Date of Award
Spring 5-16-2026
Document Type
Honors Project
University Scholars Director
Dr. Joshua Tom
First Advisor/Committee Member
Dr. Carlos Arias
Second Advisor/Committee Member
Dr. Phillip Baker
Keywords
emergent behavior, multi-agent, reinforcement learning, AI, artificial life, game theory
Abstract
Reinforcement learning (RL) algorithms can train agents to solve problems in environments using complex behaviors that are not explicitly programmed, known as emergent behaviors. The goal of our research is to investigate how different RL reward values influence the emergence of competitive and cooperative behaviors in games with teams of multiple agents. Specifically, we focus on general-sum games, in which the sum of gains and losses of each team may be non-zero, allowing situations for agents to mutually benefit or mutually fail. Using Unity’s ML-Agents Toolkit to train agents with RL self-play in bounded 2D environments, we identify high-level behaviors such as fighting other agents and mining asteroids for resources. We demonstrate that the competitive behavior of agent fighting can be decreased by increasing the agent death penalty or by increasing asteroid reward values, leading to a greater number of agents alive at the end of episodes. However, we are unable to completely prevent agents from fighting due to their short-term reward-seeking. Our findings provide insight into how emergent behaviors arise in games that are not purely competitive—like many real-life environments—and how these behaviors can be shaped through different RL reward values.
Recommended Citation
Erickson, Ethan F., "Shaping Emergent Competitive and Cooperative Behaviors in Multi-Agent General-Sum Games" (2026). Honors Projects. 256.
https://digitalcommons.spu.edu/honorsprojects/256
Copyright Status
http://rightsstatements.org/vocab/InC/1.0/
Additional Rights Information
Copyright held by author.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Included in
Comments
A project submitted in partial fulfillment of the requirements of the University Scholars Honors Program 2026.