Metacognitive Agent for Training Negotiation Skills

Training negotiators remains a difficult and expensive proposition. Negotiators require complex cognitive skills such as theory-of-mind to be successful, but these skills can be difficult to train and measure. Here we present an agent designed to model theory-of-mind for learners and serve as a practice partner for complex negotiations. This agent employs instancebased learning to make decisions about its own actions and to reflect on the behavior of the opponent. This reflection process is used to provide a source of explicit feedback on the opponent’s strategy and behavior. In this paper we present evidence that the model is a plausible opponent for students learning negotiation. It is expected that practicing with this agent will improve theory-of-mind abilities in learners and, in turn, improve negotiation performance.