Computational Modeling of (un)Cooperation: The Role of Emotions

The philosopher H. P. Grice was the first to highlight the extent to which our ability to communicate effectively depends on speakers acting cooperatively. This tendency to cooperation in language use, recognized since Grice’s William James lectures, has been a key tenet of subsequent theorizing in pragmatics. Yet it’s also clear that there are limits to the extent to which people cooperate: theoretical and empirical studies of the Prisoner’s Dilemma have shown that people prefer to cooperate if the other party cooperates, but not otherwise. This would suggest that in language use, as well, the level of cooperation depends on the other person’s cooperativeness. So far, however, it has proven remarkably difficult to test such prediction, because it is difficult to analyze cooperation and communicative style objectively, and the schemes proposed so far for, e.g., non-verbal cues to cooperation tend to have low reliability. In this study the existence of a negative correlation between emotions and linguistic cooperation is demonstrated for the first time, thanks to newly developed methods for analyzing cooperation and facial expressions. The heart rate and facial expressions of the participants in a cooperative task were recorded after uses of cooperative and uncooperative language; facial expressions and the level of linguistic cooperation in each utterance were classified with high reliability. As predicted, very high negative correlations were observed between heart rate and cooperation, and the facial expressions were found to be highly predictive of her level of cooperation. Our results shed light on a crucial aspect of communication, and our methods may be usable to research in other aspects of human interaction as well.