Differences in facial expression perception between artificial intelligence and human brains examined
Researchers at the University of York have revealed that artificial intelligence models recognize facial expressions, but do so through a process distinct from human and primate brains.
A new study conducted at the University of York has revealed that although artificial intelligence models can recognize facial expressions with high accuracy, they process this information using mechanisms different from those of human and primate brains. Published in the journal Nature Communications, the study demonstrated that general-purpose object recognition models mimic biological error patterns better than models trained specifically on faces.
Comparison of artificial intelligence and the brain
Researchers at the University of York examined how the facial expression perception processes of artificial intelligence models differ from human and primate brains. Within the scope of the study, eleven artificial neural network models, 290 human participants, and two primates were compared.
Scope and methodology of the research
Led by Maren Wehrheim, the team used 360 facial images containing six different expressions: anger, disgust, fear, joy, sadness, and shame. Perceptual processes were measured by presenting the images to participants for short durations.
Success of general-purpose models
The researchers reached a surprising conclusion by determining that general-purpose object recognition models mimic the error and behavioral patterns of primates much better than models trained specifically on faces.
Brain activity and neural coding
Neural activities recorded from 308 different sites in the temporal cortex of macaque monkeys were compared with behavioral data. It was observed that early-stage neural signals provided the closest match with behavior.
A new framework for autism research
Although the findings do not directly examine autism, they lay the groundwork for clinical studies by offering a new computational framework to understand the mechanisms behind social perception differences.