Session 2: Leveraging Technology to Improve Police-Community Relations

The talk provides a compelling and rigorous overview of a pioneering research project. The speakers are credible experts from Stanford and UC Davis, with a strong track record in this domain. The methodology is sound: they have built a massive dataset, developed robust AI tools, and are transparent about limitations. The argumentation is well-structured, moving from the problem (trust erosion) to the solution (AI analysis of body-worn camera footage) and then to specific tools and preliminary findings. The research is grounded in prior peer-reviewed work, and the team is careful to note that automatic transcription is still challenging. The sources cited are primarily their own prior work and the Hoffman-Yee grant program, which is appropriate for a research talk. The adéquation titre/contenu is excellent. The talk does not oversell the results; it acknowledges that these are early steps. However, the presentation is somewhat high-level, and the technical details of the AI models are not deeply explored. The audience is likely academic, but the talk is accessible. Overall, this is a high-quality presentation of cutting-edge research with significant societal implications.

The title accurately reflects the content, which focuses on using AI and LLMs to analyze police body-worn camera footage to improve police-community relations.

8/10

The talk presents ongoing research from a reputable academic institution (Stanford) with a multidisciplinary team, including social psychology, linguistics, and computer science. The methodology is based on large-scale data analysis and peer-reviewed foundations, but the results are preliminary and not yet fully published. The speakers are transparent about limitations and challenges.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction to the project and team by Jennifer Eberhardt.
• 02:00Discussion of the George Floyd case and the linguistic signature of escalation.
• 05:00Dan Jurafsky explains the history of the research, starting with the Oakland scandal and early analysis of 1,000 traffic stops.
• 10:00Rob Voigt describes the scaling up to 1.3 million videos and the data infrastructure built with Axon and police departments.
• 15:00Chen Shani presents new AI tools for detecting reasons for stops and analyzing cooperation dynamics.
• 20:00Preliminary results on the linguistic signature of escalation and racial disparities.
• 25:00Discussion of the potential impact on officer training and institutional interventions.
• 30:00Q&A session begins.

• Hoffman-Yee Research Grant Program — Funding source for the project, mentioned in the video description.

• Hoffman-Yee Research Grant Program — The grant program supports this research, indicating institutional backing.