Anna Welpinghus is based in Berlin, Germany. She is currently affiliated with TU Dortmund University. She is working on the nature of implicit bias and its significance for unintentional discrimination.
A post by Anna Welpinghus.
Most readers of this blog will be familiar with the basic idea behind implicit bias: Societies are structured by social hierarchies, which leave traces in the minds of its members. These can operate to some degree independently from our explicit convictions. Hence, they can lead to discriminating behavior, often without intention.
In my paper “The Imagination Model of Implicit Bias” (2019), I argued that we have good reasons to assume that imagination plays a vital role in decision making. Furthermore, if this assumption is correct, it offers an explanation for implicit bias in many considered decisions. In this blog post, I summarize the proposal of the paper and then reflect on some issues that came up in it.
The term “implicit bias” can be used for a mental state or for a disposition. In this paper, I used it for a disposition. More specifically: A person who harbors an implicit bias against members of a social group has the tendency to evaluate, perceive or judge them less favorably than members of another social group. Furthermore, this tendency is to some extent independent of her explicit convictions—biased behavior occurs without intention and sometimes despite egalitarian convictions.
I distinguish between two cases of implicit bias: bias in considered decisions and in spontaneous behavior. Take a job application process to illustrate these two cases. During a job talk, the interviewer treats a black candidate differently from a white candidate in subtle ways: he sits further away from the black candidate, he frowns a bit more and avoids eye contact, while he nods and smiles more at the white candidate (Holroyd 2016). This is bias in spontaneous interactions. An experiment by Eric Uhlmann and Geoffrey Cohen (2000) nicely illustrates bias in considered decisions. Participants judged how qualified two candidates were for a job as a police chief. They were divided in two groups. One group got a pair of CVs where the woman (“Michelle”) had received a university degree but little practical experience, while the male candidate (“Michael”) had quite some experience as a police officer but little formal education. The other group got the same pair of CVs – only the names were swapped. Most participants in both groups chose the male candidate.
In the paper I argued that imagination plays a crucial role in many considered decisions. Think of a participant in Uhlmann & Cohen’s experiment: deductive reasoning about the job requirements and candidates’ skills won’t get her to a decision. But she can imagine how the candidates will likely fare in the job. She imagines Michael having the job: which challenges would he encounter? How would he deal with them? How would colleagues and stakeholders react to the way he deals with them? She does the same for Michelle. Based on the things she imagines, she will judge how the candidates will likely fare in this job.
Imagining these things is a way to reason about counterfactual scenarios. However, it does not just consist in drawing conclusions from a limited set of premises by applying a limited set of rules of inference. Rather, imagining that p can (at least in this context) be described as a mental simulation of a counterfactual scenario. By simulating what would happen if some proposition p were the case, we draw from a wide range of general knowledge about the world. This might include tacit background beliefs about men and women in general. It may also include semantic associations. And, as Nichols and Stich (2000) put it, imagination can be described as guided by cognitive scripts (cf. Schank & Abelson 1977). Such scripts, as I understand them, contain expectations about the way women and men behave and how other people react to them. In short: the array of background knowledge we draw on contains traces of the stereotypes that are prevalent in our culture. (Note that unlike Sullivan-Bissett (2019), I do not claim that implicit attitudes are imaginings. Instead I suggest that they are among the ingredients of imagination.)
There are several ways in which these traces of stereotypes skew an evaluation: first, our biased reviewer might base her imagination on false beliefs or inaccurate scripts. Second, and maybe more important, while the scenarios she imagines are all reasonably realistic and none of them is based on blatantly false beliefs, the range of scenarios she imagines can be one-sided. Scripts, generic beliefs and maybe associations will leave some realistic scenarios unexplored: she did not imagine X although X was just as likely as the scenarios Y and Z that she has imagined. On the other hand, maybe she imagines some rather unlikely outcomes and they seem extremely plausible to her. In both cases, the scenarios she explores are one-sided. Third, affective reactions show how she evaluates the different scenarios she has imagined, but these affective reactions may get it wrong, too.
It is possible that imagination goes awry in several of these ways at the same time but one of them can already skew an overall judgment. In that case, in the mind of a biased person, the same qualification counts more for Michael than it would have counted for Michelle. This is the structure that is characteristic of implicit bias in general—a property of a person is evaluated differently as a function of her (ascribed) social group membership.
Biases regarding social groups are just one instance of a general problem with imagination. Imagination is a double-edged sword: it allows us to integrate a large body of knowledge, much more than we can handle with explicit deductive reasoning. However, among this body of knowledge is also stuff that can bias our decision-making.
Implicit bias can persist despite conflicts with other attitudes. First, implicit bias sometimes persists despite intentions to the contrary; second, it sometimes persists although a person is holding descriptive beliefs which contradict the stereotypes that apparently influence her judgments.
The first sort of conflict occurs when, for instance, a person intends that a job candidate’s gender does not influence her assessment of the candidate’s suitability for the job. Why is this intention, however sincere, not sufficient for eliminating implicit bias? It is due to the fact that we let the simulation run without intentionally choosing each and every premise. Again, what is a strength of the process also results in a weakness, in this case limited control when it would be desirable. Yet, as long as imagination plays a valuable epistemic role, we will not refrain from imagination altogether.
The second sort of conflict occurs when implicit bias persists although the person harbors egalitarian beliefs with regard to the social groups she treats or views differently. How does my model explain this? It entails that stereotypes are represented in many different ways in memory. It is not sufficient to reject some sexist assumptions in order to eliminate all those expectations that represent sexist stereotypes (and the same holds for racist and other stereotypes). One might sincerely believe that women and men are equally capable as police chiefs, and at the same time expect women to be more caring and less assertive than men; and think of police chiefs as authoritarian and assertive.
The last point is of wider significance which has not been fully explored in the paper. We tend to think of implicit bias as the biases of people whose explicit belief system is fully egalitarian. The puzzle about implicit bias is often constructed in the following way: “how can a person (apparently) sincerely believe that p but act as if not-p?” But this is misleading: people with explicit inegalitarian beliefs harbor implicit biases, too – even more so (Holroyd 2016). And many people hold some egalitarian convictions without fully understanding how some stereotypes contribute to social hierarchies. After all, these relationships can be quite complex. Another obstacle for understanding them is strategic ignorance of those profiting from a particular hierarchy, as for instance Charles Mills (2007) has argued.
Another significant upshot is that the imagination model suggests pluralism about the nature of implicit attitudes. There has been a tendency to frame the question about the nature of implicit attitudes like this: “Implicit biases are mental states – namely traces of cultural stereotypes in the mind that lead to unintentional discrimination. What sort of mental state are they?” But in framing the question like this, we lose sight of the possibility that the answer is: “Many.” A person has to have certain beliefs, emotions, desires or memories in order to have an implicit bias (understood as a disposition). But which of these mental states are decisive may differ from person to person and from case to case. And sometimes, more than one such state may bring about a particular case of biased behavior.
References
Holroyd, J. (2016). What do we want from a model of implicit cognition? (digital preprint/draft). In Proceedings of the Aristotelian Society, 116(2).
Mills, C. (2007). White Ignorance. In Shannon Sullivan and Nancy Tuana (ed.), Race and Epistemologies of Ignorance. State Univ of New York Press. pp. 11-38.
Nichols, S., & Stich, S. (2000). A cognitive theory of pretense. Cognition, 74, 115–147.
Schank, R. C., & Abelson, R. P. (1977). Scripts, plans, goals and understanding. Hillsdale: Lawrence Erlbaum.
Sullivan-Bissett, Ema 2019: Biased by our imaginings. Mind and Language, 34, 627–47.
Uhlmann, E. L., & Cohen, G. (2005).Constructed criteria: Redefining merit to justify discrimination. Psychological Science, 16, 474–480.