Computational Approaches to Human Affective Neuroscience

Lecture | September 27 | 3 p.m. | 5101 Tolman Hall

 Sonia Bishop, Professor, UC Berkeley

 Department of Psychology

Abstract: Computational modelling allows us to move beyond simple approaches to experimental design. Here, I will present two very different examples of integrating computational modelling into human affective neuroscience. In the first example, we sought to better characterize the mechanisms underlying intolerance of uncertainty in anxiety. Participants performed bandit style decision-making tasks, both inside and outside of the MRI scanner. Using a computational approach enables us to model the influence of different types of uncertainty – including risk, volatility and ambiguity – upon participants’ choice behaviors and to determine which parameters differentially influence behavior and brain function in high versus low anxious individuals. In the second example, we used multi-feature regression models to explore the cortical representation of emotional natural images. By fitting alternate models to cortical BOLD time-courses measured while participants view these images, we can examine which regions of cortex represent semantic versus structural image features. Further, we can effectively model tuning curves across cortex, identifying the image features that drive the BOLD signal in any given voxel. This in turn can be used to test theoretical models - such as whether facial emotion and identity are processed by independent cortical pathways. Reception: To follow the talk, in the Beach Room 3105 Tolman.

 nrterranova@berkeley.edu