Psychodrama is an action-oriented psychotherapy in which participants enact personal experiences or symbolic scenes through role-play, auxiliary egos, and stage enactment[5][6]. Originating with Moreno, psychodrama emphasizes spontaneity and creativity; it can elicit insight, abreaction, and confrontation with others’ feelings, thereby broadening a person’s understanding of internal conflicts[5][6]. In clinical settings, psychodrama has been applied to patients with severe personality pathology (e.g. borderline or narcissistic traits) who often have limited verbal insight[7][5]. For example, Olsson (2018) describes using psychodramatically informed techniques to engage “borderline patients who have meager capacity for insight and limited verbal skills,” facilitating emotional processing without requiring sophisticated speech[7]. In forensic settings, arts therapies (including drama and psychodrama) are widely used: reviews indicate that such experiential, non-verbal modalities can improve prisoners’ emotion regulation, anger management, empathy, and social functioning[8][9]. A recent forensic arts-therapy meta-analysis concluded that arts-based interventions yield positive effects on coping with emotions and anger control, with reported increases in self-confidence and social skills[9]. Psychodrama’s emphasis on enactment may thus be especially promising for inmates or offenders, allowing emotional tension to be expressed, observed, and reframed in a controlled setting[8][6]. Concurrently, advances in artificial intelligence (AI) and affective computing have made it feasible to quantify subtle behavioral signals of emotion and social attunement. State-of-the-art systems can analyze facial expressions (micro-expressions, action units), vocal prosody, and body motion to infer emotional states and interpersonal synchrony (the moment-to-moment alignment of nonverbal behaviors). In mental health research, AI-driven speech-emotion recognition has distinguished clinical groups from controls and identified specific voice biomarkers for disorders[10]. For example, Lombardo et al. (2025) found that algorithms could use prosodic patterns to detect mood disorders, confirming that “emotions can be successfully used as an intermediary step for mental disease detection”[10]. Likewise, human–agent interaction studies show that people naturally mimic a partner’s facial expressions and gestures in empathic engagement[11]. Sung Park et al. (2022) demonstrated significant facial synchrony (cosine-similarity of expression dynamics) when participants interacted with an emotional virtual agent, and linked the degree of facial mimicry to individual differences in affective empathy[11]. These findings suggest that measurements of nonverbal synchrony and expression intensity could serve as objective indices of interpersonal attunement and empathy. However, purely AImediated therapy (e.g. chatbots) cannot replace genuine human empathy or capture full contextual meaning[12]. AI’s role may be best as an augmenting tool – a “data-informed” supplement to human-led interventions. Understanding emotional regulation and empathy in forensic populations is particularly challenging. Individuals with antisocial personality traits or psychopathy typically exhibit profound empathy deficits, callousness, and impulsivity[3][13]. Psychopathy is defined by a “wide range of emotional deficits, including lack of empathy [and] emotion dysregulation”[13]. Empirical studies confirm that offenders often have low affective empathy and difficulty recognizing others’ distress, even if cognitive perspective-taking is intact[13][3]. Borderline personality disorder (BPD), by contrast, involves intense, unstable emotions and high emotional sensitivity[14]. BPD patients may be highly reactive to others’ emotions but struggle to identify or regulate their own affect[15]. Overall, emotion dysregulation is considered a core feature of BPD, and one review notes that “emotional sensitivity, emotion regulation and impulsivity are fundamental topics” in BPD research[14]. This dysregulation can contribute to interpersonal conflict, self-harm, and violence. Indeed, forensic research suggests maladaptive regulation (and associated impulsivity) increases risk of reoffending[16]. Given these clinical and technological trends, we propose an integrative approach: a psychodrama therapy protocol enhanced with AI-based multimodal sensing. The staged role-play allows individuals to express and encounter emotions in vivo, while AI analytics objectively measure the nonverbal and paraverbal cues underlying those emotional states. In this way we aim to “map” implicit affect. We hypothesize that offenders (antisocial and borderline traits) will show restricted affective responses and less synchronization than community controls, both behaviorally and in AI metrics, and that these implicit measures will correlate with self-report of emotion regulation (DERS) and empathy (IRI). The mixed-method design also incorporates qualitative feedback to capture participants’ subjective experience of the exercises. This study thus explores a novel “AI-assisted psychodrama” model that could advance emotion-focused forensic rehabilitation through data-driven insights.