Brain recordings hint at how AI reads facial expressions differently from us
The human brain can recognize a person's face and expression in less than a blink of an eye, but current artificial intelligence (AI) models, although fairly accurate at doing the same, use a different process than the brain. That may not be good enough for applications in health care or education, according to new research from York University.
Researchers at York University have discovered that humans and animals process facial expressions differently than artificial intelligence (AI) systems. The study, published in Nature Communications, reveals that the brain recognizes faces and expressions in less than a blink of an eye, while AI models, although accurate, use a different process. Understanding how AI arrives at its judgments could have significant benefits in healthcare and education.
Faces play a crucial role in human communication, with small changes in facial appearance influencing our perception of someone's emotional state and social responses. The study aimed to understand how AI systems differ from humans in interpreting human social signals. The researchers tested 11 artificial neural network (ANN) models, 290 humans, and two nonhuman primates using 360 images of 12 people expressing six different facial expressions at varying strengths.
The researchers found that while AI models were generally accurate in identifying facial expressions, they differed from humans and primates in specific aspects of perception. Humans and primates displayed similar patterns of facial-expression perception, while AI models, particularly those trained on object recognition, showed more resemblance to the primates' behavior in certain error patterns.
Neural recordings from the primates' inferior temporal cortex (IT) cortex provided further insights into the brain's processing of facial expressions. The brain's activity matched the monkeys' behavior better than the AI models, indicating a heterogeneous coding architecture in the temporal cortex where facial identity and emotional expression overlap. This suggests that the brain may process facial expressions using early visual signals while continuing to refine the image.
The findings validate nonhuman primates as a suitable animal model for studying human facial-expression processing and establish a foundation for further research. The study also highlights the need for AI systems to better replicate the brain's brain-like processing of facial expressions to improve their performance in applications related to health care and education.
Written by urgent.news from Medical Xpress's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
