This repo contains the code used to obtain the results from the paper: Algorithmic Collusion: An Experimental Study of Firms' CSR Investment Decisions. This paper researches algorithmic collusion of Artificial Intelligence Algorithms in a duopoly setting where firms let these algorithms completely decide their production quantity and investment in Corporate Social Responsibility.
-
All code can be found in /src:
-
/analysis_util: contains a cycle classifier and functions used for plotting and to get the statistics presented in the paper.
-
/classes: contains the Q-learning, DQN and regulator agents, as well as the economic environment and the action class.
-
/runs: contains the analysis of different runs and shows the figures presented in the paper.
-
algorithms.py: contains the algorithms used to simulate episodes in the different settings discussed in the paper.
-
Install the dependencies with the following command:
pip install -r requirements.txt
To use an algorithm just navigate to one of the files in /runs and try the example or something else. Note that most plots from the paper require the simulation data available at Google Drive. When you use simulate_episodes an h5 file called 'simulation_data' should be created or placed inside the /data directory.
MIT