10.1 Dressel and Farid (2018): Predicting recidivism

  • Q: Whats the topic of this study?

  • Dressel and Farid (2018): “The accuracy, fairness, and limits of predicting recidivism”
    • “In the criminal justice system, predictive algorithms have been used to predict where crimes will most likely occur, who is most likely to commit a violent crime, who is likely to fail to appear at their court hearing, and who is likely to reoffend at some point in the future.”
    • Compas has been used to assess more than 1 million offenders since it was developed in 1999 (since 2000 recidivism prediction component).
      • “software predicts a defendant’s risk of committing a misdemeanor or felony within 2 years of assessment from 137 features about an individual and the individual’s past criminal record.”
  • What is the research question?
    • General: Are algorithms better in predicting than humans?
    • Specific: Does the COMPAS algorithm perform better than humans in predicting recidivism?
  • What is the hypothesis?
    • They are worse/better/as good as
  • What data do they use?
    • Database of 2013/2014 pretrial defendants from Broward County, Florida
      • 7214 defendants with individual demographic information, criminal history, the COMPAS recidivism risk score, and each defendant’s arrest record within a 2-year period following the COMPAS scoring
        • COMPAS scores, ranging from 1 to 10, classify the risk of recidivism as low-risk (1 to 4), medium-risk (5 to 7), or high-risk (8 to 10)
      • Algorithmic assessment based on full set of 7214 defendants
      • Human assessment was based on a random subset of 1000 defendants, which was held fixed throughout all conditions
  • What is their finding?
    • “people from a popular online crowdsourcing marketplace - who, it can reasonably be assumed, have little to no expertise in criminal justice - are as accurate and fair as COMPAS at predicting recidivism”
    • “Collectively, these results cast significant doubt on the entire effort of algorithmic recidivism prediction”
    • COMPAS software is equivalent to a simple linear classifier (age + total number of previous convictions)
  • Q: What struck you about this study? What did you find interesting?
    • What do the authors mention about race (Tip: Correlates with..)?
    • What is the outcome variable in the study? What are false positive and false negatives in this context?
  • Insight: Widespread use of questionable prediction/classifcation algorithms

References

Dressel, Julia, and Hany Farid. 2018. “The Accuracy, Fairness, and Limits of Predicting Recidivism.” Sci Adv 4 (1): eaao5580.