In Defense of Implicit Bias
The IAT has come under fire, but that doesn’t show that implicit bias is a hoax.
Posted Mar 29, 2018
Discussions of race, racism, and racial disparities pop up in our newsfeeds at what seems like a dizzying pace. People continue to be harmed in myriad ways just because of what they look like. The prevalence of the topic in the news cycle means that people, especially white people, are paying more attention than usual to a problem that, due to our privilege, we can often avoid the burden of noticing. The persistence of the problem, however, challenges our self-conception. We’d like to think we’ve transcended race and relegated racism to the dustbin of history. But reality has a way of bursting our bubbles.
Race shapes so much of what we experience and what we do, whether we notice this or not. Even those who profess colorblindness and a commitment to racial equality exhibit tendencies shaped by and reflective of centuries of racial oppression. Many such people are quicker to identify bad words when paired with pictures of black faces than with pictures of white faces, or likely to picture a black or brown face when asked to think of a drug dealer. The ways race shapes our perceptions of things can fly under the radar, and this has been pointed out to us for a long time. People talk to each other about their experiences, and this includes people of color whose experiences reflect the biases of those in positions of racial dominance. Academic studies of these tendencies confirm what we already know.
But, as with any academic debate, there is disagreement. Criticisms of the Implicit Association Test (IAT) have made headlines recently. Chief among the complaints are that the results are not stable—one can score high on the race IAT one week and low the next—and that they are not predictive of individuals’ behavior. The first concern is acute, in large part, because many researchers want a test that meets basic, commonsense standards. What good is a measure if it varies so wildly across time, even when the same tool is applied to the same subject? The second concern is worrisome, in part, because it points to what seems like false advertising. The IAT’s developers have touted it as a tool for predicting overt behavior. But even if this were not the case, one might wonder what use there is in identifying biased tendencies that don’t manifest in any biased behavior.
In a recent article for Scientific American, Keith Payne, Laura Niemi, and John Doris explain why these complaints about the IAT do not sink claims about rampant implicit bias. One thing they point out is that it doesn’t follow from the claim that a particular tool is flawed that the phenomenon it is supposed to measure is non-existent. Even if the IAT is worthless, this doesn’t show that implicit bias is a hoax. Second, they point out that predictive tools in psychology are meant to predict average group behavior, not behavior at the individual level. Complaints about the IAT miss their mark, on this score, because they don’t target what the tool is supposed to be doing.
Moreover, Payne, Niemi, and Doris point out that there is extensive evidence of real-world racial discrimination. For example, there is good evidence that prospective employers are more willing to call back applicants with typically white-sounding names than applicants with typically black-sounding names, even when their resumes are otherwise identical. It would be folly to dismiss this on the basis of worries about a single test. They might also have added that attempts to do so follow a familiar pattern: they silence and otherwise erase the testimony of those suffering oppression. To say that racial bias is a non-issue because a particular measure of it isn’t up to snuff is tantamount to saying that you don’t believe all those people who say they suffer the effects of racism because your own racism-detection tool isn’t sounding the alarm. It is to privilege the testimony of the tool over the testimony of the person. When that person is a person of color, this comes to look a lot like an instantiation of the very phenomenon under consideration.