Human-AI Interaction for Time-Critical Sensemaking in Missing Persons Investigations

Every year an estimated 200,000 people go missing in the UK alone. Missing persons investigations involve challenging time-critical sensemaking tasks based on fragmented data sources. This paper describes a mixed-methods participatory study evaluating data science and AI-driven techniques (summarisation, fact extraction, and data visualisation) for supporting these investigations as part of a human-centered workflow. A series of human-AI interfaces were iteratively designed and tested with search officers and domain experts at Police Scotland. Based on findings, we describe:

(1) user and information needs for missing persons investigations;

(2) insights on the benefits and challenges of applying LLM-based techniques in high-risk contexts; and (3) lessons for integrating AI for sensemaking tasks in policing more broadly. We highlight that in high-stakes contexts, where accuracy and context-sensitivity are paramount, AI techniques must be balanced with other approaches and designed in close partnership with end-users.