AI personas as therapist-supervisors can support therapist training and research

The article examines using AI personas to create simulated therapist-supervisors for therapist training and psychological research. It notes potential benefits alongside risks such as prompt drift and AI hallucinations.

The article reviews the use of AI personas to create simulated therapist-supervisors for training and research in psychology. It explains that modern generative AI and large language models can be prompted to adopt supervisory personas. The author describes practical applications such as role-play with AI clients and later supervisory feedback from another AI persona. The piece also notes caveats and provides a checklist and prompt examples for shaping supervisor behavior. Key points: - AI personas can be invoked with prompts in contemporary large language models and used to simulate therapist-supervisors for training and research. - Common applications include practicing with AI clients, having an AI provide real-time supervisory commentary, and replaying sessions for analysis by another AI persona. - The author offers prompt examples and a checklist of twelve characteristics to specify supervisor style and behavior. - The article highlights risks such as prompt drift, AI confabulation or hallucination, and the chance that simulations could be treated like games rather than clinical learning. - The author stresses that AI personas are an added tool and not a replacement for human-to-human training and supervision. Summary: AI personas could broaden options for supervised practice and experimental study in clinical training while also raising reliability and ethical concerns for how simulations are used. Undetermined at this time.