10.5061/DRYAD.BNZS7H4C4
Zheng, Stephan
0000-0002-7271-1616
Salesforce (United States)
Trott, Alexander
Salesforce (United States)
Srinivasa, Sunil
0000-0002-3974-0917
Salesforce (United States)
Parkes, David
Harvard University
Socher, Richard
You.com
The AI Economist: Taxation policy design via two-level deep reinforcement
learning
Dryad
dataset
2021
FOS: Computer and information sciences
Machine learning
2021-12-02T00:00:00Z
2021-12-02T00:00:00Z
en
https://arxiv.org/abs/2004.13332
https://doi.org/10.5281/zenodo.5644182
3085638846 bytes
7
CC0 1.0 Universal (CC0 1.0) Public Domain Dedication
This dataset contains all raw experimental data for the paper "The AI
Economist: Taxation Policy Design via Two-level Deep Multi-Agent
Reinforcement Learning". The accompanying simulation, reinforcement
learning, and data visualization code can be found
at https://github.com/salesforce/ai-economist. For the one-step economy
experiments, we provide: training histories, configuration files (these
experiments do not use phases), and final agent and planner models. For
the Gather-Trade-Build scenario, the data covers 6 spatial layouts: two
Open-Quadrant (with 4 and 10 agents), and four Split-World maps with
different configurations of the high-skilled and low-skilled agents. It
also covers 4 tax policies (the AI Economist, Saez, free-market, and US
federal). In addition, the AI Economist has been optimized for two social
welfare functions: the product of equality and productivity, and
inverse-income weighted utility. The Saez tax policy also uses estimated
elasticities. Each experiment was repeated with different random seeds:
10 seeds for the Open-Quadrant scenarios, and 5 seeds for the Split-World
scenarios. For each individual experiment, we provide: Training histories
(e.g. equality and productivity throughout training) the phase 1 and phase
2 configuration files, 40 episode dense logs (the final 10 simulation
logs across 4 environment replicas), phase 1 final agent models, and phase
2 final agent and planner models. Finally, we include all data and results
used to calibrate the Saez elasticity estimates and to estimate elasticity
directly from a sweep over flat-rate tax policies: training histories, the
phase 1 and phase 2 configuration files, phase 1 final agent models, and
phase 2 final agent and planner models.
This data has been generated by applying multi-agent deep reinforcement
learning in economic simulations. It contains all key reinforcement
learning and economic metrics that support the results in the paper. We
have included tutorials and configuration files (with the raw data itself)
on 2-level RL to generate this data. These complement the code available
in the following locations: The Github
repo https://github.com/salesforce/ai-economist has all the simulation and
reinforcement learning code that produced this data. We also provide all
the visualization and analysis code on Github. All figures in the paper
can be generated by using the visualization code on Github with this
experimental data. In addition, all code is also available on Zenodo
\url{https://doi.org/10.5281/zenodo.5644182}. To reproduce this data, we
have also provided instructions on Github to independently
run reinforcement learning and the economic simulations. This data has
only been organized for clarity and not otherwise modified.
Note: the 3 GB zip file inflates to ~20 GB. For questions:
stephan.zheng@salesforce.com or aieconomist@salesforce.com.