Jurors recommend death penalty based on looks, but new training can correct the bias

Certain facial features—like downturned lips and a heavy brow—are known to make someone appear untrustworthy to others, even though these do not indicate a person’s actual character. Such facial biases influence our everyday social interactions as well as high-stakes decisions, including who we hire, or elect to political office.

But a new study by Columbia researchers shows that the effects of these judgments can be mitigated. The study outlines the results of four experiments that the authors conducted with 1,400 volunteers. Through those experiments, the researchers found that when real-world defendants have facial features that appear untrustworthy, they are more likely to be sentenced to death than life in prison. They also found that mock jurors were more likely to recommend a ruling against hypothetical defendants with an untrustworthy facial appearance. To get people to overcome these biases, the researchers developed a training intervention. Participants who underwent the training stopped relying on facial stereotypes, while participants in a control group who never received training remained strongly biased.

The findings are reported by a group led by Jon Freeman, an associate professor of psychology, in the journal Psychological Science. The other authors were Youngki Hong and Kao-Wei Chua, who were postdoctoral researchers at Columbia.

The researchers asked the participants to decide whose mugshots they felt were trustworthy or untrustworthy among 400 inmates in Florida who were convicted of murder. The inmates whose facial features were judged to be less trustworthy were far more likely to be sentenced to death than their counterparts without those features. This was even so in cases when participants’ conscious decisions showed no bias against certain facial types. Using a test known as a sequential priming paradigm, the researchers could show that these participants did, in fact, harbor unconscious biases that predicted who was ultimately sentenced to death.

The intervention trained participants to dismantle their unconscious associations between specific facial features and an untrustworthy reaction using a computer task. Unlike educating or “nudging” participants not to rely on facial appearance in a conscious and deliberate way, Freeman and his colleagues’ training works by making the implicit link in people’s minds between certain facial features and an untrustworthy reaction as no longer stable or reliable. They did this by having participants associate untrustworthy-looking facial features with trustworthy behaviors, severing the implicit link between these features and untrustworthiness.

While prior research testing interventions that raise people’s awareness of their facial bias and ask them to stop have failed to achieve success in reducing that bias, this new intervention operating on more unconscious principles was able to eliminate facial biases very successfully. The researchers were able to eliminate bias not only in participants’ conscious decisions but also in their unconscious reactions. This is important because unconscious reactions can still wreak havoc on people’s behavior, even when conscious decisions appear to be unbiased.

Racial and gender biases also strongly affect how trustworthy or untrustworthy another person is judged, biases that co-exist with the facial stereotypes the researchers studied, like downward-turned lips and a heavy brow. The researchers therefore conducted their studies only with white male faces to control for racial and gender biases. With the effects established, they are currently following up by testing the intervention with racially and gender diverse faces.

“These findings bolster prior work that facial stereotypes may have disastrous effects in the real world, but, more importantly, provide a potential inroad toward combating these sorts of biases,” Freeman said. “By exposing a cognitive pathway toward eradicating facial stereotypes, future research must investigate whether this training could be broadly applied and how to ensure the bias reduction persists over time.”

“If there are consequential judgments that are biased by facial stereotypes, our findings suggest that they have the potential to be flexibly remapped and dismantled,” the paper concludes.