abm

A simple reinforcement learning model for the game of Farolfi & Erdlenbruch, 2020, "A classroom experimental game to improve the understanding of asymmetric common-pool resource dilemmas in irrigation water management". Developed at the 2022 BESLab Summer School for Computational and Experimental Economics at Universitat Pompeu Fabra. Team members: @GabrielBayle, @gitcombat, @mrpg.

To run the central simulation,

python3 -i abm.py 0.05 20 1000 5000

The 1000 denotes the number of replications (the "outer loop") and 5000 the number of rounds (the "inner loop"). You may wish to reduce the number of replications. A random seed can be specified (our own seed is in seed.txt). To see usage,

python3 abm.py --help

Then, e.g.,

mean(welfares(rounds))

abm_profits.py outputs profits per round and replication. abm_stoch.py introduces stochasticity of the initial inflow. abm_strategies.py extracts the gameplay from an additional 5,001st round.

There are also very simple tests in simple/.

Output from June 2022 is available in data/2022-06-18.tar.bz2 that can be extracted with tar or 7-Zip. You can reproduce our plots using data/plots.R. You can also reproduce all data using the shell scripts in experiment?.sh. Except for floating-point specifics, everything should be identical to our runs.