An End to CPAT?

Controversy and indecision regarding the future of the cash bail system is never-ending. Some state lawmakers have done away with it altogether. Some jurisdictions put their faith in electronic tethers for post-release monitoring or computer algorithms for pretrial assessment. And some have done nothing, continuing to rely on pretrial services programs, own recognizance rulings, and cash bail bonding for defendants awaiting their next court date. But what works and what doesn’t? And how will decisions be made going forward?

An End to CPAT?

In Colorado, the bail bonds industry has been warning the public against the use of the Colorado Pretrial Assessment Tool (CPAT) since it was rolled out in 2012. The tool was designed to eliminate racial bias from rulings about bail and pretrial release from detention. But of course, it’s failed pretty miserably for certain racial demographics, particularly Black and Hispanic criminal defendants.

A two-year study of the CPAT, conducted by the University of Northern Colorado, confirmed the worst fears of not only bail bondsman, but of social justice advocates across Jefferson County, Arapahoe County, Denver County and elsewhere throughout Colorado. Pretrial risk assessment tools, now widely used throughout the American judicial system, are an abject failure which need to be re-assessed.

As a quick refresher, pretrial risk assessment instruments (RIAs) are costly statistical and machine learning tools utilized in pre-trial judgments to decide two things: whether a defendant will commit a new crime when released, and whether he’d disregard his next court date. Two of the most popular are the Public Safety Assessment and the Correctional Offender Management Profiling for Alternative Sanctions.

According to a report in The Brookings Institution called “Understanding risk assessment instruments in criminal justice,” RIAs are controversial and failure-prone in four regards:

“Critics have focused on four main concerns with RAIs: their lack of individualization, absence of transparency under trade-secret claims, possibility of bias, and questions of their true impact.”

A Way Forward

Much of the controversy revolves around a seemingly nationwide push to do away with the concept of cash bail, where a judge tells a defendant that she or she must post x-dollars to get out of jail pretrial. Critics have argued the practice discriminates against minorities and the poor, causing prison populations to swell. New York, New Jersey, Alaska, and California have replaced their cash bail systems with other mechanisms, but the results haven’t been what supporters asked for. Some defendants, once released, have returned to a life of crime. Colorado’s effort at bail bonding reform has had less than stellar results in Denver and other counties, with some groups calling for an end to the use of RIAs.

We’re supportive of that effort but believe it may take years for a new bill to get written, approved by lawmakers, and signed into law by Colorado’s governor.

So What Can We Do in the Interim?

Let’s get back to that Brooking Institutions report. It offers five intriguing recommendations:

1. “First, policymakers should preserve human oversight and careful discretion when implementing machine learning algorithms.
2. “Second, any algorithm used in a high-stakes policy context, such as criminal sentencing, should be transparent. This ensures that any interested party can understand exactly how a risk determination is made, a distinct advantage over human decision-making processes.
3. “Third, algorithms, and the data used to generate their predictions, should be carefully examined for the potential that any group would be unfairly harmed by the outputs.
4. “Fourth, data scientists should work to build next-generation risk algorithms that predict reductions in risk caused by supportive interventions.
5. “Finally—and perhaps most important—algorithms should be evaluated as they are implemented.”

As veterans of the bail bonds industry, we’ve learned to not get our hopes up.

What’s the Response?

Clearly, bail bond reform isn’t what it’s cracked up to be. Haphazard laws and judicial discretion policies were implemented based on the tug of emotions and sensationalized reporting. The track record of defendants who’ve been released without cash bail is pretty dismal, but some lawmakers still don’t know what to do to. In California, lawmakers are all-in when it comes to math and computer algorithms, pushing for Proposition 25, which would reform that state’s budget process and replace cash bail with RIAs.

We suspect that many a bondsman would vote No against Proposition 25 if given the chance. Too many lawmakers have put too much faith in RIAs because these tools are new and sexy, without taking the time to realize that nothing’s perfect.

It’s amazing how many people fall all over themselves to shake hands with the loudest voice in the room – in this case, opponents of cash bail, who think with their hearts and not their heads – instead of taking a step back and asking:

• What can we do better?
• Do we need to re-invent the wheel (cash bail), or make it more puncture resistant?
• Is technology and math (RIAs) the only answer?

These are tough questions to ponder, but what we’ve seen so far are off-the-cuff answers pushed into law. And they have more holes than Swiss cheese.

Here’s the rub. Cash bail opponents insist the practice discriminates against people who can’t afford bail in the first place, but jurisdictions across the country use an alternative instead – high-tech ankle bracelets – called Electronic Monitoring (EM), which most defendants don’t realize comes at a price.

The whole system is in desperate need of dispassionate review, but we’re not convinced common sense will be the ultimate victor.