Forensic Sciences
[Abbr: Acd. Jr. AJFSc]Predicting Unconscious Violence: Behavioral Analysis and Threat Assessment
by Haque, Bidisha
This study proposes and evaluates the Unconscious Violence Risk Index (UVRI), a novel threat assessment model combining behavioral and environmental indicators with machine learning to predict violent acts committed without conscious intent. Unconscious violence—aggressive acts precipitated by acute intoxication, medical conditions, or traumatic stress—is poorly captured by traditional tools. Despite extensive use of instruments like the Historical-Clinical-Risk Management-20 (HCR-20) and the Psychopathy Checklist–Revised (PCL-R) in violence risk assessment ("Structured Professional Judgment Tools"), existing methods focus on conscious intent and static traits (e.g. past violence, psychopathy) and often ignore dynamic emotional or situational cues (Ling et al. 55). In response, we developed the UVRI to integrate indicators such as acute distress, physiological dysregulation, and social isolation. In a mixed-methods study (N≈300 forensic/psychiatric clients), coded behavioral interviews and records were used to train a supervised ML model. The UVRI demonstrated superior predictive validity (AUC≈0.85) compared to HCR-20 (≈0.70) and PCL-R (≈0.65) in classifying risk of unconscious violent episodes. Sensitivity and specificity exceeded 0.80. Key predictive features included emotional dysregulation, trauma history, and contextual stressors, aligning with findings that emotion regulation deficits mediate stress-related aggression (Herts et al. 1111). Qualitative interviews revealed themes of sudden loss of control and unintentional harm (e.g. "I didn't even know I was hurting anyone"), underscoring complex subjective experiences. These results suggest that UVRI's fusion of behavioral analysis and ML enhances early detection of latent violence risk, with potential to improve preventive interventions. Implications for forensic practice, ethical considerations of automated risk prediction, and avenues for refining dynamic threat assessment are discussed.
AI-Assisted Psychodrama for Emotional Mapping in Offenders with Antisocial and Borderline Traits: A Mixed-Methods Pilot Study
by Haque, Bidisha
Understanding implicit emotional processes in individuals with antisocial or borderline traits remains a central challenge in forensic psychology. This mixed-methods pilot study evaluated an integrative framework combining psychodrama-based experiential therapy with AI-driven behavioral analytics in forensic and community samples. Eighty participants (40 offenders with documented antisocial or borderline traits and 40 controls) completed structured psychodrama role-plays under “emotionally charged” and “neutral” conditions. During sessions, multimodal AI tools (facial action coding, voice prosody analysis, and movement tracking) quantified implicit affective markers – emotional variability, facial micro-expression frequency, and interpersonal synchrony. Participants also completed standardized measures (Difficulties in Emotion Regulation Scale [DERS][1], Interpersonal Reactivity Index [IRI], Positive and Negative Affect Schedule [PANAS]) and provided written reflections. In quantitative analyses, forensic participants showed higher emotion dysregulation (DERS) and lower trait empathy (IRI empathic concern) than controls (p