From Practitioners to Algorithms: Predictors of Public Preferences towards AI-led Domestic Violence Risk Assessment

Domestic violence (DV) is a major social and population health issue across the globe, and risk assessment remains an important element in timely intervention and prevention. Conventionally, DV risk assessments have been based on human practitioners with structured instruments, although time, resource constraint, and bias are likely to restrain such methods. Artificial intelligence (AI) has recently been considered as an additional tool to provide efficiency, scalability and detecting complex data patterns. Nevertheless, the AI use in such a sensitive setting is not yet accepted by the general population. This research paper explored the psychological and attitudinal determinants of AI- vs practitioner-led DV risk assessment preferences. The study is a quantitative cross sectional survey involving adults in the general population, who completed validated scale measuring attitudes to AI, attitudes to help-seeking, confidentiality concerns and openness to experience. It was found that the majority of respondents preferred practitioner-led assessments. Although the variables explored did not significantly predict AI vs practitioner-led risk assessments in DV context, findings suggest that those who chose AI had significantly more positive attitudes to AI, more negative attitudes to help seeking, and higher confidentiality concerns. It is concluded that the public is not ready for AI use in such sensitive context, and AI must be viewed as an improving, but not a replacing, technology.