Algorithmic and technological solutions to enhance criminal justice resentencing in California

• In Event: Lightning Session: Crime, Justice, and Drugs

Abstract

In 2018, California became the first state in the US to pass a Prosecutor-Initiated Resentencing (PIR) law, which grants prosecutors the power to revisit past convictions . In addition, there are a number of resentencing and second-look laws in the state, such as AB600 and the Racial Justice Act, which seek to correct past injustices in the criminal legal system. But given that there are over 100,000 people in California's prisons, lawyers and county officials need efficient and accurate ways to identify cases worthy of a second-look instead of manually reviewing piles of paperwork. Prof. Colleen Chien at Berkeley Law coined the term "second-look gap" to capture the distance between the opportunity and implementation of second-look interventions. We estimate that there are hundreds languishing in California's prisons as a result of technological and political barriers. Redo.io (www.redoio.info) is solving this problem with a series of predictive and rules based models for resentencing eligibility determination. This presetation will deep-dive into our technology solutions and the impact we have had thus far. We have identified over 450 candidates that are serving excessive sentences for non-sexual, non-violent and non-serious offenses using an open database of prison sentences we acquired via a public records request to the Department of Corrections. These candidates have been referred to individual public defender offices through the Office of the State Public Defender. In addition, we have released an open-source AI assistant for attorneys to quickly query our dataset and find prison sentences that align with their criteria and investigate their case files further. Our predictive models embody the principles of fairness, accountability and transparency and help prioritize cases based on violent history, commitment to rehabilitation (education, mental health treatment), etc. Legal experts have helped design and fine-tune eligibility rules in an iterative review process. Our model therefore augments legal knowledge about the applicability of PIR as opposed to replacing such expertise. Redo.io has deployed the eligibility model for the Three Strikes Project at the Stanford School of Law (https://law.stanford.edu/three-strikes-project/), a legal clinical seminar where law students represent individuals serving life sentences under California’s Three Strikes Law. By supplementing a lawyer’s decision to examine a case for resentencing instead of automating the end-to-end process, our model helps stakeholders make high-impact decisions efficiently while preventing existing biases in our society from being exacerbated. It therefore exemplifies a framework for meaningful human-computer interactions in high-stakes and high-risk decision making processes.

Author

• Aparna Preeti Komarla, Independent
  Presenting Author