Reasoning About Mutable Rule Sets

This project is concerned with the research and development of intelligent agents that can reason about mutable rule sets. Agents are discrete, autonomous software systems that act in the context of an environment, which often contains multiple agents. The environment in which an agent exists is defined by sets of implicit and explicit rules. By making these rules explicit and providing a framework within which they can be changed, this project creates a software system that allows for mutable rule sets to be examined. The agents in this project use reflective modeling to reason about the effects of any changes to the rules that define their environment. Reflective modeling is the process of reflecting the state of the environment in which the agent exists into a separate, isolated model of that state. These isolated models mirror the characteristics of the original environment, including representations of the agents within it. By exploring the changes that occur through manipulating these isolated state models, agents can project how specific rule changes will affect their own local state. The capabilities of these agents and their intelligence are explored in the context of Peter Suber’s game of Nomic. The game of Nomic is a thought experiment where making a move in the game changes the rules that define Nomic. This project implements several agent behaviors which compete to accomplish differing objectives within the game of Nomic and demonstrate their ability to reason and make intelligent decisions regarding rule changes which benefit their internal objectives.