THE END OF BAIL – ARE RISK-ASSESSMENT TOOLS BIASED?
With the coming end to money bail, California courts will be mandated to rely on risk assessment conducted by Pretrial Assessment Services (PAS). Risk assessment, as mandated by the new law, will categorize those arrested for a crime as high, medium, or low risk in terms of the likelihood that the alleged offender will return to appear in court and the risk the alleged offender poses to the public. Each alleged offender will receive a “risk score” and depending upon that score, the defendant may be released on his or her own recognizance or may be held in jail. Those charged with a misdemeanor (with certain exceptions) will not be subject to the assessment and will be released on their own recognizance.
But this blog post is not about the new law, but about “risk assessment.” How will PAS make the risk assessment? The new law requires PAS to rely on a “validated risk assessment tool” approved by the court from a list of such tools that are maintained by the Judicial Council. Orange County, as will each county in California, chose what tool to use from those approved by the Council. The tools must be scientifically validated for their accuracy and reliability in assessing the alleged offender’s risk. These tools are algorithm-basedand calculate risk based on the alleged offender’s criminal history and other personal factors plus general criminal justice data. The data input, depending on the tool, may include more than 100 factors, which are weighted according mathematical formulas to assess an alleged offender’s risk. This use of artificial intelligence to determine who gets out of jail and who doesn’t is not without controversy and opposition. Indeed, over 100 civil rights-related organizations opposepretrial risk assessment, including the ACLU and NAACP.
Detractors claim that the algorithm-based assessment is biased. A major criticism is that non-white individuals are more often targeted by the police than white individuals and therefore have higher arrest rates even though they may not commit any more crimes than the population at large. In other words, the risk assessment inputs are inherently biased against non-whites due to racially disparate policing practices.
One major studyprovided support for this concern. Taking a sample of 7,000 arrestees subjected to a risk-assessment algorithm program in Florida found that black offenders were incorrectlydeemed medium to high risk compared to white offenders and conversely, white offenders were incorrectly deemed low-risk more often than their black counterparts. However, it should be noted that this study seems to be the only study thus far documenting this bias.
There is also concern regarding the transparency of the data input. The new law requires that any tool used by PAS provide a detailed accounting of the data inputs and outcome measures. The law also requires that the risk assessment tools be open to public scrutiny including information regarding line items, scoring, weighting as well as validation studies.
Others note that the data fed into the algorithms are too narrow and fail to take all factors into account that the human judgment of the court judge would consider. Some believe this new system will result in more people being held in jail pending plea or trial, while others are sure it will result in more people being released pre-trial.
When the pre-trial release or detention law goes into effect on October 1, 2019, courts will no longer set bail. This new law will affect every person arrested in California after October 1st. The courts will still have some discretion regarding continued detention or release but will no longer be able to order a bail release. In those cases where PAS has labeled the defendant medium to high risk, the hiring of defense counsel who can argue for release will of immediate consideration.
Orange County criminal defense attorney William Weinberg has over 25 years’ experience defending the accused. He is available for a free consultation. You may contact himat his Irvine office by calling 949-474-8008 or by email at email@example.com.