Date of Award

8-2026

Document Type

Thesis

Degree Name

Bachelor of Arts (BA)

Second Department

Natural Sciences

First Advisor

Gillman, David

Area of Concentration

Computer Science

Abstract

This thesis presents the development of MEIJI, an artificial intelligence (AI) agent that applies temporal difference (TD) learning to play Colonial Diplomacy within a custom reinforcement learning (RL) environment. While reinforcement learning has previously been applied to the board game Diplomacy, no reinforcement learning environment has been developed for the Colonial Diplomacy variant. To address this gap, this thesis implements a complete RL environment that models the rules and mechanics of Colonial Diplomacy and enables TD agents to learn through self-play. The project lays the groundwork for an investigation of whether the trained MEIJI agents learn to recognize the strategic importance of opposing Great Powers during gameplay. This is evaluated by comparing the number of supply centers controlled by an agent with those controlled by each opponent, thereby identifying which Great Power is closest to achieving victory. By analyzing the agents’ learned behavior, this investigation would examine immediate strategic threats in a complex multi-agent environment and contribute a new benchmark for reinforcement learning research in Colonial Diplomacy.

Rights

The author has granted New College of Florida the nonexclusive right to archive, make accessible, and distribute for educational purposes this work in whole or in part in all forms of media, now or hereafter known. The copyright of this work remains with the author.

Share

COinS