The first automated text messages have just gone out to dozens of Santa Clara County public defender clients to remind them when to appear for court hearings. The new technology, developed by Santa Clara County’s Office of the Public Defender and the Stanford Computational Policy Lab and supported by Stanford Impact Labs, is designed to help public defender clients make their court appointments and reduce the chances they will be held in jail for months or years until trial for failing to appear.
The initiative builds on research from New York City that showed simple, low-cost behavioral nudges like text message reminders can help people involved in the criminal justice system appear in court for required hearings. When people accused of a crime fail to appear for court hearings—which can be due to lack of transportation, childcare, and other support—judges can issue warrants. Individuals who may be legally innocent are then arrested and held in jail to ensure they appear in court. This kind of pretrial incarceration has significant financial costs for counties and agencies in the criminal justices system; it can cost those accused of a crime their jobs, housing, education, and families.
A novel, funded partnership
In the randomized evaluation in New York City supported by the Abdul Latif Jameel Poverty Action Lab (J-PAL), researchers found text message reminders reduced failures to appear in court by up to 26 percent, translating to thousands fewer arrest warrants per year. Santa Clara County’s Office of the Public Defender had been looking for text messaging products they could use for similar purposes. In September 2019, Charlie Hendrickson and Sarah McCarthy, from Santa Clara County’s Office of the Public Defender, heard Stanford Computational Policy Lab’s Sharad Goel present at a conference on ways technology can respond to social problems, including social justice issues. When Hendrickson and McCarthy approached Goel after the conference, they learned his Stanford team was already working with The Bail Project to develop mobile app technology that could be used for sending text messages.
Together, Santa Clara County’s Office of the Public Defender, The Bail Project, and the Stanford Computational Policy Lab applied for start-up funding from Stanford Impact Labs. They were one of five teams awarded $500,000 over two years to support the project to design, build, and test how open-source technology platforms can improve pretrial outcomes for thousands of individuals across the country every year.
“The Stanford Computational Policy Lab was part of the project from the beginning,” said Santa Clara County’s Assistant Public Defender Sarah McCarthy. “They were already working on developing the mobile app technology for The Bail Project, and the timing was perfect for us to join in the venture. We needed a mobile app, and they needed partners in developing and testing the technology as well as data to study. It was a total win-win.”
Financial support from Stanford Impact Labs covers the majority of the project and partnership. Santa Clara County will pay only for the cost of sending the individual text messages, as well as for the time and energy their staff spend on the project. Stanford Impact Labs’ funding also supports project management, something Stanford Computational Policy Lab Executive Director Alex Chohlas-Wood says is often underrated but critical. “Project management and logistics take tremendous effort,” he said, “and projects often fail when project management is done informally.”
Creating public impact and putting research into practice
The team prioritized sending out reminders to people because they expect it will have the biggest impact on reducing the number of people held in jail until their trial. “The automated reminders are really important,” says Stanford’s Chohlas-Wood. “Often clients of the public defender’s office tend not to keep their own calendars and their schedules are in flux. We think that reminders seven days before an appointment, three days before, and the day before will be super helpful for people to just make it to court.”
Missed court appearances have devastating costs for public defender clients. “When our clients are put in jail, even for a few days,” Assistant Public Defender McCarthy says, “they may lose their jobs, housing, benefits, educational opportunities. Their children may be taken from them, their cars may be impounded. The collateral consequences are many, and all have significant financial as well as mental and emotional impact on our clients.”
"When our clients are put in jail, even for a few days, they may lose their jobs, housing, benefits, educational opportunities. Their children may be taken from them, their cars may be impounded. The collateral consequences are many, and all have significant financial as well as mental and emotional impact on our clients."
The team also has an opportunity to increase their impact by incorporating research to understand what kind of reminders are most effective for people. Assistant Public Defender McCarthy says “Perhaps the biggest contribution from Stanford comes from the fact that they are not just a tech outfit, but part of a renowned research facility. Stanford can crunch the data to help us understand the impact of this technology and help us make adjustments to make it more effective.”
Stanford’s Chohlas-Wood says their research team—which includes Emma Brunskill in Computer Science and Guillaume Basse in Management Science and Engineering—will test various message personalizations that may increase effectiveness. For example, the team might test a “honey versus vinegar comparison in the type of message content.” One message might say something like “Don’t forget you have a court date in a week, here are some resources to help you get to court, let us know what we can do to help.” Another message might say, “Don’t forget to show up to court, if you don’t show up, here are the potential consequences, it could be really bad.”
Instead of guessing which approach might be best or testing one at a time, the team will use reinforcement learning techniques, a kind of machine learning meant to figure out what works for different people. Individuals opt into text message reminders. Names and personal identifying information are removed from data sets used for research.
Chohlas-Wood says, "The team is excited to put research into practice to demonstrate that the advanced techniques that have really only been shown to work and deployed in Silicon Valley big tech companies to make better advertising dollars can be used in other domains where they can create better outcomes for everybody.”
"The team is excited to put research into practice to demonstrate that the advanced techniques that have really only been shown to work and deployed in Silicon Valley big tech companies to make better advertising dollars can be used in other domains where they can create better outcomes for everybody."
“There are so many variables,” McCarthy says of why her clients miss court appearances, “and studying the nuances is what we hope may reveal how best to get people to court so we serve them better and they avoid the stress and harm from unnecessary jail time.”
The automated text message reminders are the first phase of the team’s work together. They expect to build and use other technology to help public defender clients communicate with their attorneys and coordinate transportation to court appearances.
The team is also excited about the potential for other counties and public defenders to use the tools they are building and learn from their work. The technology they have built is designed to be generalizable. “We already built it to work for two partners with two different systems and use cases,” Chohlas-Wood explained of their work with The Bail Project and Santa Clara County’s Office of the Public Defender. He hopes they have created a tool that can be used elsewhere. The team is already speaking to another major US city office. And they are thinking of building a general use website that any small office could use without building a full integration into their own office systems.
Chohlas-Wood says they also see the need for automated reminders coming up in different contexts, like diversion programs that enroll people in treatment programs instead of charging them with crimes. Those programs also come with court appearance obligations so there is potential demand for similar automated reminder technology.
With encouragement from Stanford Impact Labs, the team may also explore the potential ongoing use of online court appearances—something many counties were forced to use during the coronavirus pandemic—to increase court hearing attendance and reduce pretrial incarceration. This is one example of the new ideas and opportunities Stanford Impact Labs aims to support by funding long-term partnership and trusted working relationships between academic researchers and external practitioners.
“This is the kind of collaboration we need more of in the county,” says Assistant Public Defender McCarthy. “Working with the Stanford team has been such a pleasure and is allowing us to tackle a big social problem in a new and innovative way.”
While each member of the project brings their own experiences, expertise, and enthusiasm for different aspects of the work, they are united by a common drive to use their time and energy to solve a major social problem and help people avoid unnecessary jail time.