Contra Costa County District Attorney's Office Makes Unbiased Charging Decisions with ScaleCapacity Generative AI Solution on AWS

Contra Costa County is home to an estimated 1.16 million residents and is located in the East Bay region of the San Francisco Bay Area of Northern California. The Contra Costa County District Attorney's Office serves these residents by working to prevent crime, seek justice, and enhance public safety by prosecuting those who violate the law. On January 1, 2024, California enacted California Penal Code Section 741, also known as Assembly Bill (AB) 2778, which mandates that all state prosecution agencies must implement a Race-Blind Charging process, requiring the redaction of race information from police reports and other documents before making charging decisions. The new law aims to eradicate racial bias and promote a more fair and equitable charging process for all individuals, including minorities who have historically faced disproportionate targeting in the justice system.

The District Attorney’s Office was concerned about meeting the January 1, 2025, deadline for complying with the law’s guidelines. “We only had about 10 months to create a new solution that automatically redacts sensitive data from case materials, and we wanted to use our existing staff to do that,” says James Mount, information systems manager at the Contra Costa County District Attorney's Office. “We needed to evaluate technology and move from development to testing to production in that short timeframe and do it all cost-effectively.” The District Attorney's Office also needed to ensure that any new solution aligned closely with its complex workflow. “For each case, a clerk uploads a police report or file into an application, and then attorneys review it and determine if charges will be filed,” says Mount. “It’s a time-consuming process, and we didn’t want to burden our attorneys with more work redacting data.”

To overcome its challenges, the District Attorney's Office selected ScaleCapacity, an AWS Partner and consulting firm that specializes in justice and public safety solutions. ScaleCapacity is an AWS Public Sector Partner that holds an AWS Generative AI Competency. “We process around 17,000 cases annually and using a manual process wasn’t feasible,” says Mount. “We wanted to adopt an AI-driven solution to automate everything. Even though we evaluated a few different technology companies, ScaleCapacity had already helped us implement a transcription service, which was a success for us. That made it easy to choose to work with them again. Also, ScaleCapacity could tailor our workflow to easily incorporate the requirements of the upcoming law, which we knew would make the implementation achievable ahead of the deadline.”

ScaleCapacity designed and implemented a Race-Blind Charging solution based on Amazon Web Services (AWS) and using advanced AI and natural language processing (NLP). The solution uses Amazon Bedrock to automatically redact sensitive information from case materials, ensuring that race, ethnicity, and other identifying details are removed. Amazon Bedrock also ensures privacy by anonymizing sensitive data in case documents. Amazon Textract, a machine learning (ML) service, extracts text from PDF case documents and converts the documents into machine-readable text for further processing. Amazon Simple Storage Service (Amazon S3) stores all case-related documents and acts as a centralized location for document retrieval and storage, while Amazon DynamoDB stores metadata about cases and user information and ensures fast, reliable access to case and end-user details. Additionally, Amazon Cognito manages user authentication and authorization for secure access, and Amazon Simple Email Service (Amazon SES) emails users when cases are assigned. The AI-driven solution can dynamically evolve with minimal code changes, with any new models or other updates implemented in hours.

The solution is closely aligned with the office’s existing case processing workflow, simplifying the process of uploading police reports and other data while automatically redacting sensitive data without disrupting the workflow users are already familiar with. “Previously, there was a disconnect in the process and attorneys had to spend too much time looking at the data and redacting it,” Mount says. “Now, with the ScaleCapacity solution on AWS, attorneys have a view of a case and can make a determination about charges before having to go into another internal system.”

By working closely with ScaleCapacity and implementing the Race-Blind Charging solution on AWS, the District Attorney's Office deployed an automated solution in six months, faster than the expected ten months. “Before we started working with ScaleCapacity, we were behind the curve in identifying what we needed to implement by the deadline,” says Mount. “By working with ScaleCapacity and the AWS team, we were able to meet the requirements of Penal Code Section 741 AB 2778 by the January 1, 2025, due date. Doing it that quickly is a testament to ScaleCapacity. There is a lot of statistical collection and reporting to meet the requirements, and we were able to complete it all with this solution. We were also able to use our existing headcount and implement this solution with minimal training.”

The Race-Blind Charging solution streamlines the redaction and evaluation process, reducing the time required for manual redaction and minimizing human error to ensure accurate results. By using the solution to evaluate and review unredacted reports, county prosecutors can make unbiased preliminary decisions. “Race-Blind Charging is a great addition to the justice process, and we are pleased to support it while meeting the state’s objectives,” says Mount. “We continue to fine-tune the solution with subject matter experts and attorneys. We’re still testing, but we are seeing steady progress during every meeting with ScaleCapacity.”

The District Attorney's Office continues to work with ScaleCapacity to enhance the Race-Blind Charging solution. For example, ScaleCapacity created a feature that allows users to toggle back and forth between redacted and unredacted document views. “Previously, in an unredacted view, we wouldn't know what was redacted or not, so an attorney would have to toggle back to the redacted view. It was quite cumbersome,” Mount says. “ScaleCapacity came up with this enhancement in just one week, which was amazing. It was out of the scope of the original project, but they were still happy to help us with it. ScaleCapacity has also helped us identify some workflow issues to reduce some of our processing charges.” Vijay Musaligari, head of data analytics and AI/ML at ScaleCapacity, adds, “This has been a great partnership. The District Attorney’s Office trusts us and listens to our ideas, even without seeing the full end result. We look forward to continuing our work together.”