Umeå police headset study finds enthusiasm, not proof of safer interventions

Umeå University researchers reported positive police reactions to mixed-reality decision support in an account published on Phys.org on October 9, UTC. Their study involved 39 officers and instructors in Sweden and Catalonia using Microsoft HoloLens during a simulated high-risk vehicle stop, but it does not establish that the technology makes real interventions safer.

The current development is the university's account of the findings and plans to seek further research funding. The associated project ran from September 2023 to February 2024, and the paper was published in June 2026. Neither the experiments nor the journal publication took place this week.

What officers tested with Microsoft HoloLens

The collaboration brought together computing scientists, police researchers and practicing officers. According to the university, HoloLens projects digital information into the wearer's field of view and supports eye and voice controls. Identifying objects and potential risks are described as possible applications, rather than demonstrated detection performance.

Before the practical exercise, participants attended two workshops. One introduced principles of trustworthy AI; the other gave them hands-on experience with mixed reality and the headset. That preparation was intended to help officers evaluate both the possibilities of the technology and its ethical implications.

An important distinction concerns the familiarisation workshop: it used Remote Assist so facilitators could place visual cues. The paper says this activity introduced participants to mixed-reality possibilities rather than evaluating a specific prototype. Its findings should therefore not be treated as a performance test of a finished autonomous policing product.

Positive feedback is not an operational outcome

The university reports very positive attitudes toward usefulness, ease of use and trustworthiness. Anders Schogster, a police officer and teacher at Umeå's Unit of Police Work, said: “The results were better than I expected. The police officers were very interested in the technology and saw many possibilities.”

Schogster's assessment comes from the participating research team. The university account supplies no numerical improvement in intervention accuracy, injuries, crime prevention or public safety. Favorable perceptions do not independently establish technical reliability, fairness or whether officers' trust would be justified under operational conditions.

Jonas Hansson, a researcher at the Unit of Police Work, stressed the need for computing specialists to understand policing and for police specialists to understand technological possibilities. The collaboration combined AI, mixed-reality and ethical assessment expertise with operational experience and tactical knowledge.

Earlier police technology research shows why outcomes matter

A separate 2019 study by Hendrik Engelbrecht, Stephan Lukosch and Dragos Datcu offers relevant context. Their research evaluated a hotspot-policing system combining a smartphone, smartwatch and web application during actual police fieldwork, rather than the simulated vehicle-stop exercise used in the Umeå project.

The TU Delft research record describes a repeated-measures comparison of work with the system against a baseline without it. Researchers examined usability, task load and situational awareness in locations with high and low hotspot density. Officers gave the system high usability scores, and interview feedback was largely positive.

Nevertheless, situational-awareness measures remained stable between baseline and experimental shifts, while task load was significantly higher with the system. These findings show why user approval and measured effects need separate assessment. They concern different equipment and tasks, and do not establish that HoloLens increases workload or fails to improve performance.

Human decisions and the next research questions

In the university account, privacy, data storage, transparency and algorithmic bias are identified as potential risks. Researchers assessed functions across three risk levels and reported that greater automation increased ethical risks involving privacy, transparency and bias. Human autonomy remains a central condition of their proposed approach.

“It's important that this does not become fully autonomous. It should support police officers in their decision-making, but the decisions must ultimately be made by humans,” Hansson said. That is a design position, not confirmation that a deployed system meets those requirements.