Answers are grounded in this paper and cite their sources.
Advances in artificial intelligence make it possible to embed AI facilitation agents in collaborative settings, where they can deliver real-time recommendations based on ongoing group discussions. This experimental study examined how group members perceive a collaborative platform with or without such an agent and evaluated the agent’s effects on group communication and performance. A total of 187 undergraduate psychology students collaborated in 50 groups of three or four members on a digital platform with a synchronous text chat. Groups were randomly assigned to one of two conditions (25 groups per condition): with an AI facilitation agent (n = 96 participants) that encouraged information sharing when exchanges were insufficient, or without the agent (n = 91 participants). Participants rated the platform’s acceptability, perceived support, sense of intrusion, and affective reactions. Participants in the AI condition rated the platform as less acceptable and less supportive, but not more intrusive or unpleasant. Nevertheless, groups assisted by the agent were more efficient, completing the task faster with less communicative effort. Descriptive analyses of group interactions in the three minutes before and after the agent’s interventions showed a shift from providing information toward requesting information, expressing opinions, and agreeing, without change in overall interaction volume. These findings reveal a gap between objective performance and subjective perceptions, highlighting the need to design AI agents that improve group efficiency while fostering members’ sense of support and acceptance.
No takes yet. Share an insight, caveat, or question.
Michinov et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: