Criminal Justice

Can AI Enhance Body-Worn Camera Footage Analysis At Scale?

Body-worn cameras capture millions of police-community interactions across the U.S. every year; however, the vast majority of this footage is never viewed. Working in partnership with multiple agencies and the Natural Language Processing Lab, SPARQ developed processes to transform officers’ body-worn camera footage into big data. Using a combination of machine learning, computational linguistics, and social science methods, we evaluate current policies and practices and their impact on police-community relations. We find, for instance, that officers use less respectful language and tone of voice when addressing Black drivers during routine traffic stops. Using a dataset of over 1.3 million body-worn camera videos from multiple agencies, we are continuing to develop tools that harness artificial intelligence (AI) to better understand police-community interactions at scale. Read more about our research on respectful language and tone of voice, the language of escalation, and the conversational sequence of traffic stops. Check out a video about our research on predicting escalation.

Can Body-Worn Camera Footage Provide New Insights for Policy Evaluation?

Although police officers stop millions of drivers each day across the U.S., not every driver is told why. To address this, California Assembly Bill 2773 now requires officers to state the reason for a traffic stop; yet there has not been a way to evaluate the bill’s effectiveness. In partnership with the Stanford Center for Racial Justice, the Natural Language Processing Lab, and multiple police departments, we developed AI tools using over 300,000 hours of body-worn camera footage to evaluate this policy and its impact on communities.

Can Body-Worn Camera Footage Strengthen Police Training?

SPARQ partners with police departments to study, develop, and implement police trainings. Building on our research, we worked with the Oakland Police Department to develop and test a procedural justice training to support  officers’ respectful language use during traffic stops. Using body-worn camera footage to design and evaluate the training, we found that officers used more respectful language when addressing drivers post-training. Read more about this work.

Do Past Policies Have an Impact on Policing Today?

Past discriminatory policies, like redlining—the practice of denying home loans based on the racial demographics of a neighborhood—continue to have an impact on health and educational outcomes today. However, it’s unclear whether they affect current policing practices. SPARQ is utilizing body-camera footage to examine whether there’s a link between these historical practices and how specific neighborhoods are policed.  

Image Credits

^{Diana Robinson / Flickr | Axon | Udo S / Flickr | Wikimedia Commons}