Boston Police Department Court Overtime: Data science for policymaking

Each semester a number of small-scale data science projects are selected for the lab. These projects enable students to apply their skills in a real-world setting while solving problems that add to the success metrics of corporate partners. The lab and its courses serve as an exploration space with the goal of organizing work around particular thematic areas and making projects that are both rigorous and interdisciplinary. Thematic areas include race, gender, and economic equity and access; urban data mechanics; and sustainability.

All projects emerging from the BU Spark! Lab follow a standard step-by-step progression, which includes the following four deliverables:

1. Project Intake with Prospective Clients

Deliverable 1: In this step, student teams are expected to meet with their client (external partner), review the scope of their project, and submit a final project description. Within the project description should be information regarding all data sources that will be collected by the team, any datasets that will be accessed to enhance the project, specific research questions that will be answered using collected data, and a detailed, step-by-step plan for how the data will be transformed and the research questions will be answered.

2. Data Collection

Deliverable 2: After completing the project intake, students must collect data for the client and perform preliminary data analysis. In this step, the aim is for teams to attempt to answer one core research question relevant to their project proposal.

3. Data Pre-Processing and Cleaning        

Teams will be sorting, processing, and cleaning data as their projects progress. The final data will ultimately be uploaded to a dataset repository that can be reviewed by classmates and partners, as well as future students of the course.

4. Data Analysis and Final Presentation

Deliverable 3: At this stage, all proposed project questions should be reviewed, answered, and then submitted in a written document that outlines the project findings. Teams must also submit the data associated with their project and a description of what each label and feature in the dataset represents. Before presenting this deliverable, the team is encouraged to meet with the client to review all components of the project and discuss the project’s key findings.

Deliverable 4: The final project submission is an enhanced version of Deliverable 3.

Boston Police Department Court Overtime Project

The Problem

Taxpayer spending on police in Boston is exorbitant compared with funds allocated to other municipal departments.

The Boston Police Court Overtime Project was a collaborative effort between the American Civil Liberties Union (ACLU) of Massachusetts, Progressive Massachusetts, Boston City Councilor Ricardo Arroyo, and a team of Boston University students supervised by the BU Spark! Program based in the BU Faculty of Computing and Data Sciences and led by Ziba Cranmer. The research team was given the task of analyzing police court appearance overtime for the Boston Police Department from 2014 to 2019 in an effort to identify opportunities for redesigning the overtime policy and reduce spending. Using various sources of data, the team conducted an analysis of Boston police court overtime records as well as publicly available data from police personnel records and the 2019 city of Boston employee earnings report.

Research Findings

The BU Spark! Lab student team worked together to collect and sort the data. The students then created a portfolio of interactive visualizations reflecting their evaluation of the overtime data. The analysis yielded a number of observations:

1. Excessive court overtime hours were seen as a pay incentive and were increasingly exploited by officers.
2. The mandated minimum court appearance policy was unnecessary and wasteful.
3. Boston Police Department accountability needed to be more transparent.

Proposed Solutions

Given the findings, the research team proposed policy-based recommendations to effectively reduce the waste and exploitation of overtime hours and pay. First, the team suggested eliminating the four-hour minimum overtime court appearance policy, which came at immense cost to the city and largely resulted in overtime pay to officers for hours not worked: over $18 million over a five-year period. The team suggested that this money be redirected into city departments and social programming in the Boston community that are traditionally underfunded.

Next, the team recommended capping overtime pay as a percentage of an officer’s annual salary as well as limiting total police department overtime spending per year. The team found that in 2019, 17% to 25% of Boston police officers’ salaries — ranging from $15,000 to $53,000 — came from overtime pay. Given that not all court appearances carry the same importance and that not all police districts contribute equally to the overtime hours, the team suggested that the district attorney’s office be more selective in issuing summons to officers depending on the necessity of a court appearance.

Finally, the team recommended that the Boston Police Department be required to release quarterly data on its overtime use and that the city implement firmer accountability protocols to increase transparency between the police department and the public. The final report of data visualizations, analysis findings and policy recommendations were compiled in the form of a brief that was submitted to the Boston City Council Committee on Ways and Means hearing on Dockets #0839 and #10389 to discuss police department overtime use and oversight protocols.

This project is one of numerous projects conducted through the BU Spark! Lab each semester. At the completion of a project, insights and takeaways are gathered to help future students and teams working on public interest technology projects.

Insight for Students: Data science can and should be contextualized in the real world.

BU Spark! Lab participants contributed to this project as technical partners, providing data analysis and graphics that were subsequently interpreted by Progressive Massachusetts and the ACLU. Notably, this project provided important real-world context for data science that is socially relevant. Working with real-world data has a tendency to be messier than expected and requires strategic and thoughtful processes to merge, match, and clean data. Students may struggle to put themselves in the mindset of a client when anticipating useful data visualizations that tell a concrete story versus creating a set of discrete visualizations.

Therefore, understanding the context or associated sensitivity associated with the data students work with, specifically the assumptions involved with analyzing data about race, has become a teachable moment and opportunity for growth in this class.  Before taking this course, students do not always know what their interests outside of data science are, and for many, this practicum course is their first close encounter with contentious social issues. As such, this course can be a meaningful introduction to careers involving public interest technology.

Insight for Faculty and Staff: Distributing instructional tasks is a best practice.

Professors often want to integrate computer science instruction and real-world projects in their courses, which can be logistically challenging and costly. Professors are often more focused on teaching technical methods than on the contextual realities of projects.

A learning lab such as BU Spark! and the Faculty of Computing and Data Science who operate it, is a valuable resource in this regard. CDS is able to provide support for professors in the Computer Science Department by stepping in and contextualizing these realities through the creation of a shadow infrastructure. Supporting faculty who want to integrate these components helps to improve and gain efficiencies for various courses.

CDS is also building its own courses focused around a series of these “practicum” approaches at Boston University. This sort of program infrastructure could be beneficial if applied at other institutions.