Racial Bias Can Be Reduced By Teaching People To Differentiate Facial Features Better In Individuals Of A Different Race

Brown University and University of Victoria researchers learned this through a new measurement system and protocol they developed to train Caucasian subjects to recognize different African American faces.

“The idea is this that this sort of perceptual training gives you a new tool to address the kinds of biases people show unconsciously and may not even be aware they have,” said Michael J. Tarr, a Brown cognitive neuroscientist and a senior author of the paper.  “There is a strong connection between the way we perceive and categorize the world and the way we end up making stereotypes and generalizations about social entities.”

The research is the product of a wide collaboration. Sophie Lebrecht, a third-year Ph.D student in the Department of Cognitive and Linguistic Sciences and a member of Tarr’s lab, is the study’s lead author. Jim Tanaka, a professor at the University of Victoria and Lara Pierce, a graduate student at McGill University, collaborated on the research.

Lebrecht was interested in the interaction of visual processing with other cognitive functions such as emotion or social processing. She came up with the idea for the project with Tarr’s encouragement.

Researchers used 20 Caucasian subjects for the overall study, which incorporated a measurement developed at Brown and dubbed the Affective Lexical Priming Score (ALPS). The ALPS measure is similar to, and builds on, a test developed at Harvard University known as the Implicit Association Test (IAT), which helps to identify unconscious social biases.

The ALPS measurement involved first showing each subject a series of pictures of different races, such as African American and Caucasian faces. All the faces were shown in black and white, so subjects would focus on facial features rather than skin color.

On each ALPS trial, each test subject was shown a picture of a face, which then disappeared. The test subject then saw a word that could be real or nonsense — “tree” or “malk,” for example — and had to decide whether the word was a real word or nonsense word. Real words could imply something positive or negative.

Lebrecht found that prior to training, subjects more quickly responded if the word was negative and followed an African-American face. Subjects responded more slowly if the word was positive and followed an African-American face.

After using the ALPS to measure each subjects’ implicit racial bias, the subjects took part in about 10 hours of facial recognition training. Half learned to tell apart individual African-American faces and half learned simply to tell whether the faces were African-American or not.

The training worked on a number of levels. Individual subjects improved their ability to tell the difference between separate Africa-American faces. Those same subjects who improved that ability also showed the greatest reduction in their implicit racial bias as measured by the ALPS system. Their positive associations with African-American faces increased and they had fewer negative associations with African-American faces.

While the researchers are not claiming they can eliminate racial bias, they suggest that teaching people to tell the difference better between individual faces of a different race is at least one way to help reduce that bias.

Lebrecht said that developing a system that teaches people to make those distinctions should be helpful in reducing generalizations based on social stereotypes.

“If you give people the tools to start individuating, maybe they will make more individual (rather than stereotypical) attributions,” she said.

Funding for the study came from the Perceptual Expertise Network, a collaborative award from the James S. McDonnell Foundation; the Temporal Dynamics of Learning Center at the University of California–San Diego; the National Science Foundation, a National Sciences and Engineering Research Council of Canada award; and a Brown University National Eye Institute training grant (the National Institutes of Health).