The project investigates how participants learn to categorize alien stimuli based on visual features in a trial-by-trial learning task. We implement the Generalized Context Model (GCM) to simulate categorization behavior and compare model predictions against both simulated and empirical data.
data/→ empirical and simulated datasetsstan/→ model definitionsworkbook.Rmd→ main analysis script
We model categorization using the Generalized Context Model, an exemplar-based model where:
- Previous stimuli are stored in memory
- Similarity between current and past stimuli is calculated
- Feature attention weights determine how strongly each feature influences categorization
- A sensitivity parameter controls how quickly similarity decreases with distance
- A bias parameter captures baseline response tendencies
The model is implemented in STAN for Bayesian parameter estimation.
This repository includes:
- Simulation of artificial agents under known parameter settings
- Prior predictive checks
- Posterior parameter recovery
- Posterior predictive checks
- Model fitting on empirical participant data
Group 2: Barbora, Daniel, Mattis, Niels, Søren