Artificial Intelligence in Intimate Partner Violence Risk Pathways: A PRISMA-ScR Review of Femicide Prevention and Medico-Legal Accountability

Artificial intelligence (AI), machine learning, natural language processing and related decision15 support methods are increasingly studied in intimate partner violence (IPV), domestic-violence and 16 gender-based violence contexts. The key question is not whether AI can predict femicide as an 17 individual lethal event, but whether AI-related methods may help institutions recognise, document, 18 communicate and act on distributed signs of escalation across clinical, legal, police, social-service 19 and digital settings. This PRISMA-ScR scoping review, informed by Joanna Briggs Institute 20 guidance and structured using the Population-Concept-Context framework, mapped English21 language AI-related literature in IPV, domestic violence, coercive-control and femicide-related risk 22 pathways. Sexual violence was included only when embedded in IPV, domestic-abuse, coercive23 control, family-violence, lethality-risk or femicide-related pathways. Searches identified 4,099 24 records; after deduplication, 2,906 were screened, 166 reports were assessed at full text and 125 were 25 included in the core evidence map. The evidence was heterogeneous, spanning clinical and electronic 26 health records, police narratives, legal documents, social media or online posts, survey data, linked 27 administrative data and survivor-facing digital tools. AI-related methods were used mainly for 28 detection, classification, record linkage, risk stratification, text mining, triage or decision support 29 rather than for direct evaluation of femicide-prevention interventions. Femicide, lethality and severe 30 escalation were addressed in only part of the corpus, and few studies examined implementation, 31 human oversight, false reassurance, fairness, privacy or downstream institutional action in depth. The 32 findings do not support individual femicide prediction or demonstrate that AI prevents lethal 33 violence. Instead, they support a more defensible role for AI as a bounded component in human-led 34 risk-recognition pathways. The review develops a six-layer conceptual synthesis linking distributed 35 risk signals, AI-assisted signal processing, human contextual review, multi-agency response, legalIn review 2 36 ethical governance and medico-legal accountability. AI may support institutional recognition and 37 coordination, but it cannot substitute for professional judgment, survivor-centred practice, due 38 process or adequately resourced prevention systems.