Events Calendar

AI SEMINAR - Towards a computational framework for how we represent other people

Speaker: Damian Stanley, Caltech

Talk Title: Towards a computational framework for how we represent other people

Abstract: Predicting other peoples’ beliefs, desires, and intentions is a primary function of human cognition and is essential to thrive in our complex social world. To do this efficiently and successfully, we must form lasting representations of individuals and social groups based on information we receive through personal and vicarious experience. My research is focused on developing a computational account of the neurocognitive mechanisms through which we learn about other people, make social predictions, and are influenced by social biases. To achieve this, I employ a multidisciplinary approach, integrating a wide range of techniques from cognitive neuroscience, social psychology, neuroeconomics, computational modeling of learning and decision-making, and clinical psychology. My theoretical model of social learning and decision-making treats social group biases as a set of initial guesses (akin to Bayesian priors) that inform our social decision-making when we lack specific information about a person with whom we are interacting. Using these priors as a starting point, we form and update our mental representation of a person (as well as their social group) on the basis of observed behavior. I will present behavioral and neural data on the influence of race bias on trust estimations, as well as the computational processes through which we learn about individuals’ traits and intentions (i.e., theory of mind), and how these processes might be disrupted in individuals with social impairments (e.g. Autism Spectrum Disorder). These results suggest that while many common processes support learning about social and non-social entities, there may also exist neural computations unique to social learning.

Biography: Damian Stanley completed his Ph.D. in Neural Science at New York University in 2005, studying mid-level visual processing. In his postdoctoral work, he turned his focus toward developing a computational account of the neurocognitive processesthrough which we learn about and represent other people. In his first postdoctoral position with Elizabeth Phelps at New York University he investigated how implicit race biases influence social trust. In his current postdoctoral position, with Drs. Ralph Adolphs and John O’Doherty at Caltech, he uses computational models and model-based fMRI to characterize typical and atypical (e.g. autism spectrum disorder) social learning. This line of research is funded by an NIMH career development award (K01-MH099343).

Host: Greg Ver Steeg

Webcast: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=8d563808c16942bda353a815b33370d01d

Location: Information Science Institute (ISI) - 11th floor conference room

WebCast Link: http://webcasterms1.isi.edu/mediasite/Viewer/?peid=8d563808c16942bda353a815b33370d01d

Audiences: Everyone Is Invited

Contact: Kary LAU