Using Machine Learning to Illuminate Patterns in BPD Complaints Data

During the Fall 2023 semester, the NLG-MA Litigation Committee partnered with the Boston University Spark! program on a project to use machine learning to analyze complaints against the Boston Police Department. Spark! is an interdisciplinary program for undergraduates combining artificial intelligence, data science, and investigative journalism approaches to help organizations with real-world data projects. This project represents one facet of the Litigation Committee’s work to achieve greater police accountability. We hope the analysis will prove useful to others in the Guild and other citizen advocacy groups.

We asked the team of students to develop computer models and methodologies to analyze patterns of police behavior based on the complaints filed with Boston Police Internal Investigations Department.  While we are studying all kinds of allegations, we are focusing our efforts on cases involving excessive force and / or racial bias.  We are assessing the following questions in particular as we refine our methodology.

  • Can machine learning identify bias where it might be being hidden in procedural language?
  • Can modeling help us discern degrees of use of force in order to flag specific cases?
  • How do the allegations compare with accountability measures taken within the department? 

We hope to continue to work with the Spark! program in Spring 2024 to delve further into these questions, and continue the work of more transparency and accountability in policing.

LL Gordon, Benjamin Fanucci-Kiss, Doug Smith, David Kelston