Psychiatr Q. 2025 Oct 28. doi: 10.1007/s11126-025-10231-w. Online ahead of print.
ABSTRACT
Case formulation (CF) is central to personalized mental health care, yet little is known about how artificial intelligence (AI) may simulate theory-informed processes. This exploratory study examined how ChatGPT-4.0 generated CFs for adolescent gaming disorder using cognitive-behavioral and psychodynamic frameworks. Eight standardized fictional vignettes describing demographics, gaming behaviors, psychiatric symptoms, and family context were submitted with prompts requesting framework-specific formulations. Outputs underwent thematic analysis with structured parallel frameworks and reflexive coding. Cognitive-behavioral formulations emphasized schemas, distortions, avoidance, gaming’s psychological functions, and motivational themes, yielding three exploratory subtypes: Avoidant-Anxious, Defiant-Externalizing, and Depression-Driven. Psychodynamic formulations highlighted intrapsychic conflict, defense mechanisms, relational templates, and symbolic meanings, producing four subtypes: Shame-Regulating, Grief-Avoidant, Inhibited/Anxious-Avoidant, and Control-Oriented. Across frameworks, convergences emerged around low self-worth, avoidance, family dysfunction, and gaming as emotional regulation. These findings suggest that large language models can approximate framework-based case formulations and highlight clinically relevant themes, though they are not generalizable beyond simulated cases. With ethical oversight, such tools may support integrative clinical thinking, education, and reflective supervision. Future work should compare AI- and clinician-generated formulations with real patient data to evaluate validity and utility. Clinical trial number: Not applicable.
PMID:41148434 | DOI:10.1007/s11126-025-10231-w
