Game Theory and Agent Reasoning (II)

Building an agent that can reason the Prisoner's Dilemma

Over the last couple of weeks, I have tested a variety of different agent frameworks, approaches, and LLMs to build an agent to play a few rounds of Prisoner’s Dilemma with me. In this post, I will expand on my assumptions from my research work into Game Theory and Agent Reasoning. You can expect in this post the technical requirements I needed to cover, my learnings from implementing them, and my next steps in evaluating the quality of my work.

More importantly, to start with the bottom line up front, the first iteration of the ReAct agent that I want to use is complete. However, I have not run a full evaluation yet. I plan to make this part III of this Series.

But I am getting ahead of myself.