Artificial Intelligence and Technology-Based Interventions for Mental Health
1 - (OP14) An Autobiographical Approach with AI for Interpreting Personal Narratives: Using Synthetic Personas to Test the Timing of Cognitive Distortion Feedback
Friday, June 26, 2026
9:05 AM - 9:22 AM PDT
Location: Yerba Buena Salon 10, B3 Level
Keywords: Research Methods, Cognitive Processes, Cognitive Biases / Distortions Recommended Readings: Fischer, E. (2011). How to Practise Philosophy as Therapy: Philosophical Therapy and Therapeutic Philosophy. Metaphilosophy, 42(1–2), 49–82. https://doi.org/10.1111/j.1467-9973.2010.01680.x, de Miranda, L., & Loughlin, M. (2023). Philosophical health: Unveiling the patient’s personal philosophy with a person-centred method of dialogue. Journal of Evaluation in Clinical Practice, 29(7), 1161–1170. https://doi.org/10.1111/jep.13871, Chen, Z., Lu, Y., & Wang, W. (2023). Empowering Psychotherapy with Large Language Models: Cognitive Distortion Detection through Diagnosis of Thought Prompting. In H. Bouamor, J. Pino, & K. Bali (Eds.), Findings of the Association for Computational Linguistics: EMNLP 2023 (pp. 4295–4304). Association for Computational Linguistics. https://doi.org/10.18653/v1/2023.findings-emnlp.284, ,
PhD student Stanford University Palo Alto, Stanford, California, United States
Personal narratives of past experiences are shaped by individual views and cognitive appraisals. By examining how a person narrates an impactful negative experience, we understand how they are psychologically affected by such an experience and can identify their interpretations of it that shape their sense-making of the very experience. It is very common that personal narratives contain cognitive distortions – false beliefs that sabotage psychological well-being and prohibit individuals from forming helpful ways of learning, making sense of, and eventually healing from past experiences. In this study, we did an experiment with synthetic personas. We conducted an experiment with 210 synthetic personas who had gone through suicidal thoughts, and had them narrate and reflect upon a past challenging experience in life through phases of guidance with or without AI assistance of cognitive distortion reflections. In the four experimental conditions, we asked individual personas to narrate and reflect first, while introducing AI interventions to respond and inform them of the cognitive distortions they revealed in autobiographical narratives on different stages. We found that, AI personas that were only guided to complete the multiple-stage narrative therapy or were informed of distorted beliefs only after they had finished all stages of narrative, reported to perceive less psychological suffering and higher sense of meaning toward life than personas who were informed of their distorted beliefs halfway or right from the beginning of the narrative process. This shows promising signs in using autobiographical narrative therapy to assist individuals with reflections and reappraisals of past negative events while also informs researchers the necessity of selecting a right timing to break cognitive distortions to individuals who have maladaptive beliefs in their interpretation of lived experience. We propose that using generative AI agents assigned with specific synthetic personas facilitates scrutiny of experimental ideas and provides feedback that helps iterate treatment designs.
Learning Objectives:
Upon completion, participant will be able to get an idea of how one can develop experiments to test cognitive interventions on synthetic personas.