of getting a few good soundbites in hand in time for the live transmission.
In this case:
(i) The stereotype “men are more likely than women to attend football matches” was accurate.
(ii) The stereotype “men are more likely than women to attend football matches” was relevant to the scenario (trying to maximise the chance of a good quote in a timely fashion in Manchester centre).
(iii) No epistemic injustice occurred5 because (a) nobody was wronged in their capacity as a knower; (b) there was no epistemic cost to Jonno or their audience in Jonno’s focus on men; (c) there was no prejudicial aspect to Jonno’s behaviour (they would have responded quickly to counter evidence from any woman who spoke to them).
POST-MATCH PUB CHAT
Elsie, a City fan wearing a City shirt, is sat in the pub after attending that same City match earlier that day. The rest of the crowd around her are men. Everyone is joining in with a large discussion, dissecting the match, the strengths of players on both sides, and the actions of the referee, in great detail. Elsie tries to explain why she thought the ref made a bad call early in the game, but when she speaks, she is ignored. A few minutes later, a man in the pub makes the same point and everyone joins in with their views in response to him.
In this second case:
(i) The stereotype “men are more likely than women to attend football matches” was accurate.
(ii) However, the stereotype “men are more likely than women to attend football matches” was irrelevant in the given context of listening to Elsie. Additional content—that the specific woman present, Elsie, was known to have been at the football match—over-ruled any reasonable exclusion of Elsie from the discussion about what happened in the game on the grounds of rules about the general population.
(iii) Epistemic harm occurred because: (a) Elsie was wronged in her capacity as a knower; and (b) there was an epistemic cost to the group who failed to hear Elsie’s insights.
Was there also (c), a prejudicial aspect to the group’s behaviour, which would make this testimonial injustice?
Arguably, yes. The prejudice may have entered in through background prejudicial beliefs and acts which lead to fewer women watching football matches in the first place, and the other members of the group having poor responsiveness to evidence due to holding those prejudicial views about women. All of this together was likely to have led to the downgrading of Elsie’s testimony, in the presence of the accurate stereotype about the general population.
There is no easy way to prove that this is the case. And there may be cases where it is not true. But we might say that where an accurate negatively valenced stereotype is true in virtue of conditions of prejudice within the broader environment and the acts this leads to, it is very likely that one cannot have escaped the prejudicial beliefs that led to that stereotype being true. As such, one is not ethically non-culpable for epistemic harm in a testimonial interaction with someone whose credibility is assessed based on that stereotype. This would not entirely correspond with Fricker’s use of the concept but it does, I think, represent a reasonable adjustment to it which takes into account the important sub-genre of cases Puddifoot has identified, and which accords with the spirit of what testimonial injustice is intended to capture: prejudicial treatment through use of stereotypes.
The role of accurate stereotypes in ordinary testimonial injustice
That said, I think that there may be something more fundamental about the role of context when considering any special considerations we ought to make for cases of testimonial epistemic harm, and the use of accurate stereotypes. And this is the question around whether or not the stereotypes being invoked really are accurate, given the context of application.
To examine this, first consider the stereotypes at play in paradigm cases of testimonial injustice. What stereotypes was Greenleaf drawing on, in unjustly failing to give Sherwood her epistemic due, in Fricker’s own example? Candidates, with their accuracy, include:
That women are not capable, or are less capable than men are, of rational thought (may or may not be accurate for a given interpretation of rational—if, for example, rationality is understood in a way that requires a particular kind of thought that women are less likely to be trained in, then women may count as less rational on that account).
In the Sherwood case there would also have been, given the time and location of the setting, many other beliefs and biases lurking in the background, which were not stereotypes about a given group (women), but which might come about due to stereotypes about a given group: that intuition plays no part in rational thought; that a person who is emotionally invested cannot be rational; that rationality of a certain kind is actually needed to understand the character of someone who has gone missing and to therefore be of use in finding out what has happened to him, etc. These are all aspects of the social reality of the time period and place.
My point here is not to say that in this paradigm case from literary fiction, or in parallel ones from real life dismissal of women’s testimony, testimonial injustice is in fact based on accurate stereotypes. Rather, that there are a number of readings where the stereotype being held by Greenleaf could be one that reflects some aspect of social reality, given the society and time in which the story is set and the characters are speaking. Nevertheless, somewhere along the line, an inaccurate stereotype came into play, for that specific context, in a way which made Greenleaf irresponsive to evidence in a given testimonial interaction, and it is this which arguably drives the poor epistemic behaviour and leads to this being a case of testimonial injustice. And this may not be dissimilar to what happens in the type of cases Puddifoot introduces.
Stereotypes are only accurate with regard to a given population. They are only reliably accurate insofar as they are applied to that same population. They are non-prejudicially applied within a given context only when they are responsive to evidence in egalitarian ways. In the case of the woman in STEM, those around her who don’t take her scientific expertise seriously, and who downgrade her input, are not merely relying on accurate stereotypes. They might be understood as inferring beyond what the empirical evidence supports, based on any number of further background beliefs and assumptions, and therefore relying on an (implicit or tacit) inaccurate stereotype.
One way to understand this is to attempt to make visible the hidden sub-clauses which are unstated but required for a stereotype to be genuinely accurate. The stereotype:
“women lack scientific expertise”,
is, for example, only accurate within a given context if we add sub-clauses:
“women (taken as a whole in the base rate population of this particular society) lack scientific expertise (because fewer women than men receive scientific training)”.6
While this is accurate, it can now clearly be seen that this stereotype is actually irrelevant in assessing the credibility of a person who does have scientific training. And so, the person using this stereotype is guilty of one of two things. Either they have failed to attach good reasons to their judgements (because the stereotype is irrelevant), in which case they are epistemically culpable for this in itself and likely to be driven by other motives (of prejudice or simple culpable irrationality) in making their assessment. Or, they have formed a judgement which is not based in the claimed stereotype at all, but in a more precisely expanded and inaccurate stereotype, such as that:
“women (who work in science) lack scientific expertise”.
If this is the case, then these instances more closely fit the standard model of testimonial injustice. And the problem of a lack of an ethically culpable wrong or wrongdoer in Puddifoot-style cases dissipates.
Conclusion
Katherine Puddifoot has drawn attention to the important fact that stereotypes which are accurate can be seen to feature in important cases of testimonial harm. This is an invaluable addition to the scope of what we need to consider when it comes to Fricker’s concept of epistemic injustice and to thinking about epistemic harm more broadly.
Further, it might be difficult, given that the stereotypes in these cases are not inaccurate and do not initially appear to involve prejudicial attitudes and act, to recognise epistemic harms in these cases as instances of injustice, for which hearers and audiences are culpable. This might be thought to leave a worrying responsibility gap in attributing responsibility which dispositional evaluation is not fully equipped to deal with.
In this short response I have, however, offered two ways in which these newly categorised Puddifoot-style cases might still be thought to be cases of testimonial injustice after all: (i) they may involve prejudice, although this may not be located where Fricker initially proposed in her account; or (ii) they may turn out to involve inaccurate stereotypes. If I am right in either case then it is possible that we can identify that in Puddifoot cases, there are ways in which persons are culpable for their use of stereotypes. As such, we can also conclude that an epistemic injustice, not merely unfortunate epistemic harm, has been enacted after all.
REFERENCES
Fricker, Miranda. 2007. Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.
Puddifoot, Katherine. 2021. How Stereotypes Deceive Us. Oxford: Oxford University Press.