Exploring emotion modelling with ACT-R and the Iowa Gambling Task

Code for First Year Project.

An ACT-R agent plays the Iowa Gambling Task with a dual-route structure: a rational route that estimates each deck's value from experience, and an emotional modulator that scales the choice according to the losses each deck has produced. The modulator acts on the action, not on memory.

Files

• IGT.ipynb — notebook with the full analysis: sweeps, tables and figures
• igt_model.py — the task, the agents and the emotional modulator
• igt_analysis.py — running, scoring (P and Q indices) and plotting
• model.py, dmchunk.py — ACT-R declarative-memory implementation, from N. Taatgen's teaching materials, used unmodified

Reproducing the results

Running IGT.ipynb reproduces the sweeps and figures reported in the project. The random seed is fixed (np.random.seed(0)) before each sweep, so the reported runs are reproducible.

Requirements: Python 3 with numpy and matplotlib.