The First Simulation Card for Ethical AI Simulations

By: Stephan Zheng

We recently released Foundation, an open-source framework to build economic simulations. Foundation has been designed with flexibility and AI research in mind, and can be modified by anyone.

AI simulations offer researchers the power to generate data and evaluate outcomes of virtual economies that capture a part of the real world. Our research on the AI Economist showed the potential of using reinforcement learning to model more realistic economic agents in economic simulations and to achieve higher social welfare (i.e., improving both equality and productivity) with economic policies designed by AI.

To document and highlight the ethical dimensions, benefits and risks of simulations, we have released a Simulation Card, the first of its kind. Its purpose is to document the use, risks, and potential sources of bias in the published version of the simulation.

The objective of Simulation Cards is similar to that of Model Cards and Data Sheets. However, simulations differ crucially from trained models and datasets, as simulations are typically designed to create scenarios of interest. These scenarios can contain bias, which might be purposefully built-in or an unexpected side-effect of the design choices made during creation. Because simulations create many datasets of various shapes and sizes, the potential for misuse is greater than that of a single fixed dataset that might contain bias. An unethical simulation poses an order-of-magnitude larger ethical risk. As a result, our commitment to transparency is all that much more critical.

We encourage researchers and developers to publish similar Simulation Cards for software releases, to broadly promote transparency and the ethical use of simulation frameworks.