Abstract
Recent advancements in AI have led to chatbots, such as ChatGPT, capable of providing therapeutic responses. Early research evaluating chatbots' ability to provide relationship advice and single-session relationship interventions has showed that both laypeople and relationship therapists rate them high on attributed such as empathy and helpfulness. In the present study, 20 participants engaged in single-session relationship intervention with ChatGPT and were interviewed about their experiences. We evaluated the performance of ChatGPT comprising of technical outcomes such as error rate and linguistic accuracy and therapeutic quality such as empathy and therapeutic questioning. The interviews were analysed using reflexive thematic analysis which generated four themes: light at the end of the tunnel; clearing the fog; clinical skills; and therapeutic setting. The analyses of technical and feasibility outcomes, as coded by researchers and perceived by users, show ChatGPT provides realistic single-session intervention with it consistently rated highly on attributes such as therapeutic skills, human-likeness, exploration, and useability, and providing clarity and next steps for users’ relationship problem. Limitations include a poor assessment of risk and reaching collaborative solutions with the participant. This study extends on AI acceptance theories and highlights the potential capabilities of ChatGPT in providing relationship advice and support.
•Users interacted with GPT-4 in a single session intervention for relationship difficulties.•GPT-4 provided realistic single-session intervention that was rated high on technical outcomes.•Most users described GPT-4 as human-like, helpful, and empathic.