Supporting Bystander Intervention Through Conversational AI: Experimental Evidence from Undergraduate Samples

Bystanders play a critical role in supporting victims in situations of harm. Across campuses and workplaces, bystander intervention training aids individuals in developing effective intervention strategies. However, people’s reluctance to practice and intervene when needed remains a persistent challenge. One potential solution is to provide a safe, judgment-free space where people can repeatedly practice skills and receive personalized feedback. Here we test a chatbot-based bystander intervention practice tool, where an AI agent guides users through evidence-based training. Across two experiments with undergraduates, one conducted on a local college campus and the other online, we found that chatbot-guided training improved participants’ bystander efficacy and attitudes, with positive trends for measures of psychological safety and norm perceptions. To our knowledge, this research provides the first empirical evidence of the effectiveness of a chatbot in bystander intervention training. These findings reinforce the potential of AI as a uniquely useful pedagogical tool for promoting prosocial motivation, and, more broadly, this work contributes to efforts to cultivate safer and more inclusive social environments.