Eliminating Human Bias: The Regulatory Landscape Governing Predictive Policing Technology

Long, Jennifer, School of Engineering and Applied Science, University of Virginia
Foley, Rider, EN-Engineering and Society, University of Virginia
Basit, Nada, EN-Comp Science Dept, University of Virginia
Norton, Peter, EN-Engineering and Society, University of Virginia

Technical Report Summary:
As part of a research team led by Professor Nada Basit and Professor Robbie Hott, I conducted a behavioral research project on analyzing impact of time-tracking methodologies in online assessments. As students move to online learning, efforts have been made to streamline instructional and assessment software to reduce distractions. Prior behavioral research has found that visible timers on online assessments are a source of anxiety that can influence performance and outcome. Additional studies on color theory suggest that timer color may be another factor. Through literature review, timer UI design and implementation, and voluntary behavioral studies of undergraduate subjects taking CS 2110 (Software Development Methods) and CS 4750 (Database Systems), we plan to analyze the correlation between various types of virtual timers and exam performance.

STS Research Summary:
This STS research investigates arguments for and against the use of predictive policing technology in the criminal justice system. In the interests of improved efficiency and reduced human error, crime assessment algorithms and other machine learning models are increasingly used to inform societal regulations. However, some researchers assert that historical crime data is biased due to overpolicing, meaning this approach does little more than reinforce systemic problems like racial profiling and the disproportionate targeting of low-income neighborhoods.
The human and social dimensions are important in this research because of the social context that needs to be considered in developing technology that serves public safety and affects the livelihoods of citizens. The Social Construction of Technology (SCOT) framework will be used to analyze the stakes that various social groups have in predictive policing technology. Under the framing of interpretive flexibility, the central tenet of SCOT, I studied each stakeholder's relationship with the technology, the problems they need addressed, and potential solutions this technology can bring.
To conduct this STS research, I analyzed arguments from both sides of the predictive policing debate through a study of case law. I surveyed federal and state court cases from the past ten years (2010-2020) to identify relevant cases, compiled arguments from case briefs, and summarized the current legal boundaries of predictive policing technology. Over time, case law seems to be shifting from more permissive of algorithm usage to more restrictive, as in the landmark case U.S. v Curry (2020). I found that the courts are more favorable towards the use of risk assessment tools like COMPAS to inform sentencing, and are more critical of the use of predictive policing tools, for example, crime hotspots or "heat lists," by law enforcement personnel. Higher courts also tend to prioritize transparency from government agencies.
Through this research, I observed judicial opinions shifting over time towards skepticism of crime prediction algorithms, as more research cast doubt on their efficacy. When considering predictive policing technology and STS research in concert, any efficiencies the technology brings need to be weighed against its adverse effects on civil liberties. While crime prediction and risk assessment algorithms may expedite police work, use of historical crime data without consideration for social context means that this technology may require further development to be effective.

BS (Bachelor of Science)
predictive policing technology, crime risk assessment, machine learning algorithms, case law, court cases, legal, law enforcement, police, justice system

School of Engineering and Applied Science
Bachelor of Science in Computer Science
Technical Advisor: Nada Basit
STS Advisor: Rider Foley
Prospectus Advisor: Peter Norton

Issued Date: