Legal limits on facial recognition technology

/in /by

What limits, if any, should federal , state, and local governments impose on the use of “facial recognition technology” (FRT) by law enforcement? For
instance, given FRT’s demonstrated accuracy problems – at least three people have been falsely arrested because of faulty FRT identifications- should any
use of FRT by law enforcement be barred until the FRT is demonstrably more accurate? Or, should FRT be used only when law enforcement officers have a
search warrant, backed by a finding of probable cause? Should there be an exception to any warrant requirement for use of FRT by DHS officers at the
borders? Or should DHS officers at least be required to document “reasonable suspicion” when they use FRT at the borders? Should the FRT databases be
limited to mugshots of those who have been arrested, or to photos from motor vehicle databases? Should use of FRT from companies such as Clearview AI
be barred altogether because their databases are developed through “web-scraping” (II, below) which allegedly violates state “biometric information privacy
laws” and the Federal Computer Fraud and Abuse Act (CFAA)? Would such a bar unlawfully infringe on Clearview’s First Amendment to Free Speech ? The
use of FRT by law enforcement obviously raises important legal and public policy issues. Please write a paper (10-15 pages) in which you analyze the issues
and explain your view of how FRT use by law enforcement should be regulated. ## Slides 11-34 of the Week 7 Lecture -AI and Decision-Making in the
Criminal Justice Process- may help get you started in your research. But, use FRT is one of the hottest issues in the law right now (May 2021). New cases,
articles and proposed legislation on FRT are appearing regularly.

The following slides are based on  law review articles by Kiel Brennan-Marquez,  Professor at the University of Connecticut School of Law : “Plausible Cause”: Explanatory Standards in the Age of Powerful Machines (Plausible Cause) 70 Vand. L. Rev. 1249 (2017) ; Emily Berman, a Professor at the University of Houston Law Center; A Government of Laws and Not of Machines, 98 B.U. L. Rev. 1277  (October 2018) (“Laws Not Machines) and Andrew Guthrie Ferguson,  a Professor at the UDC Clarke School of Law:  Policing: Predictive Policing,  (Policing) 94 Wash. U. L. Rev. 1109 (2017)

“PREDICTION TOOLS’’  

  • Police department are increasingly relying on “predictive policing software” identify and target: places where crime is likely to take place; “likely perpetrators of gun violence” and likely victims of gun violence. Policing, pp. 1113-14

oData on past crimes are “fed into computer algorithms” to predict where crimes are likely to take place in the future. Policing, p. 1113

  • “Place-based” use of computer algorithms to predict sectors of a city where crime are likely to take place have long been recognized as an efficient and effective means to help police prevent crimes before they are committed. Policing, p.1132

oThese “prediction tools” (e.g.  PredPol) have been used to help to identify locations where gang violence and  other violent crimes are likely to take place.

v  “Person-based”- Scholars and legislators have become increasingly concerned as the use of complex algorithms have been used to target individuals (“person-based”)who who are:

  • 1. Likely to commit violent crimes(e.g., domestic violence); or
  • 2. Likely to violate bail conditions; or
  • 3. Likely to reoffend once released from prison. Policing, pp. 1137-43
  • vLeading Case Discussing Use  of “Person-Based Prediction Tools” 
    • As of Spring 2021, there were few reported cases that discussed the use of prediction tools in the criminal justice system.

    v Leading Case  – In fact, the leading cause is still (March 2021) State v. Loomis, 881 N.W.2d 749 (2016)- a case in which the Wisconsin Supreme Court upheld the trial judges’ use of the prediction tool COMPAS(“Correctional Offender Management Profiling for Alternative Sanctions) but set explicit limits on how COMPAS was to be used.

    v  In Loomis,  the Court explained that COMPAS can be used in:

    o1. Deciding at sentencing whether a convicted defendant is likely to reoffend (“risk of recidivism”);

    o2. Deciding pre-trial whether a person charged with a crime  is likely to comply with court-ordered “conditions of release”;

  • o3. Deciding whether an individual is likely to engage in violent conduct upon release.  881 NW2d 754
    • The COMPAS report, which was attached to the Department of Corrections pre-sentence report to the sentencing judge (PSI), predicted that the Defendant Loomis, was a high-risk respect to each of the three of the criteria: (1) likely to reoffend; (2) likely to violate release conditions and (3) likely to engage in violent conduct. 881 NW2d 755
    • The trial judge used the COMPAS report in making his decision to sentence Loomis, who was a repeat offender, to prison (2 years) rather than probation.

    v The judge emphasized that the COMPAS report was not “determinative” in his decision to sentence Loomis to prison.  881 NW2d 753-54