Changing implicit bias doesn’t necessarily change behavior, research suggests.
The concept of implicit bias has made its way into the general consciousness, most often in the context of racial bias. More broadly, however, implicit biases can affect how people think of anything.
“All the little ways in which our everyday thinking about social stuff is unconscious or uncontrollable,” writes Calvin Lai, assistant professor of psychology at Washington University in St. Louis, in an article in DCist. “The stuff that we don’t realize is influencing us when we make decisions.”
It seems logical enough to assume that if you change the bias, changes in the behavior will follow. If true, reducing implicit bias could be put to practical use for anything from ending discrimination (removing a bias in favor of white men) to losing weight (dialing down a cookie bias).
In a meta-analysis of research papers about implicit bias, however, Lai found that the evidence does not show this kind of causal relationship. The results appear in the Journal of Personality and Social Psychology.
Why the missing link?
Lai worked with Patrick Forscher, of the University of Arkansas, to systematically review 492 studies that dealt with changing people’s “automatic mental processes,” the uncontrollable, unconscious mental processes that have come to be known in particular contexts as “implicit bias.”
The studies contained more than 87,000 participants. After crunching the numbers, Lai and Forscher saw that studies suggest biases can, in fact, change—although not dramatically.
When they homed in, looking at 63 studies that explicitly considered a link between changes in bias and changes in actions, however, they found no evidence of a causal relationship.
“We definitely didn’t expect this,” Lai says. “And it challenges assumptions about the relationship between implicit bias and behavior.”
Lai suggests four possible reasons that the meta-analysis didn’t establish a link:
- Measurement errors: The way outcomes were measured may have picked up on changes unrelated to the underlying bias. For example, Lai says, such a measurement would be analogous to “moving the mercury around within a thermometer rather than changing the heat in the room.”
- Confounds: After tests to measure an implicit bias, something happened, unrelated to the changed subjects’ behavior.
- Measured too narrow of a bias: Appeared to assess the same associations, but maybe the effects were too broad to capture a change associated with the change in bias. For example, the implicit bias measured was about broad attitudes toward white vs. black people, but the behaviors measured were about behavior toward a specific person of a particular race. In that case, the attitude measured may have been too general.
- No causal relationship: Implicit bias doesn’t affect behavior at all.
This last option doesn’t sit well with Lai. “It would open a theoretical can of worms because there are decades of experiments in other lines of research showing evaluation without conscious intention or control,” he says.
Change the structure instead
However, Lai says there is a more effective way to change these behaviors; one that doesn’t rely on changing people’s implicit biases: ridding society of the features that cause people to act in a biased way.
For example, reducing subjectivity makes it more difficult for a person’s biases to affect decision-making. Instead of relying on a “gut feeling” for a hiring decision, for example, lay out the requirements first, and stick to them.
Or, in the cookie realm, don’t have any on hand—not at home or at the office—and don’t drive past the bakery on the way home.
On an individual level, Lai says, “Equip people with strategies to resist the environment’s biasing influence.
“The power of counter-stereotypes is not to be underestimated,” Lai writes in a paper describing possible ways to counteract implicit biases. “And if counter-stereotypical encounters become typical, shift in attitudes and beliefs will follow.”
Lai points out that this study was heavily constrained by the available literature. The studies they included focused on brief interventions and assessments and was heavily skewed toward a certain demographic: university students.
“At this time, we can’t distinguish between these potential explanations for the results,” Lai says. “Carefully controlled studies need to be conducted to rule out—or rule in—these explanations.”
