Preference learning from emotional expressions contributes integrative solutions between human-AI negotiation

Learning about the counterpart, especially their preference and limits, through pre-play communication is crucial for reaching a win-win solution in negotiations. Although emotional expressions are considered to be honest and credible signals in human-human negotiations, it is still unclear how emotional expressions encoding preference and limits affect the outcome of human-AI negotiations. In the present study, we designed that preferences learning phase of counterpart AI agent from emotional expressions in pre-play communication and four-issue ultimatum game with an AI agent. Participants are going to experience either one phase or two phases of pre-play communication before deciding on their final offer. We hypothesized that participants who experienced two phases, i.e., the preference learning phase in addition to a phase in which simultaneous inference of preferences and limits was possible, were able to reach more integrative solutions than those who did not.