What makes a police stop feel respectful? To find out, researchers from the USC Dornsife College of Letters, Arts and Sciences are leading a multi-university team that has analyzed body camera footage from 1,000 Los Angeles traffic stops.
Why it matters: Trust between police and the communities they serve hinges on day-to-day encounters - especially traffic stops, one of the most common.
- Yet, little is known about what actually happens during those stops or what makes them feel respectful to the people involved.
- AI could help, but only if it's trained on what respect looks like.
By using body camera footage to examine these encounters in detail, researchers with the Everyday Respect Project aim to identify patterns that foster trust and those that erode it. Their findings will help inform AI tools that will eventually scale that work to many thousands of interactions.
The Everyday Respect Project process:
- Community-driven definitions of respect: Surveys, interviews and focus groups help define what "good" communication looks like from multiple stakeholder perspectives.
- Training AI with diverse human input: A team of 43 annotators - including community members and retired officers - rate bodycam footage to train machine learning tools.
- Scaling up: These AI models will eventually analyze footage from roughly 30,000 traffic stops by L.A. Police Department officers.
- Understanding escalation and de-escalation: Researchers will identify what drives respectful communication and what causes interactions to break down.
- Transparency and impact: The tools and findings will be shared publicly to support improvements in traffic stop practices nationwide.
How the study protects privacy: Bodycam footage was analyzed under strict privacy protocols; neither officers nor drivers are identified.
- The goal is to understand how tone, language and behavior shape perceptions of respect - and to support practices that build mutual trust.
What they're saying: "Body camera footage gives us a rare window into everyday interactions that shape public trust in policing," says Ben Graham, coordinating principal investigator of the Everyday Respect Project and associate professor of international relations at USC Dornsife.
- "We want to learn what respect looks and sounds like and how it can be encouraged."
Initial findings, which were recently presented to the Los Angeles Board of Police Commissioners, are available in the Everyday Respect Project Report.
What's next: The team is refining tools that can teach AI to recognize respectful communication - along with a wide variety of other stop features - and evaluate some 30,000 videos of real-world traffic stops.
- They hope the model can help cities nationwide evaluate - and improve - everyday policing practices.
Who's behind it: In addition to USC Dornsife, the project brings together researchers from the USC Sol Price School of Public Policy; USC Information Sciences Institute; Georgetown University Massive Data Institute; University of Texas at Austin; and University of California, Riverside.
Who's funding it: Support for the project comes from the National Science Foundation, Arnold Ventures, the Microsoft Justice Reform Initiative, Google (Award for Inclusion Research), and a USC Zumberge Interdisciplinary Research Grant.
