Using Reinforcement Learning models to characterize human learning and decision making

The use of computational models to describe and characterize cognitive processes underlying human behaviour has been emerging as a useful paradigm beyond cognitive psychology. It is now used in cognitive neuroscience, computational psychiatry, neuroeconomics and AI, and has a potential to affect other fields as well. In this workshop we aim to discuss the general approach of fitting computational models to behaviour, specifically examining the use of reinforcement learning (RL) models to describe learning and decision making. 

Content

This workshop will include theoretical and hands-on parts. 

In the theoretical parts we will discuss model fitting to behaviour, and introduction to RL algorithms in human learning and behaviour, namely q-learning and Pearce-Hall. 

We will also carry hands-on fitting of RL models to human data from a learning and decision making task. We will use a pre-prepared dataset in R coding environment.

The data analysis will be carried using R, but can also be easily implemented in Matalb/Python as well. For the hands-on part we will assume background in programing language (Matlab/Python/R). A basic mathematical understanding is also required for better understanding of the operations within the model fitting procedures and RL models (logarithms, exponents, integration).

The workshop will take place over two days, with the first focused mostly on theory and the second on hands-on experience. The workshop will be given in Hebrew unless English speakers will be attending it.

Day 1 (9-14)

Lecture - Models and behaviour

Hands on - Fitting models to data, evaluating parameters, comparing models

Day 2 (9-14)

Lecture - Learning in humans and reinforcement learning models

Hands on - Building a RL algorithm and fitting it to behaviour