Text-messaging with a conversational AI as an out-group member: Preliminary evidence for a new application of intergroup contact.

The application of artificial intelligence (AI) for electronic contact (e-contact) offers new possibilities for intergroup contact. This pilot study examined whether exchanging instant messages with an AI-generated out-group member reduces prejudice, using explicit and implicit pre- and post-contact evaluations with a pure control condition. The contact was designed in accordance with Allport's optimal contact conditions and strategies derived from categorization theories. Sixty-one Japanese adults participated in this study. One month after the pre-evaluation, participants in the contact condition engaged in a 15-min interaction with a text-based AI designed to represent a Chinese individual, during which they introduced themselves and discussed possible joint efforts toward achieving the Sustainable Development Goals. The results showed that the contact condition increased explicit measures of positive attitudes and intergroup empathy while reducing negative attitudes relative to the control condition. However, intergroup anxiety and implicit attitudes measured using the Implicit Association Test showed no contact effects. Our findings suggest that a brief, semi-structured contact involving personalized interaction and cooperation task may mitigate some aspects of explicit intergroup bias. This preliminary evidence indicates new possibilities for applying conversational AI in instant messaging environments within intergroup contexts and provides implications for effective e-contact strategies.