Date: Monday, April 14, 2025 at 3-4pm CT / 4-5pm ET
The virtual panel on “Findings from Ingram Olkin Forum on Statistical Analysis of Police Use of Force” held on April 14, 2025, summarized and expanded on key topics related to the statistical analysis of police use of force. Each panelist was an author or co-author of one of the publications resulting from the Ingram Olkin Forum hosted by the National Institute of Statistical Sciences (NISS) and Carleton College in November 2023. That forum was motivated by the urgent issue of excessive use of force by police and the challenges involved in effectively analyzing data on policing practices. The publications are listed at the bottom of this news story.
The panel featured Gregory Lanzalotto, PhD Candidate at the University of Pennsylvania Wharton School; Dr. Elizabeth Brault, Assistant Professor of Criminology and Criminal Justice at Merrimack College; Dr. Cristian Allen, Assistant Professor of Mathematics at Wartburg College; and Dr. Jennifer Wyatt Bourgeois, Postdoctoral Fellow at the Center for Justice Research at Texas Southern University. The discussion was moderated by Dr. Claire Kelling, Assistant Professor of Statistics in the Department of Mathematics and Statistics at Carleton College. Together, they explored four key areas of research related to police use of force data: causal inference, the use of unstructured data, standardization of data, and spatial analysis. The panel aimed to provide an accessible overview for statisticians new to the field while identifying key directions for future research.
Our panelists discussed the challenges and importance of statistical analysis of police use of force data, emphasizing the need for better standardization, more accessible data, and stronger collaboration between statisticians and police departments. They also examined the limitations of administrative police data and the role of AI in processing unstructured data and improving transparency, underscoring the critical role of community involvement. The session concluded with a discussion on bias detection in spatial methods, highlighting issues around funding, data collection, and the need for open-source solutions.
Spatial Analysis Challenges in Police Data
Dr. Elizabeth Brault presented on spatial analysis issues, highlighting challenges with geographic boundaries, socioeconomic factors, and data privacy. She emphasized the importance of considering different levels of aggregation, matching policing data to administrative boundaries, and accounting for spatial dependence and heterogeneity. She also noted the need for better standardization, more accessible data, and collaboration between statisticians and police to develop fair and privacy-protecting methods for analyzing use of force data.
Causal Inference in Police Use of Force
Gregory Lanzalotto discussed the importance of causal inference in understanding disparities in police use of force. He used the example of Portland, where there was a clear disparity in the use of force across racial groups, but the statistics alone couldn't explain why. He emphasized the need for tools that can help go beyond describing disparities to understanding why they occur. He also introduced the concept of evidence-based policing, which applies empirical research to public safety. Lanzalotto outlined the types of questions causal inference can help answer, such as estimating the proportion of force incidents that are unjustified and investigating whether force is applied differently to civilians of different races. He also highlighted the challenges in analyzing police use of force, including incomplete or selectively recorded data.
Improving Police Data Quality Strategies
Lanzalotto continued by discussing the limitations of police administrative data, including missing denominators, selection bias, and mismeasurement, which can lead to systematically distorted data. He suggested several strategies to improve data quality, such as fusing police data with third-party sources like mobile location data and traffic sensors and implementing body-worn camera audits. He also emphasized the need for transparency in handling uncertainty and the development of better tools for handling non-random missingness in administrative data sets. He highlighted the challenges of implementing randomized control trials in policing interventions and the need for an interdisciplinary approach to address these issues. His ongoing work focuses on defining causal estimates, addressing bias, and exploring the use of generative AI to support this work.
AI Transforms Unstructured Data into Action
Following Lanzalotto's remarks, Dr. Cristian Allendiscussed the challenges of unstructured data and the importance of community involvement in organizing and making data accessible. Allen presented four case studies, including the Guatemalan Civil War and the Puerto Rico Police Bureau, where human rights organizations used AI to transform unstructured data into structured data. These efforts led to significant outcomes, such as the conviction of a dictator and the exposure of systemic civil rights violations.
AI in Policing and Accountability
Dr. Allen also discussed the potential of AI in policing and accountability, highlighting the work of Trina Reynolds Tyler, who won a Pulitzer Prize for local reporting. Tyler’s work focused on gender-based violence and the difficulties of categorizing incidents. Allen introduced the machine learning model Judy, developed by HRDAG and the Lucy Parsons Lab, which reclassifies police reports to identify gender-based violence. The model successfully identified patterns of abuse, demonstrating the power of AI in uncovering hidden trends in police data. He also referenced the Innocence Project’s use of large language models and AI to develop metrics for wrongful convictions and emphasized the importance of digitization, categorization, and the creation of public databases for transparency and accountability.
Standardizing Police Use of Force Protocols
Dr. Jennifer Wyatt Bourgeois focused her talk on the importance of developing standard protocols and data collection methods for police use of force incidents. She emphasized that consistent reporting is critical for public trust, officer accountability, and accurate analysis of policing practices. Bourgeois identified the challenges of different jurisdictions defining and categorizing use of force inconsistently, which hinders meaningful comparisons and policy evaluation. She proposed an expanded and layered version of the traditional use of force continuum to categorize officer responses and stressed the need for coordinated stakeholder engagement—including law enforcement, researchers, policymakers, and community organizations—to support standardization.
Q&A Discussion
The webinar concluded with a panel discussion on bias detection and measurement in spatial methods. Panelists noted the impacts of funding cuts on research and data collection, while also expressing optimism for future support. The importance of transparency and the use of open-source software in police use of force research was highlighted, and the conversation ended with a call for continued involvement in the field of law and justice statistics.
Acknowledgement and Recognition
The passion and expertise of the speakers and moderator were clearly reflected in their presentations and discussions. NISS sincerely thanks Gregory Lanzalotto, Dr. Elizabeth Brault, Dr. Cristian Allen, Dr. Jennifer Wyatt Bourgeois, and moderator Dr. Claire Kelling for sharing their insights and advancing important conversations at the intersection of statistics, policing, and social justice.
See Publications from Police Use of Force forum:
Bourgeois, J. W., Haensch, A., Kher, S., Knox, D., Lanzalotto, G., & Wong, T. A. (2024). How to Use Causal Inference to Study Use of Force. CHANCE, 37(4), 6–10. https://doi.org/10.1080/09332480.2024.2434435
Brault, E. E., Kelling, C., Bourgeois, J. W., Taheri, S. A., Jones, A., Charles, C., … Banks, D. (2024). Toward Standardization of Police Use of Force Data. CHANCE, 37(4), 24–30. https://doi.org/10.1080/09332480.2024.2434438
Kelling, C. (2024). Statisticians Address Analysis of Police Use of Force. CHANCE, 37(4), 4–5. https://doi.org/10.1080/09332480.2024.2434432
Kelling, C., Allen, C., Brault, E. E., & Matos, P. (2024). Issues in the Spatial Analysis of Police Use of Force Data. CHANCE, 37(4), 11–17. https://doi.org/10.1080/09332480.2024.2434436
Shah, T., Allen, C., Ibrahim, A., Kefalas, H., & Stevens, B. (2024). The Use of Unstructured Data to Study Police Use of Force. CHANCE, 37(4), 18–23. https://doi.org/10.1080/09332480.2024.2434437