A Case for Leveraging Generative AI to Expand and Enhance Training in the Provision of Mental Health Services

TL;DR

This paper argues that generative AI should be leveraged to expand and enhance training in mental health service provision rather than directly delivering care via therapist Chatbots. It advocates a human-in-the-loop approach using Retrieval-Augmented Generation to create scalable, interactive training with simulated clients, real-time feedback, and competency assessment. A case study of the HomeTeam program with veterans demonstrates scalable, diverse, and low-risk practice that can improve clinician readiness and throughput. The work highlights the need for accuracy, cultural responsiveness, and benchmarking, while outlining significant potential to close the global mental health treatment gap through enhanced training infrastructure.

Abstract

Generative artificial intelligence (Generative AI) is transforming healthcare. With this evolution comes optimism regarding the impact it will have on mental health, as well as concern regarding the risks that come with generative AI operating in the mental health domain. Much of the investment in, and academic and public discourse about, AI-powered solutions for mental health has focused on therapist chatbots. Despite the common assumption that chatbots will be the most impactful application of GenAI to mental health, we make the case here for a lower-risk, high impact use case: leveraging generative AI to enhance and scale training in mental health service provision. We highlight key benefits of using generative AI to help train people to provide mental health services and present a real-world case study in which generative AI improved the training of veterans to support one another's mental health. With numerous potential applications of generative AI in mental health, we illustrate why we should invest in using generative AI to support training people in mental health service provision.

Figures (7)

• Figure 1: By using AI to deliver personalized didactic training, to generate simulated clients for practice, and to provide feedback on learning and performance during training, there is potential for a single clinician supervisor to focus their full effort on supervision and extend their reach to a greater number of trainees. This figure was made using generative AI.
• Figure 2: Specific client simulations can be developed to support specific types of skill acquisition or competency development. As one example, a trainee hoping to improve their skills in motivational interviewing could practice with a simulated client experiencing ambivalence about a behavior change, while receiving feedback on their motivational interviewing skills and strengthening competence over time. This figure was made using generative AI.
• Figure 3: Generative AI enables the development of a diverse set of simulated clients experiencing a similar presenting concern (e.g., suicide risk), but having different cultural backgrounds or life experiences. These approaches give trainees opportunities to strengthen their cultural competence and practice intervention skills in more challenging or higher risk contexts without risk to human clients. This figure was made using generative AI.
• Figure 4: Generative AI allows for real time, in-the-moment feedback similar to bug-in-the-ear and bug-in-the-eye technologies. This figure was made using generative AI.
• Figure 5: Generative AI could be used to summarize strengths and areas of improvement for trainees and their clinician supervisors, track changes in those areas over time, and identify specific examples (e.g., video clips) that represent those areas. This could not only save trainees and clinician supervisors the time needed to identify these strengths and weaknesses and scroll through video recordings of sessions to find examples, but ultimately may improve the quality of that supervision and clinical care. This figure was made using generative AI.