From Detection to Healing: A Vision for Interactive, Context-Aware, Human-Centred XAI in Cyberbullying Intervention

Cyberbullying constitutes a mounting mental health emergency: rates of victimization have increased dramatically over the last ten years and strongly correlate with self-harm and suicidal actions (John et al., 2018). Explainable AI (XAI) is intended to connect automated detection with human comprehension (Kandala et al., 2025), but current multi-modal systems despite their high accuracy still fail to provide the contextual, compassionate explanations required in safety-critical settings. We outline a socio-technical research agenda grounded in latency-optimized cross-modal fusion, human-in-the-loop interactivity, tiered transparency, and human-centered evaluation, reframing XAI from a mere debugging aid into a vehicle for victim support, remediation, and psychological closure