BSI x CNP Symposium: Computational modeling of learning and motivation to understand psychopathology

This BSI x CNP Symposium — Computational modeling of learning and motivation to understand psychopathology —  consists of a colloquium lecture by dr. Massimo Silvetti, pitches by local researchers (Bingjie Li, dr. Mingqian Guo, Upasana Shah & dr. Guusje Collin), and a practical workshop on reinforcement learning by dr. Tim Vriens. This event seeks to illustrate how computational models can act as a bridge between neuroscience and behavioural science, generating testable predictions about human behavior across both healthy functioning and psychopathology.

Program Overview

It is possible to register for part of the program (only Colloquium or only workshop). When attending the workshop, attendance of all parts is advised.

Colloquium Abstract

Every day, our brains make countless decisions—not just about what action to take, but also about how much effort to invest, how much attention to allocate, and how quickly to adapt when the environment changes. Recent findings in cognitive and computational neuroscience suggest that this process is best understood as meta-learning—the brain’s ability to tune its own learning strategies to maximize outcomes (e.g., reward and information).

In this talk, Massimo Silvetti, an expert in computational modeling, will introduce the Reinforcement Meta-Learner (RML). The RML is a neurocomputational model that bridges decision-making, motivation, effort allocation, uncertainty and volatility management within a system-level neuroscience perspective. This model proposes that adaptive control of neuromodulators (i.e., dopamine and norepinephrine) allows the brain to flexibly adjust its learning and decision-making in real time, creating a feedback loop between brain dynamics and behaviour.

Specifically, Massimo Silvetti will show how the RML explains core processes such as cognitive control and effort-based decision-making, as well as volatility estimation (detecting environmental change via prediction error and adaptive gain) in foraging decisions (choosing whether to exploit or explore). Next to this, he will discuss how the RML framework can generate mechanistic hypotheses about two high-prevalence psychiatric disorders—ADHD and MDD—and present preliminary data evaluating these predictions.