New study sets a higher bar for AI-powered mentoring support

A new study accepted for publication in Translational Issues in Psychological Science (Hersch et al., in press). raises questions about how AI should be used to support mentoring relationships. In the study, “Understanding Undergraduates’ Attitudes Toward AI-Assisted Peer Mentoring: Balancing Authenticity with Efficiency” Hersch et al. (in press) surveyed 65 undergraduates at a public minority-serving institution. Students rated mentoring without AI-generated advice as more useful and more logical, and expressed greater confidence in recommending it to a friend.

Students’ reservations persisted even though the proposed AI would suggest answers for mentors to copy and personalize rather than respond directly to students. Nearly two-thirds wanted to know if their mentor used AI, and just over half said discovering that use from someone else would affect the relationship. Students were most comfortable with AI assistance for career questions and least comfortable with its use for mental health questions.

Other platforms emphasize AI-guided ice breakers and conversation prompts or chatbots that conduct automated check-ins. Such communication, however, leaves open the question of whether students receive guidance grounded in a person’s understanding and judgment. The authors propose a different emphasis. “Our findings suggest that a ‘human-at-the-helm’ approach, which maintains transparency and uses AI to reduce cognitive load and offer on-demand expert advice, rather than replace it, might be a more acceptable and effective approach than current companion chatbot models which tend to foster student-AI interaction at the expense of human relationships.”

MentorPRO’s human-at-the-helm, or HATH, approach makes human judgment and responsibility the starting point for AI design. In the mode,l described in the paper, mentors lead the conversation while AI helps organize information and identify relevant resources.