Elnaz Nouri, Kallirroi Georgila and David Traum: “A Cultural Decision-Making Model for Negotiation Based on Inverse Reinforcement Learning”

August 1, 2012 | Sapporo, Japan

Speaker: Elnaz Nouri, Kallirroi Georgila and David Traum
Host: CogSci 2012

Abstract: We learn culture-specific weights for a multi-attribute model of decision-making in negotiation, using Inverse Reinforcement Learning (IRL). The model takes into account multiple indi- vidual and social factors for evaluating the available choices in a decision set, and attempts to account for observed be- havior differences across cultures by the different weights that members of those cultures place on each factor. We apply this model to the Ultimatum Game (a well-known simple negoti- ation game) and show that weights learned from IRL surpass both a simple baseline with random weights, and a high base- line considering only one factor of maximizing gain in own wealth in accounting for the behavior of human players from four different cultures. We also show that the weights learned with our model for one culture outperform weights learned for other cultures when playing against opponents of the first cul- ture. We conclude that decision-making in negotiation is a complex, culture-specific process that cannot be explained just by the notion of maximizing one’s own utility, but which can be learned using IRL techniques.