Social Cognition with Deep Recurrent Neural Networks

In our current social cognition research, we use the power of deep recurrent neural networks to represent social mental states, including states of cognitive dissonance, to measure and predict human reactions to information and then apply the results to improve the messages sent to persons. These results could automatically detect and alert to attempts at psychological manipulation that take advantage of human cognitive dissonance and tribalism, or they could simulate realistic social reactions on an individual level, such as feedback between Little Sophie robot and a “parent.” Our scientists originally wrote similar programs using the Boltzmann machine to simulate population reactions to information operations. This tool can be applied to any case with multiple, possibly dissonant, social information messages, whether it focuses on the individual or a population.