Prosocial or Selfish? Agents with different behaviors for Contract Negotiation using Reinforcement Learning

We present an effective technique for training deep learning agents capable\nof negotiating on a set of clauses in a contract agreement using a simple\ncommunication protocol. We use Multi Agent Reinforcement Learning to train both\nagents simultaneously as they negotiate with each other in the training\nenvironment. We also model selfish and prosocial behavior to varying degrees in\nthese agents. Empirical evidence is provided showing consistency in agent\nbehaviors. We further train a meta agent with a mixture of behaviors by\nlearning an ensemble of different models using reinforcement learning. Finally,\nto ascertain the deployability of the negotiating agents, we conducted\nexperiments pitting the trained agents against human players. Results\ndemonstrate that the agents are able to hold their own against human players,\noften emerging as winners in the negotiation. Our experiments demonstrate that\nthe meta agent is able to reasonably emulate human behavior.\n