Racial Bias Warps How We See Faces

Your Brain Warps People’s Faces To Match Stereotypes, New Research Shows


CREDIT: Shutterstock/Laurel Raymond

Implicit biases are incredibly powerful. Unconscious stereotypes affect how we evaluate who is running for political officewho we promote at work, who we encourage academically, and, perhaps most devastatingly, who we regard and react to as threats.

A huge body of research shows that black Americans are perceived to be more threatening than white Americans, and are stereotyped as more likely to be involved in criminal behavior. During split-second decisions, those biases can have devastating consequences — particularly when it comes to policing and the justice system. In 2015, young black men were nine times as likely to be killed by the police as other Americans, even though about 25 percent of them were unarmed, according to reporting by the Guardian.

Now, new research published in Nature Neuroscience shows that stereotypes about race and gender might be even more powerful than we thought — distorting what we see from the earliest steps of processing so that we actually see people as what we expect them to be.

“Our findings provide evidence that the stereotypes we hold can systematically alter the brain’s visual representation of a face, distorting what we see to be more in line with our biased expectations,” Jonathon Freeman, one of the study’s authors and a psychology professor at NYU, said in a press release.

In order to reach that conclusion, Freeman and his colleague, Ryan Stolier, showed study participants a series of faces that varied in race (black, white, or Asian), gender (male or female), and emotion (happy or angry). They measured participants’ neural activity using an fMRI machine and asked them to do a mouse-tracking task, where they had to click on the correct descriptor for the face on screen. By tracking the trajectory of the mouse as participants made their choices, they could identify people’s early, split-second categorizations.

They found that people displayed a lot of initial associations that don’t make objective sense, but that strongly correlate with problematic cultural stereotypes. For instance, female faces were initially perceived as happy, and male faces as angry — even when it was actually the opposite. Asian faces of both genders were initially seen as female, and black faces tended to be seen as male. Black faces also tended to be seen as angry even when they were objectively happy — particularly when the face was both black and male.

In a follow-up study, the authors confirmed that these erroneous associations matched culturally-held stereotypes, and also showed that the more strongly a participant identified that stereotype, the more strongly it influenced their behavior.

How stereotypes warp vision in the brain

The fMRI portion of the study backed up these results — and, importantly, suggests that these biases are kicking in during basic visual processing.

Two areas of the brain, the fusiform cortex and the orbito frontal cortex, lit up when participants saw faces. The fusiform cortex is located at the back of the brain and is involved in recognizing faces, while the orbito frontal cortex sits at the very front of the brain and is involved in “top-down” processing — learned associations and context clues that help us navigate the world, including social categories and implicit stereotypes.

This suggests an interaction between these two locations as the brain tries to figure out what its seeing — with top-down expectations warping even the first steps of visual recognition.

Even more compellingly, the neural-activation patterns were more similar for associations of gender, race, and emotion that matched incidents of bias — for example, when people saw male faces and when people saw angry faces, their brains reacted in similar ways, suggesting an interaction between them in the brain. People who showed more bias in the mouse-tracking task also displayed a stronger pattern of neural-activity matching.

Together, the two experiments show that learned associations actually change what we see into what we expect to see. The systematic first step on seeing an angry woman’s face, for example, would be to process it as happy — literally altering the first image we get of her.

Stereotypes are learned from culture

Although this study doesn’t get into where the stereotypes come from, the authors note that “in all likelihood, [the associations] were acquired across participants’ lifespans through cultural transmission and implicit learning.”

In other words, our cultural environment is full of stereotypes associating groups with different traits that actually have little to do with gender and race identity. Women are constantly told to smile — reflecting stereotypes of femininity — while men are associated with masculine aggression. Black men in particular are stereotyped in a threatening light, assumed to be incredibly strong, aggressive, and angry. At some level, that’s what our brains see even when it’s objectively false.

Importantly, though, this study doesn’t suggest that these biases are innate. Participants’ biased responses reflected learned stereotypes — absorbed from a cultural soup of media, advertising, and entertainment. More nuanced, balanced, and diverse portrayals could help lessen the effect, and learned biases can be unlearned. Since this new research provides important information about how biases work, that could help people address and eliminate them.

“Ultimately, this research could be used to develop better interventions to reduce or possibly eliminate unconscious biases,” said Freeman in a press release.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s