Knowledge Transfer Portrayals of Mentally Disordered Offenders: A Large Language Model (LLM)-Assisted Thematic Analysis of Public Forums

The portrayal of mentally disordered offenders (MDOs) in public forums represents a critical vector for informal knowledge transfer, profoundly shaping societal attitudes, stigma, and policy discourse. This study investigates the mechanisms of this digital knowledge transfer by applying an advanced, large language model (LLM) assisted thematic analysis methodology to a large-scale hypothetical dataset of public forum discussions. By integrating multi-agent supervised fine-tuning and human-in-the-loop validation, we identify recurring themes in how laypersons and professionals exchange, distort, or clarify information regarding MDOs. We tackle major unanswered questions in the field, particularly the reduction of algorithmic bias in qualitative coding and the preservation of research interpretive authority when using generative artificial intelligence. We eventually suggest important future research directions, such as the investigation of longitudinal changes in MDO language and the use of hierarchical Bayesian modelling to follow information dissemination throughout several online ecosystems.