A MACHINE LEARNING APPROACH TO PREDICTING SATISFACTION IN NEGOTIATION

Abstract: In this paper, we examine whether negotiators can accurately predict their partners' satisfaction and explore how language used during negotiations can predict satisfaction. In Study 1, participants negotiated via online chat over a used car. Results showed that while negotiators achieved marginal accuracy in predicting their partners' satisfaction, they relied on their own satisfaction as a cue, despite it being an unreliable indicator. Using the Linguistic Inquiry and Word Count (LIWC) dictionary and ridge regression modeling on negotiation transcripts, we found that language patterns predict negotiator satisfaction more accurately than participant predictions or economic outcomes. Language associated with satisfaction centered around feeling and perception, while dissatisfaction correlated with offer rejection and competitive drives. Our findings offer a novel approach to understanding and predicting negotiation satisfaction through language analysis. Study 2, featuring in-person negotiations, is ongoing.

Keywords: negotiation, satisfaction, machine learning.