Effects of self-disclosure on attributions in human-iot conversational agent interaction

Abstract

This study examined effects of self-disclosure on relationship closeness with an Internet of Things (IoT) conversational agent (IoT-CA) and attributions of responsibility. Participants and IoT-CAs worked as dyads for two interdependent-outcome tasks: A creativity and a learning task (for measuring dyadic creativity ability and IoT-CAs' understanding of people's preferences, respectively). Dyadic success or failure feedback was determined. Results showed in contrast to self-serving bias (SSB), people did not credit personal responsibility for dyadic success or blame the IoT-CA for dyadic failure. However, when people had previously engaged in self-disclosure with IoT-CAs, they showed reversed SSB, and tended to attribute success to the IoT-CA and accept more personal responsibility for failure. The effect of self-disclosure on attributions of responsibility was mediated by closeness of the relationship. In terms of attributions of responsibility between tasks, people who engaged in self-disclosure with IoT-CAs believed IoT-CAs understood them more and were more likely to attribute success to the learning task. RESEARCH HIGHLIGHTS: Voice interaction is becoming the mainstream mode of human-computer interaction. Intimate self-disclosure effectively induced closeness between IoT-CAs and individuals. People may tolerate and even accept imperfection in their personal agent. Participants' attribution style may affect their SSB when interacting with computer.