“Hey Siri, Should this Defendant be Awarded Bail?”: Artificial Intelligence in Modern Maryland Bail Considerations

March 13, 2019

3–5 minutes

As electronics continues to flourish with only the passage of time, we can expect that areas of our lives will begin to be unrecognizable when compared to the past. From computers the size of classrooms, to having a computer, phone, camera, media player, and entertainment system in your pocket, several areas of emerging technology and innovation are intended to make our lives easier. But what areas of our lives do we want to keep a “human touch”? The care of children? The medical field and bed-side manners? The government? One unexpected, and potentially troubling, area where artificial technology is seeing implementation is the criminal justice system.

In line with a national trend, numerous counties in Maryland seem to be embracing artificial intelligence, or “AI”, in to bail determinations for individuals arrested and charged with a crime.[1] Currently, a total of ten counties in Maryland have begun to calculate what have been called “risk scores” for defendants in criminal cases, which is intended to indicate the likelihood of reoffence or failure to appear for trial if the defendant were released on bail.[2] The factors that are included in the scoring of a defendant are their criminal history, their general demographic data, like gender and age, and information about the crime they are being charged with where bail is at issue.[3]When analyzing this data, certain indicators have a “point value” that increase or decrease the score for a defendant.[4] For example, serious crimes like rape and murder have a value of 9 points, and female gender has a value of -1 point.[5]

Once these scores are calculated, they are compared to categories that have been pre-determined based on the score: a score of less than 6 are recommended for release without supervision, a score of 6 to 12 is suggested for pretrial release with supervision, and scores of 13 or more are recommended to stay in jail pending their trial.[6] This recommendation is then presented to the judge, along with all other relevant issues, for the judge’s consideration in bail determinations.[7] It is important to note that the AI’s report and recommendation is not the deciding factor and the judge still has discretion to rule for or against awarding bail.[8]

Unfortunately, there is an understandable fear that racial bias will creep in to the algorithm used by the AI, much like it has in other areas of criminal justice.[9] Although race is not a factor that the AI considers, it does consider areas that are plagued with racial disparity to this day, like homeownership, employment, and criminal history.[10] As such, there is worry that individuals of a minority race who are disproportionately impacted by racial bias in other areas of their life will suffer for those biases in the criminal justice system for bail determination. Even more troubling, the only way to check that the tool is not impacted by any racial bias would require extensive testing, which would cost the state money, and enough of a pool of the population to check for significant differences in outcome based on race.[11]

On its face, the implementation of technology to give a completely objective determination based on real statistics is not a bad idea. In effect, it would be the removal of any racial bias or prejudice, whether known or latent, that a judge may have in their bail determinations. But the criminal justice system is never that easy. So long as there are areas of everyday life where racial disparity thrives without being checked, there will always be the possibility that those bias will be picked up, unintentionally, by the AI. This is because the AI is not able to distinguish between legitimate and tainted data, and therefore may negatively impact defendants in bail determinations.

Harry Snoots is a third-year law student at the University of Baltimore. He serves as a Staff Editor with University of Baltimore Law Forum, Treasurer with the University of Baltimore Chapter of the American Constitution Society, and is a Student Attorney with the University of Baltimore Bronfein Family Law Clinic. Before coming to law school, he attended Salisbury University and graduated in 2015 with a Bachelor’s Degree in Psychology and Philosophy. During summer sessions, he works with Mid-Shore Pro Bono, a regional legal non-profit, as an intern and volunteer in Easton, MD. He is currently employed as a law clerk for FreeState Justice, Inc., a legal non-profit that seeks to serve the LGBTQ+ community in their unique legal issues. His interests include family law and emerging topics of law for families and parents, including surrogacy, IVF, and parental rights for LGBTQ+ families and the associated rights.

Untitled