Data for Justice

This project was my Capstone Project as part of the NYU Data Science Master’s Program. My team and I (on behalf of the Marron Institute worked with the District Attorney’s office in Fort Bend County, Texas to modernize their data infrastructure, and determine if any bias existed in the prosecutor’s office.

Most people think of the criminal justice system as unifed, monolithic entity, but the reality is that each part of the criminal justice process functions largely independently, and only recently have many offices adapted modern electronic recordkeeping systems. As part of our project, my team and I built a data dashboard that monitors the case flow of the prosecutors office, reducing the time it takes to fill out monthly reports, and also providing overviews of crime type, sliced by demographic and geographic variables. Furthermore, we examined to what extent bias against protected classes (e.g. race and gender) existed in prosecutorial decision-making. We found that large amounts of bias existed in only a few decision-making points - this runs counter to the standard narrative of small amounts of bias existing at every decision-making point (what we normally think of as small prejudices adding up over time).

Due to privacy reasons, we cannot share the Github repo, but the final paper is available here.