Conversational agents have the potential to support intergroup relations when psychological or linguistic barriers prevent direct interaction. Based on intergroup contact theory, we propose GroupEnvoy, a text-based conversational agent that represents outgroup perspectives during ingroup discussions. Its dialogue is grounded in data from a prior outgroup-only discussion. To evaluate this approach and derive design principles, we conducted a mixed-methods, between-subjects study with university students, in which host-country students formed the ingroup and international students formed the outgroup. Ingroup students performed a collaborative task while engaging with outgroup perspectives, either by interacting with GroupEnvoy (AI-mediated contact) or by reading a static document (passive exposure). Quantitatively, AI-mediated contact demonstrated a directional reduction in intergroup anxiety and an improvement in perspective-taking. Qualitatively, AI-mediated contact enhanced outcome expectancies and directed empathy toward the outgroup's evaluations of the ingroup, whereas passive exposure fostered future contact intentions and elicited empathy toward the outgroup's lived experiences. These findings present AI-mediated contact as a promising paradigm for improving intergroup relations.

翻译：摘要：当心理或语言障碍阻碍直接互动时，对话代理有潜力支持群际关系。基于群际接触理论，我们提出GroupEnvoy，一种基于文本的对话代理，在内群体讨论中代表外群体视角。其对话基于先前仅限外群体讨论的数据。为评估这一方法并推导设计原则，我们针对大学生开展了一项混合方法、组间设计研究，其中东道国学生构成内群体，国际学生构成外群体。内群体学生在执行协作任务的同时接触外群体视角，方式包括与GroupEnvoy互动（AI介导接触）或阅读静态文档（被动接触）。定量分析显示，AI介导接触在减少群际焦虑和提升观点采择方面呈现方向性改善。定性分析表明，AI介导接触增强了结果预期，并引导共情指向外群体对内群体的评价，而被动接触则促进了未来接触意愿，并引发对外群体生活经历的共情。这些发现将AI介导接触确立为改善群际关系的一种有前景的范式。