States are using tech to wipe criminal records clean automatically

One in three people living in the United States has a criminal record, and while almost every state has a process to clear those records after a certain period, doing so can be challenging.

Petitioning a state can be complex and require lots of burdensome paperwork. It also might involve expensive court fees, or even require hiring an attorney to navigate the process. Waiting periods to get convictions expunged can be years long, too. Meanwhile, the collateral consequences of having a criminal record can be enormous: People may be unable to live in certain housing, get a job or reunite with their child. That can lead to re-offending.

Some state governments have looked to buck the trend and help their residents clear their criminal records more easily. Twelve states — California, Colorado, Connecticut, Delaware, Michigan, Minnesota, New Jersey, New York, Oklahoma, Pennsylvania, Utah and Virginia — have passed fully automatic record clearance policies, meaning around 15.6 million people will get full or partial relief as a result.

Seven states — Illinois, Kentucky, Maryland, Massachusetts, Missouri, Ohio and Texas — are in various stages of working on policies of their own, known in many places as “clean slate laws.” Making them automatic puts the onus on the government to do the work, rather than rely on the petitioner to navigate what the National Conference of State Legislatures said can be a “baffling and cumbersome” process and is typically free.

Code for America, the civic tech nonprofit that works to deploy new technologies to government, has been at the heart of the push towards automatic records clearance and embraced tech in doing so. The organization started its work in this area in 2016 and helped more than 20,000 people get relief in some California counties, then expanded in 2018 as the idea of the government playing a more active role in helping residents clear their records started to take hold.

The biggest step is having the technology flag someone’s eligibility and determining what records need to be cleared. That requires a comprehensive view of someone’s criminal history, which can be challenging, as systems are not organized by individual people, but by cases. New entries into the court system, even if it is a repeat offender, are entered as new cases.

“Anytime I have a new entry into the legal system, I have a new case generated,” said Alia Toran-Burrell, program director for Code for America’s Clear My Record initiative. “That means there's a ton of duplicated cases, there are duplications of information about one person. In order to determine eligibility, we need to have a way to de-duplicate those parties within a system.”

That requires entity resolution, which reconciles criminal cases for the same person and helps deal with common problems like misspelled names and the different formatting that law enforcement and criminal justice agencies use in their case management.

Code for America uses machine learning for what Chris Correa, the organization’s director of solutions engineering, called “probabilistic entity resolution” to calculate the probability of whether different records refer to the same entity or person.

“There's a list of what we believe are the actual people, actual records, and then a link to this is related to these three records that may have differences in how they're described, and a map to figure out which records relate to which people, and then they integrate that into the eligibility determination process to figure out who's eligible for the clean slate process,” Correa said.

For administrative staff, using technology has sped up the records reconciliation process, similar to other states that have used it to make employees work more efficiently. Correa said Code for America used the method to find 3 million individual records in Utah, then de-duplicated those records to connect them with 1.2 million people who had interacted with the courts. Those records can then be parsed to determine eligibility for automatic records clearance.

Correa said it would have taken a small team of court staff over a decade to manually review records and link the matching ones. Using machine learning, it takes a matter of hours.

Research across states with automatic record clearance and clean slate laws has found significant impacts. A report from the Clean Slate Initiative, which advocates for the policy, found that when people’s records are cleared, various aspects of their lives improve, including their employment prospects, housing and health.

“People who get record clearance earn more, are able to support their families better, are able to participate in their local economy more effectively,” said Toran-Burrell. “There are big benefits to economies when people are able to actually participate.”

The push for automatic records clearance and more tech is part of a wider effort to help ease some of the burdens of the criminal justice system. The Los Angeles County Public Defender’s Office, for example, has digitized many of its records and is looking to use AI and automation to help provide effective defenses.

And while experts say AI cannot be relied upon too heavily in such sensitive work, it can help with some of the more mundane tasks like finding information and making the first draft of certain court documents.

“There's a way that AI means that information can be accessed and used helpfully, rather than arbitrarily,” said Jesse Rothman, a senior fellow at the Council on Criminal Justice think tank, who leads its AI initiative. “There's a way that has the potential to really make things more reasonable, more accessible, democratic. That possibility is very exciting.”