Summary

Innovating the training of peer counselors through advanced AI-driven multi-level feedback systems. The study designs a taxonomy with psychotherapy supervisors to evaluate emotional support conversations thoroughly.

• Contextualized feedback using LLMs.
• Public dataset with 400 emotional support conversations.
• Enhanced feedback generation reducing harmful or low-quality responses.

Significance

This method critically addresses the lack of experienced mentorship available for trainee peer counselors by providing structured, multi-tier feedback. It is particularly beneficial in scenarios where the scale and quality of training have to be matched with the increasing demands for mental health support services. ### Further Research By iterating and expanding the feedback mechanisms and their datasets, the project can significantly contribute to the quality of mental health services, ensuring a consistent and reliable training standard for future counselors.