Simulated Forensic Role Interactions Using GPT-4: A Psycholinguistic Dataset

This dataset comprises 30 simulated conversational sessions generated by GPT-4 (OpenAI, March 2024 release), each designed to emulate one of three role types commonly observed in forensic contexts: Analyst, Friend, and Witness. Prompts were systematically crafted to evoke role-consistent psycholinguistic behaviors and were used to test the applicability of generative AI for profiling and training in investigative scenarios. For each session, role-specific prompts, AI-generated responses, model metadata, and annotated linguistic features are recorded. This dataset supports the reproducibility and transparency of the findings presented in the article "Generative AI in Forensic Role Simulations as a Human-Computer Interaction Phenomenon".