Exploring the impact of frequency of robot's proactive utterances in product introduction dialogues

This study investigates the impact of the frequency of proactive utterances by interactive agents in product introduction dialogues. Proactive interactive services, unlike traditional reactive services, actively provide information based on user states and situations. This research aims to understand how different frequencies of proactive behaviors affect user evaluation, objective attainment, and the impression of the agent. We developed an interactive agent system capable of delivering proactive utterances at two different frequencies: low and high. The low-frequency condition involved the agent making proactive utterances following its own reactive responses, while the high-frequency condition included additional proactive utterances during silent periods in the dialogue. The experiment compared these two conditions in terms of dialogue duration, information conveyed, user participation, and user impressions. Results indicated that high-frequency proactive utterances led to higher agent proactivity but reduced user proactivity. Moreover, the high-frequency also improved the perceived speaking skills and animacy for the agent. The detailed analysis of high-frequency condition implied that moderate frequency of proactive behaviors are rated higher by users than extremes frequency. These findings highlight the importance of balancing proactive behavior to enhance interactive service experiences without overwhelming the user. Future research should explore these dynamics in more diverse and naturalistic settings to further validate and expand upon these results.