Exploring conflict-related behavioral intentions in AI and human role-playing interventions: evidence from behavioral experiments and linguistic analyses.

Interpersonal conflict often impairs decision-making and communication. This study examined whether AI-enabled role-playing (AI-RP) can regulate behavioral intentions in conflict situations and how it compares with the human counsellor-guided role-playing. Two experiments and a linguistic analysis (LIWC and BERTopic) were conducted. Participants were assigned to AI, the human counsellor, or no-interaction conditions. Behavioral intentions and interaction texts were analyzed. Both AI- and human-guided role-playing reduced revenge intentions and increased understanding compared with the control condition. These effects were context-dependent and did not transfer to parallel scenarios. The AI and human counsellor conditions did not differ in revenge or understanding outcomes, but the AI condition was associated with higher willingness to communicate in the specific context of Study 2a. Linguistic analyses showed AI responses were more structured and function-oriented, whereas the human counsellor responses emphasized emotional validation. AI-RP and the human counseling play complementary roles in conflict intervention. AI may facilitate rapid de-escalation and communication initiation, whereas in this study the human counsellor's responses were more oriented toward emotional validation and relational repair.