Harhen-Bornstein-2023-Overharvesting-as-Rational-Learning

Please feel free to get in touch ([email protected]) if you plan on using/adapting the task or model! I'm happy to walkthrough the code or answer any questions.

Experiment

This contains the code for running the task and for analyzing the data.

Important Files

• experiment/run_exp/static/js/task.js This contains the "meat" of the task code and includes the functions for the different trial types.
• experiment/run_exp/static/exp_struc/10017437_best.js This defines the experiment structure including the order of planet types observed by all participants. Important to note: we only use the struc variable not r_0 (initial planet richness) and decay (decay rate). The initial planet richness and decay rates are drawn from distributions as necessary.
• experiment/data_analysis/analyze_data.ipynb Statistical tests are run and plots generated within this jupyter notebook.
• experiment/data_analysis/data_combiner.py Data_cleaning functions.

Model

This contains the code for fitting the models (Bayesian Structure Learning model, Marginal Value Theorem + learning, and Temporal-Difference Learning) to participant's choice data and performing 10 fold cross validation.

Important Files

• model/fit_by_planet.jl Running this file will fit a subject's choice data to one of the three models (example command: julia fit_by_planet.jl 10 0, to fit the Bayesian structure learning model to subject number 10's data).
• model/10_fold_cv.jl Running this file will compute the 10-fold cross validation for a subject for a given model (example command: julia 10_fold_cv.jl 10 0, to compute the cross-validation score for subject number 10 with the Bayesian structure learning model).
• model/plot_modeling_results.ipynb This contains the code for generating the model plots. This uses the parameters in fit_params folder and the cv scores in the cv folder.