Abstract:
Digital humans are a form of animated character that are starting to be used as conversational agents for health care and psychological therapy. The expansion of digital humans into these areas has great potential to extend the reach of health interventions. For digital humans to be acceptable to users, it is important that they foster trust and rapport. One factor that has been shown to impact rapport with a digital human is its eye gaze. Eye gaze is a socially salient communicative signal used in conversation to signal thinking, turn-taking and to modulate intimacy levels. Direct gaze as an affiliative signal can also influence perception formation. It is not known how these three forms of eye gaze together influence rapport with a digital human. This research had two aims. The first aim was to investigate how human dyads used these three forms of eye gaze when paired with a stranger in a close conversational task and how these patterns were associated with rapport. The second aim was to program similar gaze patterns into a digital human and investigate whether this improved rapport compared to a random eye gaze model and a static direct gaze model. It was hypothesised that the more realistic and humanlike gaze model would result in better social closeness, rapport, likeability and trust than the other gaze models. Two studies were conducted to meet these aims. The first observational study collected gaze and questionnaire data from 36 healthy female participants. Participants completed the Relationship Closeness Induction task in pairs. Analyses examined gazing behaviour as the intimacy of the questions increased with each set, and the relationship between gaze and four key psychological outcomes; social closeness, rapport, likability and trustworthiness, as well as how the interaction influenced levels of stress, loneliness, relaxation and support. The results showed that gazing behaviour varied as a result of changes in question intimacy level. Increased gazing towards the conversation partner was associated with increased positive perceptions of their partner. The interaction produced significant improvements in feelings of stress, loneliness, relaxation and support. In study 2, 72 female participants were recruited and randomised to complete the relationship closeness task with a digital human in one of the three gaze conditions. Study two produced no support for the primary hypotheses, with the results showing no significant differences between gaze condition on ratings of social closeness, rapport, likability and trustworthiness. Talking to the digital human was not as effective at reducing stress, loneliness, and boosting relaxation and support as talking to a real human. In conclusion, eye gaze patterns between people change as a function of the intimacy of a conversation. More eye gaze at a partner is associated with greater social closeness, trust and rapport. These results support theory and previous research. The different programmed gaze patterns of a digital human did not influence psychological outcomes. The lack of significant findings may have been due to issues with the digital human‘s base programming, which is a key limitation of this study. One implication of this research is that digital humans‘ gaze patterns could be altered to match the intimacy level of conversations to more closely match human gaze patterns. It is possible that gazing behaviour that more closely reflects human gaze could improve rapport, but more research is needed to create and test such detailed patterns in a robust way.