Modeling Human Adaptation in a Non-Stationary Two-Armed Bandit Task

This project compares a Basic Rescorla-Wagner model against a Complex Curiosity-Based model to explore how humans rapidly adapt in a non-stationary two-armed bandit task. The complex model tests the hypothesis that humans re-compute the utility of non-chosen actions using knowledge about the environment's hidden structure. Model comparison favors the complex model, suggesting humans leverage counterfactual curiosity for faster adaptation.

Report

Full Report

---

Running the Project

To run this project locally:

1. Install uv (https://docs.astral.sh/uv/getting-started/installation/)

2. Then run uv sync

3. Run the Jupyter notebook as usual, select the created venv as the Python interpreter.