Justice Driven Data Science for Prosecutorial Impact
In the next 100 years, what if California could fundamentally disrupt the pipeline to incarceration, ensuring justice and equity in criminal prosecutions?
The use of data to inform decisions is a key tenet of “progressive prosecution,” a pathbreaking approach to criminal prosecution that seeks to reverse the role of prosecutors as drivers of mass incarceration by shifting the goals of prosecution from conviction and punishment to fairness and accountability. In partnership with DA offices across the state of California, the California 100 Innovation Team and The People Lab at UC Berkeley are undertaking a research and demonstration project that will pilot the use of new empirical methods to assist prosecutorial decision making in the pursuit of justice.
Case data are often complicated amalgamations of arrest circumstances, administrative records, police report narratives, body camera footage, and data elements from other record-keeping systems from multiple jurisdictions. While working within this data ecosystem and meeting quick turnaround deadlines for decisions, Assistant District Attorneys (ADAs) must parse and process all relevant information pertinent to whether a case can be charged or is eligible for diversion.
By creating a structured dataset of case characteristics and using data science methods to build a decision aid for use by ADAs in their case reviews, we will pursue three distinct but interrelated goals:
- first, to reduce the prosecution of cases that pose a low risk to public safety and thus are shown by a growing body of research to be more effectively addressed without formal prosecution;
- second, to reduce DA caseloads through earlier and more effective identification of cases for discharge or diversion;
- and third, to improve DA’s ability to prosecute high priority cases that pose a significant risk to public safety.
This pioneering effort, if successful, has the potential to transform prosecution across California and the United States by demonstrating a replicable model of transparent, fair, and just data-driven prosecution.
Integrating a Fractured Safety Net
In the next 100 years, what if California could completely redesign the implementation of social welfare programs, to ensure all Californians can benefit from social programs?
In California, low-income households are eligible to access a patchwork of programs designed to provide a supportive safety net. However, the fragmented nature of safety net administration in the state results in these programs falling short of their true reach, ultimately leaving millions of households underserved. This is especially true when it comes to programs like the Child Tax Credit, the Earned Income Tax Credit, and other pandemic-related aid delivered through the tax code. Many families that are eligible for these anti-poverty benefits are unaware of their eligibility to receive this type of aid, or do not file taxes at all.
The California Department of Social Services (CDSS) and the California Policy Lab (CPL), in collaboration with The People Lab (TPL) and California 100’s Innovation Team, are using administrative data to integrate the process of delivering tax benefits to low-income Californians. This aid is projected to lift millions of Californians out of poverty–provided that eligible households actually receive the aid.
Californians must file taxes to receive these benefits, but the administrative burden of filing taxes is large: filing taxes can be costly and complicated for any Californian, but especially so for the most vulnerable. In the first phase of this pilot project, we will pilot a new SMS-based method for improved targeting of eligible individuals, relying on linked administrative data across government agencies. This model, if it works, has the potential to fundamentally transform how multiple government agencies can work together to ensure access to and uptake of benefits by targeting outreach to those who are least likely to claim and who stand to gain the most from tax benefits of this kind.