How AI and Brain Activity Uncover Racial Biases in Face Perception

Recent research conducted by scientists at the University of Toronto Scarborough has uncovered new insights into how we perceive faces from different racial backgrounds. By combining artificial intelligence (AI) with brain activity measurements, specifically EEG (electroencephalography), researchers are exploring the well-known phenomenon called the Other-Race Effect (ORE). This effect describes how individuals tend to recognize faces of their own race more accurately than those from other races.

In their studies, scientists used advanced AI techniques, such as generative adversarial networks (GANs), to generate visual representations of faces based on participants’ responses. In one notable experiment, two groups—one East Asian and one White—viewed numerous faces and rated their similarity. The AI then reconstructed these faces from the mental images of the participants. Results showed that faces of the same race were reconstructed with greater accuracy, while faces of other races appeared more averaged, younger, and even more expressive in these mental images.

Another significant aspect of this research involved analyzing brain activity during face perception. Using EEG data collected within the first 600 milliseconds of viewing images, researchers digitally reconstructed how faces are processed in the brain. Findings indicated that faces from different races are processed differently: same-race faces elicit more distinct neural responses, whereas other-race faces tend to be processed more uniformly with less detail. This neural differentiation—or lack thereof—may contribute to the difficulty in recognizing and distinguishing faces from other races.

Interestingly, the studies also revealed that faces from other races are perceived as younger and more expressive, adding another layer to understanding perceptual biases. These findings suggest that the brain’s representational processes influence how facial features are perceived and may reinforce racial biases in face recognition.

The implications of this research are far-reaching. Better understanding of how racial biases form in the brain could inform efforts to reduce prejudice and improve facial recognition technologies. Moreover, this approach could serve as a diagnostic tool for mental health conditions like schizophrenia or borderline personality disorder, where perceptual distortions are common. By deciphering how individuals process emotional and facial cues, clinicians could develop more targeted interventions.

Ultimately, this innovative work highlights the potential to utilize AI and neural data to understand human perception better and address societal biases, paving the way for more equitable social interactions and advances in mental health diagnostics.