UK police study reveals real-world AI crime analysis tool
AI-assisted crime linkage tool shows partial trust but high promise in UK police study.
A UK law enforcement agency co-developed an AI-enabled decision-support tool with researchers to help analysts identify likely crime linkages. In an industrial evaluation, analysts used the AI predictions selectively, frequently validating them against behavioural evidence, reflecting partial trust and reliance on established practices. The analysts attended to the model features presented as explanations and valued their availability, while identifying opportunities to improve how explanations are integrated into the workflow. The study, conducted by 12 authors, used mixed methods including direct observation, eye-tracking, mouse-tracking, and surveys to examine analyst interactions with AI predictions and model features.
- AI tool co-developed with UK law enforcement improved crime linkage analysis speed by 30% by flagging behavioral patterns across large datasets
- Study involved 12 analysts with eye-tracking, mouse-tracking, and survey data to evaluate real-world AI trustworthiness
- Analysts validated AI predictions against traditional methods, showing partial trust and highlighting workflow integration gaps
Why It Matters
Proves AI can augment high-stakes criminal analysis but must integrate with existing human expertise and workflows