Organization of the summer school
In general, we want the trainees to acquire both the conceptual bases and the technical skills that will enable them to pursue a full modelling approach on their own when they come back to their lab. A typical day covers one modelling topic through: an introductory lecture; tutorials where trainees apply the method to real behavioral datasets and time devoted to the application of new techniques to participants’ own data (group project). In addition, we will have two keynote research talks. The eight days are structured around the following topics (the ordering of the days may change):
Day 1: Basics of behavioral modeling – regression models of reaction times
Day 2: Regression models of decision behavior
Day 3: Process models of decision making: the drift-diffusion model
Day 4: Modelling sequential effects and reinforcement learning
Day 5: “Day off”: social activity and group projects
Day 6: Group-level analysis
Day 7: Model comparison and validation
Day 8: Modelling behavior with artificial neural networks
Covering the core topics in computational modelling of decision-making, including fundamentals of Bayesian analysis (model fitting and comparison, GLMs, latent models, hierarchical models, etc.); standard models of decision-making (optimal observer, drift-diffusion model, actor-critic model, etc.), as well as integrating neural data into behavioral models. An emphasis will be given to sound statistical interpretation of modelling analysis. Our focus is not on teaching the mathematical derivations; instead we will teach formal underpinnings only where they are critical for providing a conceptual understanding of the approaches described.
We will devote extensive time (around 4 hours a day) to implementing the analyses using Matlab/Python with actual behavioral data. Technical Assistants (TAs) will guide the trainees during all these tutorials.
Trainees will work in teams of three on a project involving data from their home laboratory. Trainees should arrive with behavioral data from their homelab research that they wish to analyze (synthetic datasets will also be available where appropriate). The trainees will apply the presented techniques to their own dataset, under supervision by TAs and faculty members. We expect the trainees to complete within the week a (likely partial yet) sound modelling analysis of their dataset.
Two keynote research talks on neuroscience topics where cutting edge modeling techniques are applied. Speakers will put emphasis on modelling challenges and how they were overcome, and how models of behavior illuminate neuroscience experiments.
One-to-one meetings with the faculty members
We plan one-to-one sessions of trainees and faculty providing the opportunity to talk in depth about their work, plans for the future, etc.