Utilitarian Minecraft Agent
Ongoing •
Updated today
This is my ongoing undergraduate research project at Carleton University. I am running an experiment investigating the optimal amount of time to think over a plan. The experiment has three agents. One which generates a plan and executes it without further evaluation, one which evaluates a little, and one agent which evaluates and replans until it generates a plan which scores really well in the evaluation.
Throughout each agent's play the session, the world utility (sum of all entities health / max health * moral weight) is calculated at a set interval. The question is which of these three agents will produce the most utility within the "confines" of Minecraft
Project Entries
- Starting the Overleaf Document • 2026-02-27
- Describing Minecraft Worlds • 2026-02-02
- Omnipotent Minecraft AI Loop • 2025-11-15
- Only Text of MineDojo Wiki Dataset • 2025-11-10
- Voyager AI Package Dependency Errors • 2025-10-11
- Villager Antagonistic Mobs • 2025-10-05
- Utilitarian Ethics in Minecraft • 2025-10-04