AI-supported Video
Analysis for Law

Enforcement Investigators

In this user centered project we collaborated with Europol Innovation Lab aswell as the swedish police. We worked closely with bringing the police stories to life to create a UI dedicated to supporting their work. This project examines how AI and computer-vision can support law enforcement investigators in an agency-driven, efficient, and iterative manner.


We were also selected to participate in Europool Innovation Days to present our projects, due to their satisfaction with the class’s performance.

2025 / 6 weeks / Team / Interaction Design

Special thanks to my teammates: Pius Burkhart, William van der Bijl, Yadu Sharon M G

In collaboration with

X


AI-supported Video Analysis for Law Enforcement Investigators

In this user centered project we collaborated with Europol Innovation Lab aswell as the swedish police. We worked closely with bringing the police stories to life to create a UI dedicated to supporting their work. This project examines how AI and computer-vision can support law enforcement investigators in an agency-driven, efficient, and iterative manner.


We were also selected to participate in Europool Innovation Days to present our projects, due to their satisfaction with the class’s performance.

2025 / 6 weeks / Team / UI, Interaction Design

Special thanks to my teammates: Pius Burkhart, William van der Bijl, Yadu Sharon M G

In collaboration with

X

In this user centered project we collaborated with Europol Innovation Lab aswell as the swedish police. We worked closely with bringing the police stories to life to create a UI dedicated to supporting their work. This project examines how AI and computer-vision can support law enforcement investigators in an agency-driven, efficient, and iterative manner.


We were also selected to participate in Europool Innovation Days to present our projects, due to their satisfaction with the class’s performance.

2025 / 6 weeks / Team / Interaction Design

Special thanks to my teammates: Pius Burkhart, William van der Bijl, Yadu Sharon M G

In collaboration with

X

AI-supported Video Analysis for Law Enforcement Investigator

Law Enforcement Agencies face the challenge of processing large amounts of footage in a small amount of time

Brief


In this project, our class received a brief from Europol to design a user interface that helps law enforcement investigators efficiently, adaptively, and responsibly process large volumes of video to identify critical information. The project focused on how AI-driven computer vision can meaningfully support these investigative practices.

Understanding context


The workflow of processing video material in an investigation is part of a broader process that begins with a case being reported to the police and may end with an investigator testifying in court about findings from video analysis. Cases can span from a few weeks to several months, with work continuously moving between investigators, analysts, and prosecutors.


For the scope of this project, we focused on the organizing, processing, and analyzing stages, while keeping the broader investigative context in mind to ensure a meaningful and cohesive experience.

Selected user stories from

Stockholm and Umeå police


Analyst,

Umeå

Investigator, Stockholm

Intelligence officer,

Stockholm

Riding the curiosity wave


Police work is in its nature iterative, and investigators often find themselves being “briefed by the data.”

Riding the curiosity wave


Police work is in its nature iterative, and investigators often find themselves being “briefed by the data.”


If I know your folderstructure I know you


In image and video analysis, context such as folder structure or EXIF data can be just as important as the visual content.

If I know your folderstructure I know you


In image and video analysis, context such as folder structure or EXIF data can be just as important as the visual content.


Looking for a needle in a haystack… or is it a screw?

Sometimes investigators face tasks where both what to look for and where to look are unclear.

Looking for a needle in a haystack… or is it a screw?

Sometimes investigators face tasks where both what to look for and where to look are unclear.




Spectrum of work


Investigative work is inherently unpredictable and highly situational. Each process varies depending on who conducts it, the nature of the case, and the stage of the investigation. Rather than following a linear path, investigative work, much like design practice, adapts continuously to its context. As a result, no single workflow or automated logic can account for every situation.


Spectrum of work


Investigative work is inherently unpredictable and highly situational. Each process varies depending on who conducts it, the nature of the case, and the stage of the investigation. Rather than following a linear path, investigative work, much like design practice, adapts continuously to its context. As a result, no single workflow or automated logic can account for every situation.


Who


Roles range from highly specialized analysts to generalist investigators, working in setups that vary from large multi-monitor workstations

to laptops used on the move.

Who


Roles range from highly specialized analysts to generalist investigators, working in setups that vary from large multi-monitor workstations

to laptops used on the move.

What


Wide range of investigative tasks: ranging from looking for a very specific clue or looking for patterns to discover crimes that hasn't even happened yet.




What


Wide range of investigative tasks: ranging from looking for a very specific clue or looking for patterns to discover crimes that hasn't even happened yet.

When


Investigative work changes over time as new clues emerge, briefs evolve, and directions shift through coordination with prosecutors.


When


Investigative work changes over time as new clues emerge, briefs evolve, and directions shift through coordination with prosecutors.


—————>

Design a tool that helps law enforcement investigators review and manage large quantities of video material in an Iterative and Agency-driven way to identify key insights?

Interface


The result is a light weight system designed to support diverse

investigative practices through an open UI architecture and multiple data representations. Searches are structured as stacked steps to enable iterative work while maintaining transparency and reducing errors. Gallery, timeline, and cluster views support different investigative scenarios.

Interface


The result is a light weight system designed to support diverse

investigative practices through an open UI architecture and multiple data representations. Searches are structured as stacked steps to enable iterative work while maintaining transparency and reducing errors. Gallery, timeline, and cluster views support different investigative scenarios.

Search


Searches, or prompts, are structured as stacked steps to reflect the iterative nature of investigative work. Investigators can progressively refine and combine parameters to find what they need while maintaining a clear mental model of the process.

Stacked searches ->

Stacked searches


Searches are built as stacked steps to maintain transparency and control. Because results are probability-based, combining multiple searches can introduce errors.


Stacking makes each step visible and traceable, giving investigators full control over inputs and a clear understanding of results. Aswell as supporting iterative work.

Computer
Vision Visuali-zation


An AI algorithm scores each video based on the probability of a search term, shown as peak and average values.


Scores are visualized with a color gradient and graph for quick recognition.


Investigators can filter results to focus on the probability ranges that matter most.

Computer Vision Visual-ization


An AI algorithm scores each video based on the probability of a search term, shown as peak and average values.


Scores are visualized with a color gradient and graph for quick recognition.


Investigators can filter results to focus on the probability ranges that matter most.

Computer Vision

Visualization


An AI algorithm scores each video based on the probability of a search term, shown as peak and average values.


Scores are visualized with a color gradient and An AI algorithm scores each video based on the probability of a search term, shown as peak and average values.


Scores are visualized with a color gradient and graph for quick recognition.


Investigators can filter results to focus on the probability ranges that matter most.

Gallery


The data browsing view enables quick scanning of search results. Supported by probability scores, investigators can identify relevant material efficiently, and by hovering over items, preview thumbnail videos.


Rather than relying on a single fixed view, flexible data viewers allow investigators to approach the data from different perspectives, depending on their task and context.

Gallery


The data browsing view enables quick scanning of search results. Supported by probability scores, investigators can identify relevant material efficiently, and by hovering over items, preview thumbnail videos.


Rather than relying on a single fixed view, flexible data viewers allow investigators to approach the data from different perspectives, depending on their task and context.

Gallery


The gallery view supports quick browsing and comparison, making it easier to scan large amounts of data and narrow down relevant items.

Timeline


Provides contextual understanding of the data by revealing the sequence of events and highlighting key moments over time.

Cluster


Acts as a thread starter for exploratory investigation, grouping data to reveal patterns and surface potential leads.


Through a card-like flip, investigators can quickly acsess EXIF-data and pathways.


Through a card-like flip, investigators can quickly acsess EXIF-data and pathways.

Video analysis


A focused workspace for investigative video review. Users can closely examine footage, access contextual metadata such as time and location, and extract relevant moments as evidence. Findings can be added to active searches and organized through saving, categorization, or export.

Video analysis


A focused workspace for investigative video review. Users can closely examine footage, access contextual metadata such as time and location, and extract relevant moments as evidence. Findings can be added to active searches and organized through saving, categorization, or export.

Concept video ->

Process ->

Collaborative class workshop

Collaborative class workshop


Police study visits


Embodied explorations of research


Active Interviewing and story exctraction

Invited to present at Europol innovation days in Den Haag

Reflections & learnings


What I loved most in this project was meeting and designing for police investigators, and seeing the important work they do. It was exciting to hear how they imagined using the tools and to see them feel understood.


The biggest challenge was scoping, because the context was large and complex and time was short. I learned how to create shared materials and metaphors as a team to navigate this complexity and focus our design.



+Research +Embodied prototyping +UI design