How observers resolve social ambiguities: two complementary studies; Investigating predictive processing: psychophysics and computational methods

On November 17, 2020 at 12:00 pm till 1:00 pm
Setayesh Radkani (Saxe + Jazayeri Lab), Isaac Treves (Sinha + Gabrieli Lab)

Zoom Webinar URL: https://mit.zoom.us/j/94423030575

(Please note: an earlier talk announcement had the talk abstracts swapped between the speakers. This message has the correct order!)

Setayesh Radkani

How observers resolve social ambiguities: two complementary studies

By observing other people’s behavior, we make inferences about those people and about the world. However, these inferences do not merely depend on others’ actions and the outcomes they bring about. We can see the same action, with the same outcome, but end up making really different interpretations which have really different consequences for what we think of those people and what we think of the world.

In this talk, I will argue that the background information and our priors affect how we interpret our observations. I will present my two projects on studying such inferences at a cognitive and neural implementation level, investigating 1) how the different pieces of information interact to influence our attributions and inferences, and 2) how these inferences are carried out by the neural computations.

Isaac Treves

Investigating predictive processing: psychophysics and computational methods

Prediction, put simply, means using past information to guide our actions and interpret future events. Predictive processing theories argue that prediction is a fundamental, unifying brain function, with a shared mechanism across vision, audition and other sensory domains. I aim to investigate predictive processing using a three-pronged approach based on psychophysics, computational modeling, and neuroimaging. In this talk, I present two novel predictive psychophysical paradigms. One is sequence termination, where participants view sequences of stimuli and terminate the sequences when they can predict the next stimulus. The other is an anchored serial reaction time task, where participants react serially to structured sequences of letters or numbers. I explain how computational models allow us to specify precise hypotheses about how these predictions are made, asking what kinds of structure in the sequences guide participants’ predictions. I hope my research will provide new insight into autism, joining the emerging field of computational psychiatry.

Zoom Webinar