Common test of ‘implicit’ racial bias fails to predict prosocial behavior in new study

A common test used to measure unconscious bias may not be a good predictor of actual behavior, suggests new research published in the journal Games. The study examined the Implicit Association Test, a popular tool that is used to measure people’s biases against certain social groups.

The IAT measures differences in reaction time taken to categorize faces into groups, depending on whether they are paired with positive or negative words. People are quicker to associate negative words with things they are biased against.

“I first became interested in the IAT during my second year in grad school. My father (who is an employment lawyer) was at a conference where he and his colleagues took one as a didactic exercise. In telling me about it, he stressed that he could feel himself slowing down when the less associated categories were paired,” explained Daniel J. Lee of Rice University, the author of the study.

“Now I’m an economist, so naturally my first thought was how does this map onto behavior ‘in the wild’. I found this question of particular importance because people were using this assessment to draw conclusions about complex economic constructs like labor market matching.”

“As I explored the IAT research more deeply, one thing that struck me was there hadn’t been much work that explored how the IAT predicted costly behavior, and this is where I think the toolbox of experimental economics can be used to shed light on the topic of implicit bias,” Lee explained.

The study of 456 participants found no evidence that scores on the IAT were associated with racially-biased behavior in a dictator game. In this popular experiment, a participant is given a fixed amount of money that he or she can choose to share — or not share — with another person. The one-sided game is used to measure prosocial behavior.

Implicit bias did not predict how much money was shared with people of another race. For example, “dictators” who had a pro-white bias as indicated by the IAT were no more or less likely to share their money with a black person.

“There is a difference between having a bias and acting on one, which is not to say there is no overlap between the two, but it is to say we need to understand when and why that overlap exists,” Lee told PsyPost.

“The second takeaway stems from this point which is an overall need to interrogate our assumptions, and do think and act more critically.”

The study — like all research — has limitations. Lee only used a single measure of behavior but other behavioral tests could lead to different results.

“The dictator game is a very straightforward game. On the one hand, this is very exciting because we can think of it as ‘bounding’ the effects of implicit bias. On the other hand, perhaps what we are learning is people act on biases when decisions are harder, fuzzier, or they have to rely on heuristics,” Lee explained.

“Unfortunately, we live in a world where bias is real and bias is profound. However, the science of bias, particularly implicit bias, is still nascent. I’d encourage everyone reading this to take an IAT ( but then to turn back and ask themselves what are they learning from it, and how we can use that information to improve the world around us.”

The study was titled: “Does Implicit Bias Predict Dictator Giving?“.