California's governor signs a bill that replaces cash bail with an algorithmic system to calculate risk of flight or reoffending from October 2019
Dave Gershgorn / Quartz : Tweets: @davegershgorn and @mediajustice Tweets: Dave Gershgorn / @davegershgorn : California moved to replace cash bail with algorithmic risk assessment last week (following New Jersey a year ago), and there's debate about whether the state is swapping one biased system for another. Read my story here: http://qz.com/1375820 Center for / @mediajustice : Replacing cash bail with risk assessment tools is NOT bail reform! Why? “Because machine learning systems used to calculate these riskscores have been shown to hold severe racial biases, scoring people of color more likely to commit future crimes.” http://qz.com/...
Context & Ripple Effects
California is following the path New Jersey already walked: swap cash bail for a pretrial risk-assessment algorithm, effective October 2019. The state's move lands amid an established debate — courts have long leaned on proprietary third-party recidivism scores in sentencing and bail decisions, and civil-rights groups like the Center for Media Justice argue that machine-learned risk scores carry documented racial bias, so abolition of cash bail by algorithm may just re-encode the disparity.
First-order effects
- From October 2019, pretrial release in California hinges on an algorithmic flight/reoffending score rather than ability to pay, immediately shifting power from bail bondsmen to whichever vendor supplies the state's risk model.
- Defendants flagged high-risk by the tool face detention without any cash alternative, making the algorithm's accuracy and bias the direct determinant of who waits in jail.
Second-order effects
- New Jersey's experience under its own PSA-based system becomes the live test case: jail populations fell there, but racial disparities persisted — evidence both sides will cite as California implements its version.
- Vendors of criminal-justice risk models gain their largest market yet, intensifying scrutiny of opaque proprietary scoring at exactly the moment federal reform like the First Step Act also leans on algorithms.
Third-order effects
- If the pattern holds, US pretrial justice consolidates around algorithmic gatekeeping regardless of reform politics — cash bail falls, but the discretion migrates to unaccountable models, unless transparency mandates or judicial override practices catch up.
- The debate pushes toward structural questions about due process when life-altering decisions come from third-party software, foreshadowing the governance frameworks now being built around predictive systems in policing, sentencing, and probation.
The trend: Criminal justice reform in the US is increasingly implemented through predictive algorithms rather than abolished discretion, trading explicit cash-bail inequity for harder-to-audit statistical inequity.