The problems with implicit bias training

Last week, Scott reported that Dartmouth will impose mandatory “implicit bias” training on all students, faculty, and staff. Even Dartmouth’s board of trustees “has committed itself to participating in the training.”

I don’t doubt that there is such a thing as implicit bias. The bias can be against Blacks, Whites, Latinos, Jews, etc. It can also be against America, although these days much of the anti-American bias on display is pretty damn explicit.

But does what we know about implicit bias — e.g. how to measure it and what its real world consequences are — justify making implicit bias training mandatory? To explore this question, I turned to a paper by Gregory Mitchell, a professor at the University of Virginia.

Mitchell teaches at UVa’s law school. He has a law degree and doctorate in psychology from the University of California, Berkeley. His 2018 paper, “An implicit bias primer,” was published in the Virginia Journal of Social Policy & the Law.”

Mitchell begins by defining implicit bias. Bias, of course, is the tendency to favor or disfavor something. It is manifested by reacting differently to stimuli based on particular characteristics of the stimuli. Thus, if one tends to react more positively to White people than to Black people, then one is biased in favor of Whites and against Blacks.

Bias can be measured through direct or indirect measures. A direct measure asks someone to report how one feels about a particular group. If the responses show bias, the bias is explicit.

But people may be disinclined honestly to report their feelings about a group or incapable of assessing through introspection how they really feel. Hence the case for using indirect measures, in which researchers have people react to stimuli on a task that does not require verbal reporting of attitudes or beliefs about the stimuli. Researchers then infer from the pattern of responses whether a bias may exist. (Mitchell describes the most prominent of these measures at pages 32 of his primer.) If it does, the bias is said to be implicit.

This seems pretty straightforward. However, there are several complications, described by Mitchell at pages 33-39 of his primer. The upshot is that “indirect measures of bias have been found to have low to moderate test-retest reliability.” Moreover, one might receive a high score on one type of test for implicit bias and a low score on another type of test for the same phenomenon.

But even assuming that a particular test reliably measures implicit bias, the question remains whether individuals who show such bias have a tendency to engage in acts of discrimination against the group towards whom they are implicitly biased. According to Mitchell, the evidence to date is that tests for implicit bias have little ability to predict who will and who won’t discriminate in a given situation. Indeed, some studies show that “high bias” persons behave more positively towards minorities than “low bias” persons. One such study involved police officers. (Details at pages 47-48 of the primer)

Thus, if college students in a mandatory implicit bias course take an implicit bias test, it’s easy to imagine those whose test scores indicate implicit bias against Blacks concluding (1) that they hold racist views and (2) that they are prone to treating Blacks unfairly. But both conclusions might well be unjustified because (1) the test is unreliable and (2) even if reliable, it might not accurately predict behavior.

This analysis leads Mitchell to the questions of (1) whether implicit bias can be changed through training and education and (2) whether, even if it can be, the effort will have positive real world effects. Mitchell doubts that either question can be answered in the affirmative.

He does acknowledge the possibility that training programs might “increase vigilance about behavior and increase scrutiny of outcomes for possible group-based disparities.” He warns, however, that the possibility of such benefits should be weighed against the potential opportunity costs if adoption of implicit bias programs crowds out other interventions that may more directly or effectively prevent discrimination.

In addition, any possible advantages should be weighed against the potential disadvantages of implicit bias training Mitchell cites. They include loss of trust and increased suspicion within an organization, more impersonal and uncomfortable interactions among members of different groups, and compensatory measures aimed at promoting particular groups at the expense of merit-based decisionmaking.

Given the junk science nature of implicit bias education and the potential harms of undertaking it, I think that, other things being roughly equal, a student is better advised to attend a college that doesn’t require implicit bias training than a college that does.

Responses