A recent study conducted by researchers at UCLA reveals that biological brains and artificial intelligence systems develop similar neural patterns during social interactions. The study, published in the journal Nature, compared the neural network patterns of mice and AI agents during social and non-social tasks.
UCLA researchers identified high-dimensional “shared” and “unique” neural subspaces when mice interact socially. These patterns were also observed in AI agents engaging in social behaviors. This discovery could enhance understanding of human social disorders and aid in developing AI capable of understanding and participating in social interactions.
The research team, comprising members from UCLA’s departments of neurobiology, biological chemistry, bioengineering, electrical and computer engineering, and computer science, used advanced brain imaging techniques to record activity from neurons in the dorsomedial prefrontal cortex of mice during social interactions. They developed a computational framework to identify these neural subspaces across interacting individuals. This framework was then applied to train AI agents for social interaction analysis.
Findings revealed parallels between biological and artificial systems’ neural activities during social interaction. In both cases, activity could be divided into a “shared neural subspace,” showing synchronized patterns between entities, and a “unique neural subspace,” specific to each individual.
GABAergic neurons showed larger shared spaces compared with glutamatergic neurons. When applied to AI agents, disrupting shared components reduced their social behaviors significantly.
“This discovery fundamentally changes how we think about social behavior across all intelligent systems,” said Weizhe Hong, professor at UCLA and lead author of the study. “We’ve shown for the first time that the neural mechanisms driving social interaction are remarkably similar between biological brains and artificial intelligence systems.”
The research team plans further investigations into shared neural dynamics within more complex interactions. They aim to explore disruptions’ roles in contributing to social disorders while developing therapeutic interventions for restoring healthy inter-brain synchronization patterns.
The study was led by Weizhe Hong along with Jonathan Kao at UCLA. Co-first authors Xingjian Zhang and Nguyen Phi contributed alongside collaborators Qin Li, Ryan Gorzek, Niklas Zwingenberger, Shan Huang, John Zhou, Lyle Kingsbury, Tara Raam, Ye Emily Wu, and Don Wei.
