Research Brief: May 2026
- Darwin
- 2 hours ago
- 3 min read
Here are the studies we found this month that may be worth your attention. As ever, if you spotted something we missed, let us know and we'll add it into the next brief.
TL;DR Summary
Strip searches in police custody and ethnicity Black detainees face significantly higher odds of being strip-searched in custody, even after controlling for offence type, vulnerability and use of force.
What shapes police officers' attitudes toward AI-assisted policing Organisational justice and supervisory support significantly shape whether officers accept AI tools.
A community policing communication field experiment A brief collaborative message during door-to-door visits significantly improved citizen perceptions and reporting intentions.
School exclusion, missing children and serious harm School exclusion, exploitation and missing incidents frequently overlap in the most seriously harmed child cases.
Automated feedback and AI body-worn camera review Automated positive feedback emails improved officers' acceptance of AI-driven body-worn camera review.
Strip searches in police custody and their association with ethnicity: evidence from an English force
Authors: Ali and Dymond
Study design: Quantitative analysis of administrative custody data (binary logistic regression, N=25,676)
Setting: England and Wales
Summary: Using four years of custody records from an English force, this study finds that Black detainees face over twice the odds of being strip-searched compared to White detainees, even after controlling for offence type, vulnerability indicators and use of force. Black men and Black children face disproportionately higher odds than either characteristic alone would predict, providing evidence of racialised policing in the custody suite.
You'll be interested if: you work in custody, equality, professional standards, inspection or research, or if you are responsible for ensuring policing is fair and lawful. Relevant to all forces given proposed PACE Code C changes.
What shapes police officers' attitudes toward AI-assisted policing
Authors: Jian, Sun, Wu et al
Study design: Vignette experiment
Setting: Taiwan
Summary: This vignette experiment tests what contextual factors shape police officers' willingness to accept AI-assisted tools. It finds that organisational justice, supervisory support, and the perceived purpose of the AI system all significantly influence officer attitudes. Officers are more accepting when they believe the technology supports rather than replaces professional judgement.
You'll be interested if: you are implementing AI tools in a force, leading technology change, or researching the human side of AI adoption in policing.
How police communication influences citizen perceptions and cooperation: a community policing field experiment
Authors: Canales and Santini
Study design: Randomised field experiment (N=980 households)
Setting: Mexico
Summary: Officers conducting door-to-door visits were randomly assigned to vary their uniform type and whether they delivered a brief collaborative safety message. The message condition significantly improved citizen perceptions of police and their willingness to report crime. Uniform type alone had little effect. Findings support procedural justice theory and suggest low-cost communication changes can meaningfully improve public cooperation.
You'll be interested if: you work in neighbourhood policing, community engagement or public confidence. Note: Mexican context.
School exclusion, missing children and serious harm: disentangling the interlinking factors
Authors: Fox
Study design: Mixed-methods analysis of Child Safeguarding Practice Reviews using machine learning-assisted content analysis
Setting: UK
Summary: This preprint analyses Child Safeguarding Practice Reviews to examine how school exclusion, missing child incidents and exploitation co-occur in the most seriously harmed cases. Findings show these factors frequently intersect, with exclusion functioning as both a vulnerability indicator and a pathway to further harm. Uses accessible machine learning methods to analyse sensitive text data at scale.
You'll be interested if: you work in child protection, exploitation, vulnerability, or multi-agency safeguarding. Also relevant for those interested in how AI-assisted text analysis can support research with sensitive case review data.
Automated feedback and the acceptability of AI-generated body-worn camera review: an implementation science natural experiment
Authors: Watts, del Pozo, White and Malm
Study design: Natural implementation science experiment (within context of two randomised controlled trials)
Setting: United States
Summary: Automated emails highlighting positive professionalism scores from AI-driven body-worn camera analysis were associated with significantly improved officer acceptance of the technology. This is the first implementation science study in a police setting and finds that feedback design, not just tool validity, matters for successful adoption.
You'll be interested if: you are implementing body-worn camera AI, leading technology adoption, or interested in what implementation science can offer to evidence-based policing.
