Learning in Public Goods Games with Non-Linear Utilities: a Multi-Objective Approach

Nicole Orzan, Erman Acar, Davide Grossi, Roxana Rădulescu

Keywords: Multi-Objective Reinforcement Learning, Public Goods Games, Non-Linear Utility Function

Abstract: Addressing the question of how to achieve optimal decision-making under risk and uncertainty is crucial to both understanding human decision-making processes, and enhancing the capabilities of artificial agents that collaborate with or support humans. In this work, we address this question in the context of Public Goods Games. We study learning in a novel extended version of the Public Goods Game where agents have different risk preferences, by means of multi-objective reinforcement learning. We introduce a parametric non-linear utility function to model risk preferences at the level of individual agents. These attitudes are represented as preferences over the rewards received from the game. We study the interplay between such preference modeling and environmental uncertainty, which is constructed as noise over the level of incentive alignment in the game the agents play. We observe that different combinations of individual preferences and environmental uncertainties sustain the emergence of cooperative patterns in non-cooperative environments (i.e., where competitive strategies are dominant), while others sustain competitive patterns in cooperative environments (i.e., where cooperative strategies are dominant).

Type Of Paper: Work-in-progress paper (max page 6)

Anonymous Submission: Anonymized submission.

Submission Number: 15