By Colleen V. Chien, Founder, Paper Prisons Initiative and Professor, Berkeley Law School
One in three adults has a criminal record. As such, one of the most important and unresolved questions in criminal justice policy is whether people can move beyond their past records and re-integrate into society. The evidence suggests that they can — when people clear their criminal records (via petition), they experience gains in income. The realization that millions eligible to get their records cleared have not done so, due to the “second chance gap,” in expungement, has led states to pass automated expungement laws, under The Clean Slate Initiative, and the founding of the Paper Prisons Initiative.
But do Clean Slate” policies work? To date, a growing body of careful criminal justice research — including work by Amanda Agan, Partha Dasgupta, Randi Hjalmarsson — has failed to find the same positive labor effects of petition-based clearing to be associated with automated clearance.
Yet a question remains, at least in my mind, due to the fact that despite automated clearances, people often don’t know about them: how can people be expected to change their behavior if they don’t know they are free and the opportunities that are available to them as a result?
This question is at the heart of Paper Prisons’ Digital Juneteenth research agenda, which treats notification not as an administrative afterthought, but as a core policy mechanism. Read our agenda here. To execute on it, we have entered into a partnership with the County of Santa Clara Public Defender’s Office to carry out rigorous randomized control trials to test notification strategies (attempting to engage in allegory-of-the-cave thinking). Though our work is just beginning, we share some early results from a very small pre-pilot (N=63) as a way to encourage others to share what they are doing in this space as well and to cultivate a broader conversation about notification in an automated age.
Our setting
In California, SB731 clears arrests and most felony and misdemeanor convictions after a waiting period of 1-4 years. As a result of this transformative law and its predecessor, AB1076, millions of records have been cleared, at least at the state level. Santa Clara County is unique because it has also cleared records at the county level.
In the fall of 2025, several members of the Paper Prisons Initiative Dylan Margolis (‘26, Data Science) and Magali Urresti (MSW ‘26) embedded in the Santa Clara County Public Defender’s Office to start to develop the infrastructure to notify as many of the 130K people (with additional clearances happening every month) receiving automated expungements. At full scale, this would require a roughly 2000× increase in outreach capacity from the 63 person pre-pilot just carried out —almost certainly involving AI, with all the governance questions that raises.
But before scaling, we wanted to understand something more basic: what actually happens when people are told they’re free?
Early signals
In our pre-pilot, implemented for operational and methodological learning, we sent an email notification to people whose records had already been cleared. These emails were simple, sent from the Public Defender and identifying the cleared records. But they were also behaviorally informed. Just as previous work has shown that contacted people are more likely to vote if they make a plan, our notifications provided three follow up options for engagement, as described below.
Among non-bouncing emails (N=50):
- 42% engaged (sending a response email to the Public Defender)
- Of those who engaged, 57% asked for one of the three additional services— employer letters, help with entire record, or answers to specific questions
These numbers aren’t generalizable. The sample is small and selected. But they suggest that notification, done thoughtfully, may unlock something that clearance alone does not: engagement.
Notification + Engagement
What we found interesting wasn’t just the percentages, but also the survey responses (33% response rate). For example, one respondent (“Danny,” name changed) wrote:
“This is awesome of you and your team. It will open up jobs and I want to go back to school.”
When asked what expungement meant to him, he said:
“My ability to get a place to live… Can’t get a real job and this should help a lot.”
This suggests the possibility of labor attachment, not as an abstract outcome variable, but as lived experience. Asked what he planned to do next, Danny said:
“College. I want to do dental assistant work… I like putting smiles on peoples’ faces.”
This is a potentially promising indicator that engagement on records can set into motion other types of engagement. Santa Clara County is one of the wealthiest counties in the nation, yet access to its labor markets often hinges on credentials and educational pathways. That this respondent’s newfound sense of possibility translated into interest in education—he separately asked for a help to connect with dental assistant colleges —suggests not only hope, but a willingness to invest in one’s own human capital. In fact the County and State do have potentially relevant educational offerings, such as the Rising Scholar community college program. Credit for the ability then to connect Danny to a social worker for follow up is owed to the Public Defender’s Office’s holistic approach to post-conviction services.
Also notable and unexpected was an indicator of Danny’s impression of the Public Defender’s Office:
“We used to call them the public pretender… and because of this that changes.”
If this finding replicates, it may create greater buy-in from other public defender offices as well, to follow Santa Clara’s lead and invest in notification.
Scaling Up
As we scale the work, we will be paying attention to these indicia of engagement put in motion by our notifications, which are not accounted for by current studies. Our pre-pilot also remind us why mixed methods matter. Here quantitative design will help us measure scale and effect. But the qualitative responses we are receiving can help us understand meaning, motivation, and momentum.
We’re just at the beginning of our project and big questions lie ahead, including in terms of raising funds to scale and access a variety of administrative and other outcomes. But these early signals suggest that the promise of Clean Slate may not lie only in automation—but in what happens after people learn they’re free. Visit our website and subscribe to updates to follow along on the journey and tell us your testable hypotheses as well, as to how an initial bit of good news can set off a longer term process.
Read more about our partnership with the Public Defender and our Digital Juneteenth Research Agenda.
With thanks to Paper Prisons partners Santa Clara County, Santa Clara County Public Defender’s Office, the California Policy Lab, and Paper Prisons team members and project advisors: Jake Rhodes, Dylan Margolis, Alonzo L Harvey, Janet Schwartz (Duke Center for Advanced Hindsight), Dylan Martinez , Kyla Bourne, JJ Prescott, Al Barrentine, Johanna Lacoe, Iris Cisneros.
