Dr. Peter Dayan Group Talk

Alireza Modirshanechi

Title: An integrative framework for the human sense of control

Abstract: From managing daily schedules to making major life decisions, how much we feel in control of our lives is central to our behavior and well-being. Yet, there is currently no precise, unifying account of what it means to feel in control. Here, we propose that humans feel in control when they are ready to tackle tasks that may arise in the future. We formalize this perspective via a precise yet interpretable measure of controllability that quantifies three key factors: (i) the range of available actions, (ii) the certainty or (iii) the possibility of achieving desired outcomes. This measure unifies previous conceptions of control and successfully predicts human preferences in two experiments in which participants prepare themselves to seek rewards or avoid punishments. Our framework offers a principled foundation for understanding how environmental structure shapes the human sense of control, paving the way towards an integrative science of agency.

Emerson Harkin

Title: Dopamine: From reward prediction to error

Abstract: Emerson will present an interactive, web-based timeline of dopamine research showcasing the interplay between theoretical and experimental work from the 1980s to today. In this short chronological tour of the dopamine field, we will begin by reviewing the origins of temporal difference learning as a model of animal conditioning in the early 1980s. Next, we will survey some of the primary dopamine physiology data collected by the Schultz lab in the late 1980s and early 1990s, and how the reward prediction error hypothesis published in the mid 1990s managed to unify these puzzling data by drawing on ideas from temporal difference learning. Finally, we will explore how this hypothesis has been at times reinforced at and times challenged over the years, with a recurring theme being the tension between learning- and behavioural-activation-centric views of dopaminergic function.

Wenting Wang

Title: Decomposing Response Inhibition: A POMDP Framework for Understanding ADHD

Abstract: We introduce a Partially Observable Markov Decision Process (POMDP) framework to decompose response inhibition, integrating Bayesian inference to explain perception with value-based decision making. When applied to behavioral data from more than 3,000 children, we uncover a subtle but robust difference in ADHD. Even though the population is highly heterogeneous, we show that specific computational attributes map directly to altered perceptual and intrinsic valuation dynamics.

Website: https://www.kyb.tuebingen.mpg.de/computational-neuroscience

https://sites.google.com/view/modirsha

https://efharkin.com

Access to the meeting: Zoom link