Cognitive bots and algorithmic humans: toward a shared understanding of social intelligence

Questions of social behavior are simultaneously among the most fundamental in neuroscience and the most challenging in artificial intelligence. Yet despite decades of work, a unified perspective from the cognitive and computational approaches to the problem has yet to emerge. Recently, however, excitement around the challenges posed to reinforcement learning by multiplayer video games, coupled with the adoption of more complex modeling strategies in social neuroscience, has broadened the interface between the two fields. Here, we review recent progress from both directions, arguing that advances in artificial intelligence provide neuroscientists with valuable tools for modeling social interactions. At the same time, the study of humans as efficient social learners can inform the design of new algorithms for multi-agent systems. We conclude by encouraging a joint approach that incorporates the best of both domains to advance a shared picture of social intelligence.

Duke Scholars

Author John Pearson Neurobiology