Pretrial Risk Assessment: Computers vs. Man, Fantasy vs. Reality

As technology has advanced over the years, many debates and questions have been asked around the topic of machine versus man.  These questions include things like can computers do a task better than a man? Can computers make thoughtful decisions that involve experience and emotion?  Can man be as effective and as accurate as a computer without the ability to process and analyze millions of data points?  While no one has been able to completely answer these questions, and trust me when I say that I do not plan on answering them in this blog post, I will, however, give my thoughts on the current man vs. machine debate that is happening in the criminal justice system.

A few weeks back I watched a “Ted Talk” video about risk assessment.  In this video, the speaker talked about risk assessments and their likeness to an approach used by the professional baseball team the Oakland Athletics called “moneyball.”  The moneyball philosophy (just in case you never saw the Hollywood movie made on the topic) is based on a deep analysis of data and statistics.  The theory is that you can use this data to tell you which players you should draft and/or acquire based on their statistical performance in certain situations.  For example, should you pay millions of dollars for a big name superstar that hits 50 home runs a year or do you spend a few hundred thousand dollars on a lesser known player that consistently hits for average with runners in scoring position.  This analytical approach to baseball has been very beneficial to the Oakland A’s organization allowing them to compete with teams that have much higher salaries and cost structures.

The question that I would like to raise is does this type of “moneyball” statistical analysis or “risk assessment” as it is referred to in the criminal justice system work like it does in baseball?  Is it a fair analysis to equate the two together and through the transitive property assume that both are effective?  I would argue no, it is not the same thing.

The concept of pretrial release is not something that can be distilled down to common statistics across a broad range of different environments.  Every defendant is different; every state is different; every county is different; every criminal act is different; and every reason a person commits a crime in the first place is different.  Criminal Justice is not that simple.  Additionally, the people who make the ultimate decisions in the criminal justice system are judges.  And judges don’t just become judges; they earn a judgeship through experience and practice.  Sitting in front of thousands of unique defendants is more valuable to understanding how criminals think and act than just looking at questionnaires and statistics.

The same goes for the bail bond industry.  Bail bond agents deal with thousands upon thousands of defendants.   These are defendants that they understand and are familiar with.  Why?  Because they have grown up and live in the communities that they serve.  They know the people; they know the families; they know the employers and the businesses.  They have a world of local knowledge and insight that no computer program or public sector pretrial risk assessment can quantify and make sense of.  Call it gut or call it experience, but the bail bond industry brings an intangible component to their own decision making process that is not merely about risk assessment but rather about something more important “risk management.”  Ultimately when a bail agent decides to financially secure a defendant’s release, they are focused on managing the risk of the defendant and ensuring they show up for court as opposed to just assessing the risk and trying to play a guessing game of whether that person will show up or commit another crime.

Based on the uniqueness of the criminal justice from locality to locality and the reality of the pretrial world, how can, as the speaker in the video explains, a computer program that spits out a 7 question risk assessment be enough information to accurately claim that a defendant is dangerous to the community and should not be let out of jail pending his trial.  In my opinion it can’t for the very reasons I have mentioned above.  Computers are not people.  They are tools that people use to make a complex task easier.  In the world of baseball that might work.  In the world of the criminal justice system and pretrial release, where public safety is at stake, I think that these computer modeled risk assessments are a shiny new toy for the public sector pretrial release community to rally around.  Why, because they have nothing else.

If statistics and data are so important, than why doesn't the public sector pretrial community just look at the data that is already in existence.  For example, there is plenty of data on pretrial release successes and failures that have been published by the Bureau of Justice Statistics.  The problem is that data doesn’t support pretrial release through a publicly funded pretrial release agency, but rather shows that release through a financially secured commercial bail bond is the most effective way to get defendants back to court. And when that happens, we all know that the public is protected, victims get a chance at justice and the system has accountability.

Summing things up, should the criminal justice system rely on a moneyball approach to pretrial release?  I guess the answer to that question is that the jury is still out, but as far as I know, the Oakland A’s still haven’t won a championship and public sector pretrial release still hasn't outperformed private sector financially secure release…so you decide. I look forward to your comments.