A proof of concept study on the use of large language models as a client in typed role plays for training therapists
Year: 2024
Type: article
Abstract: In cognitive behavioural therapy (CBT) training, case vignettes are used for discussion exercises and provide a brief for behavioural role-plays between peers. Some challenges implementing role-plays may include participants' limited knowledge, skills and confidence, which could affect performance and learning for the trainee in the role of the therapist. Large Language Models produce naturalistic language that has yet to be tested as an alternative to traditional role-play in training. The present study was a proof of concept investigation into the interactive capabilities of Large Language Models (LLM) powered vignette in a text-based role-play. The study aimed to demonstrate appropriate text-based conversational responses of a LLM in the role of a depressed client receiving a CBT intervention. ChatGPT 4 was engineered with prompts ensuring that the LLM consistently behaved as a client presenting with depression. During a text based role-play the author took the role of the CBT Therapist and the LLM as the client presenting with depression receiving a cognitive intervention. The LLM was able to interact and provide responses during the text-based role-play and complete a cognitive intervention for depression. The study provides evidence that the LLM is capable of text-based conversations in the role of a depressed client to work through a cognitive intervention. These findings highlight the potential for LLM text-based role-plays in training of CBT Therapists. The practical implications of these findings in education are discussed and the potential for text-based role plays in meeting diverse learning needs are considered.
Source: Discover Psychology
Authors: Leah Fung, Rory Laing
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $1,490
OpenAlex ID: W4405550450