Introduction
How Stereotypes DeceiveUs (2021) is a truly wonderful book. It argues for some rather startling conclusions, but it does so incredibly well and incredibly carefully. The startling claims come to seem not just inevitable but even commonsensical. This is a remarkable feat, and an indication of the book’s importance. It is the sort of book that really has the potential to change the way that people see some very important issues. What I will do here is to suggest that the book might have even more radical consequences than Katherine Puddifoot has herself argued for. First, I will rehearse some key arguments from the book, then I will suggest the possibility that they lead to even more startling conclusions than those which Puddifoot has already drawn out.
1. Key arguments from how stereotypes deceive us
I’ll start by laying out some of the startling, or at least surprising, claims made.
Social attitudes that do not reflect reality may sometimes be epistemically good.
Acts of truthful stereotyping may sometimes be epistemically bad.
These claims are, we will see, importantly interrelated: according to Puddifoot, it is because 2 is true that 1 is true.
As is obvious from these claims, Puddifoot rejects the idea of any summary judgement of stereotyping as always bad. She adopts a very broad and non-normative definition of “stereotype” and “stereotyping”. On her definition, stereotypes are “social attitudes that associate members of some social group more strongly than others with a certain trait or traits” (2021, 23). I will pause to note here that this is an extremely broad definition of stereotype, one that is in certain ways at odds with ordinary usage. (This is one of very few places where an opponent might find something to object to in Puddifoot’s arguments.) To see this, consider the group of people who teach at Durham University. This is clearly a social group. I associate this social group more strongly than certain others (e.g. people who teach at Southampton University) with teaching in the north of England. It seems strange to say that I stereotype people who teach at Durham University as teaching in the north of England. On Puddifoot’s view, however, this is precisely what I am doing. I don’t think that she needs to be troubled by this, though, as her chief goal is not one of perfectly capturing ordinary usage. But it seems perhaps worth noting that there is something here to think about.
How, then, does Puddifoot argue for her surprising conclusions? A crucial step here is Puddifoot’s view that there are many aspects we must consider when deciding whether an attitude is epistemically good or bad. (She calls her view Evaluative Dispositionalism.) Amongst these, crucially, are:
Does the stereotype dispose the person who applies it to respond in a way that is fitting with accurate statistical information (…). Does the application of the stereotype lead information about the specific case to be distorted or ignored? For instance does the application of the stereotype lead to: distorted remembering, the misinterpretation of ambiguous evidence, false assumptions about similarities/dissimilarities among groups and group members, aspects of the social identity of the person who is stereotyped being missed, testimonial silencing, testimonial injustice? (Puddifoot 2021, 56).
The next important step is realising that acts of truthful stereotyping can be epistemically bad. Take, for example, the belief that men are more likely than women to be scientists. This is true, and on Puddifoot’s view it is a stereotype. It has some obviously epistemic strengths: along with being true, it will allow one to make a reasonably good guess in the absence of other information about whether any randomly chosen scientist is a woman, for example. But it can also lead to errors, due to the power of the largely automatic associations that are likely to come with it. For example, someone with this belief who is evaluating a job candidate would likely be prone to:
(i) memory distortions that would make them selectively remember features of the candidate; (ii) viewing ambiguous behaviours of the candidate as evidence of lack of expertise; (iii) failing to notice differences between the candidate and other, previously encountered female scientists; (iv) failing to notice similarities between the candidate and male scientists; (v) the tendency to assume that any behaviours that are stereotypical of non-experts (…) are indicative of the dispositions of the candidate rather than the situation that she is placed in; and, finally, (vi) the tendency to make associations with a candidate that are inaccurate. (Ibid. 79-80)
This is, in brief, the case for accepting that some acts of truthful stereotyping can be epistemically bad—and it’s a convincing case.
The fact that truthful stereotyping can be epistemically bad is crucial to the argument that it may sometimes be epistemically good to have social attitudes that don’t reflect reality. A key case for Puddifoot is that of a person—call him Albert—who lacks the stereotype that men are more likely than women to have scientific expertise, and instead has the false belief that women and men are equally likely to have scientific expertise. This false belief, Puddifoot argues, brings significant epistemic benefits: Albert will respond to a woman candidate for a position in science in such a way as to avoid the many epistemic faults just outlined above. The case Puddifoot makes here is both startling and undeniable: it is clear that Albert’s false belief is doing a tremendous amount of good epistemically. Puddifoot firmly establishes, then, that when making epistemic judgements we must look to far more than the truth or falsity of the beliefs or stereotypes at issue.
This leads to a further suggestion of Puddifoot’s—that Albert’s false belief may be epistemically innocent: “a cognition is epistemically innocent if there is no alternative cognition that would confer the same benefits without the costs” (2021, 78).1 She suggests that this may well be the case for a false belief like that described above. After all, it is extremely difficult to avoid automatic stereotyping if one has beliefs like women are less likely than men to have scientific expertise. Albert’s false belief confers quite considerable epistemic advantages which may well outweigh its costs. Puddifoot admits that it may in fact be difficult to determine whether there is an alternative cognition available (ibid., 79), and also that there are methods which can be used in order to avoid such automatic stereotyping-meaning that there is a real possibility of alternative cognitions with the same benefits and fewer costs. (We will return to this latter point shortly.)
Puddifoot’s preferred method of assessing beliefs/stereotypes is Evaluative Dispositionalism, which requires examining both the dispositions that give rise to the cognition in question and the dispositions flowing from it. This would allow us to say that the person in the example above has an epistemically flawed belief, both in terms of its falsehood and in terms of the epistemic faults that gave rise to it, while at the same time appreciating the epistemic benefits that it brings. We can also identify epistemic flaws that may flow from it—such as the disposition to not appreciate arguments for working to increase the representation of women in science (we see these in Roger, a later example Puddifoot discusses). There is a wonderful complexity in all of these discussions, one which makes very clear the folly of simple-minded judgments of stereotypes or beliefs as simply good or bad.
Nonetheless, Puddifoot insists that her approach can give clear guidance in how to evaluate acts of believing stereotypes:
[T]he approach provides clear prescriptions about how to approach acts of believing stereotypes: check the dispositions associated with so believing. It allows us to distinguish satisfactorily between two different acts of stereotyping, explaining, for example, how two people can believe the same stereotype, under the same circumstances, will be deserving of different levels of praise and criticism. (2021, 12)
2. Reflections on How Stereotypes Deceive Us
As I reflected further on this rich and interesting picture, I began to wonder whether Puddifoot should take this complexity yet further—by abandoning the very idea of epistemically assessing individual beliefs or stereotypes themselves at all (except perhaps in terms of truth or falsehood). Or, at the very least, abandoning the idea that this is a simple matter—and that Evaluative Dispositionalism can guide us successfully in deciding which individual cognitions deserve praise or blame.
My suggestion here is that it is no easy matter to “check the dispositions associated with so believing” (ibid., 12). We can never rest content that we have looked at all the relevant dispositions in order to correctly apportion praise and blame. We can certainly praise or blame people for the dispositions that lead up to their beliefs or stereotypes. And we can praise or blame them for particular acts which flow from those beliefs or stereotypes. But we cannot possibly know enough about all the dispositions that they have in order to properly apportion praise and blame for those. Moreover, I will suggest that it will often be impossible to single out particular cognitions for praise or blame. I have two broad reasons for suggesting this: the very many dispositions that there are, and the difficulty of singling out a particular cognition as responsible for any dispositions.
2.1 Very many dispositions
The concern which I call Very Many Dispositions is that any act of stereotyping or believing something will give rise to a huge range of dispositions, arguably an infinite number of them. At any rate, the number is so great that we can never reach any firm evaluation. Consider, for example, our evaluation of Albert’s false belief as having more epistemic benefits than costs. This seems right, given the dispositions that we considered. But now let’s add some more. Suppose that if Albert were to be asked to approve funds for a study of the under-representation of women in science. This would have huge costs not just in terms of the epistemic injustice done to the person suggesting this, but also in terms of the potential loss of the knowledge which could be gained by such a study. However, things might not go that way. Rather than being asked about implementing such a study, Albert might be asked to make a huge number of hiring decisions, thereby increasing the benefits that come from his lack of tendency to stereotype women as less expert at science. Or perhaps Albert might leave science for a career as a bartender, and have no notable further dispositions arising from his beliefs about gender and science. Any of these things might happen, and Albert would have dispositions with respect to each of them. And I have only begun to scratch the surface. Any evaluation of Albert’s belief in terms of the very many dispositions it gives rise to begins to look pretty unfeasible.
2.2 Which cognition?
It is also important to note that no cognition gives rise to dispositions completely on its own. Any act of believing or act of stereotyping is carried out by a person with various other beliefs, dispositions, habits, skills, preferences, and the like. This point has already been touched on, by noting that one thing a person might do if they realize that there are more men than women in science is take action to prevent themself from acting automatically on the basis of this stereotype. Now consider the case of Betty, who believes that men are more likely than women to have scientific expertise. This belief gives rise to automatic associations which cause her to make all the bad epistemic moves that Albert avoided, leading Betty to underrate the competence of women scientists that she encounters, and making her more likely to hire men than women as junior scientists. It may seem obvious at first that this is a case in which Betty’s stereotyping belief is at fault—despite its truth, it has a large range of negative epistemic consequences and should receive a poor epistemic evaluation on an Evaluative Dispositionalist account. So far so good for that account.
But now imagine something further. Betty has been to an equality and diversity training session about implicit bias. She knows that there is a real risk of her true belief giving rise to automatic stereotypes which caused her to underestimate the worth of women scientists. The trainers were extremely skilled, and explained a variety of techniques that Betty could use to try to keep her true belief about gender and science from giving rise to biased future cognitions. Betty was not interested in doing this and decided to shred the helpful handout that she was given. Let’s also imagine that Betty’s colleague, Caleb, reacted very differently to the training. He held the same belief as Betty about the relative likelihood of men and women having expertise in science, but he was horrified to learn of the consequences this might bring. He decided to work very hard to keep this belief from bringing with it future biased cognitions. He kept the handout, did further research, and tried out all the techniques that he could find. Through these efforts, he succeeded in blocking that belief from bringing about biased cognitions. After all of this is in place, it starts to seem quite strange to place the blame for the biased cognitions on the belief about the frequency of men and women in science. It seems as though the real blame should focus instead on Betty’s belief that it isn’t worth doing anything about future biased cognitions, on her decision not to take any action to combat future biased cognitions, and so on.
Puddifoot herself would readily admit that these two different cases deserve different judgments. Indeed, one advantage that she cites of Evaluative Dispositionalism is the “ability to explain how two people with same stereotype, believed under same circumstances, may deserve different levels of praise and blame” (2021, 12). However, crucially: for Puddifoot, these different levels of praise and blame attach to the act of believing the stereotype. My question here is whether that’s really what the praise and blame should attach to. My suggestion is that in real cases, once we spell out enough details to understand the differences between the believers involved, it becomes difficult to single out any one cognition as the proper target for praise or blame. It seems more likely to me that, to put it in a Quinean way, our cognitions face the tribunal “not individually, but only as a corporate body” (Quine 1951, 38).
Let’s consider another example to motivate this thought. Imagine that Dorinda is a woman scientist. She is well aware of the under-representation of women in science, and indeed quite devoted to fighting it. As a result, she is aware that men receive unfair advantages relative to women, and has adopted the policy of assuming that where a man and a woman in science look equally well-qualified on paper, the woman is actually more qualified—since she has managed these achievements despite the barriers of being a woman in science. Now consider Edith, who is just like Dorinda except for one thing: Edith is also very aware of the under-representation of Black men in science. When Edith encounters a man and woman in science who look equally well-qualified on paper, she also takes time to think about the man’s race before concluding that the woman will have faced more barriers. Dorinda and Edith will behave quite differently when they encounter the CV of a Black man in science. Dorinda will falsely assume that he is unlikely to have encountered many barriers, since he is a man. Edith will assume that he is very likely to have encountered barriers, since he is a Black man.
Once more, Puddifoot can certainly capture the difference between these cases. She can say that the stereotyping belief held by both Dorinda and Edith leads to different results in the two cases, and assess it as epistemically problematic in Dorinda’s case but not Edith’s. But once more, it seems to me puzzling to single out the belief in this way. It seems like we capture the situation better if we look at the whole picture of Dorinda’s and Edith’s beliefs. When thinking about praise and blame, surely we should focus on their attention or lack of attention to race, rather than just on their shared belief about gender.
My thought here is that what dispositions people have arising from any belief depends on many other facts about them—including, crucially, what other things they believe. Given this, I wonder why an Evaluative Dispositionalist should want to single out a particular belief for evaluation.
3. Conclusion
The Very Many Dispositions worry makes me wonder how an Evaluative Dispositionalist could ever confidently make a judgment about whether someone’s dispositions are on balance good or bad. The Which Cognition? worry makes me wonder whether an Evaluative Dispositionalist should ever single out a particular cognition for evaluation anyway. Together, these lead me to the thought that perhaps Puddifoot’s arguments could lead us to an even more surprising place—one where we can’t assess the epistemic worth of any individual belief or stereotype. We could still discuss the epistemic merits of how it was formed, and discuss the epistemic consequences that this belief or stereotype has for a particular person in a particular situation. But overall epistemic evaluations of stereotypes, or even acts of stereotyping, would be something that we should try to avoid.
Acknowledgments
I’m very grateful to Marina Trakas for organising this symposium and special issue, and to Leonie Smith for helpful comments on this paper.