There's No App for That
I received a proposal to help in the creation of an algorithm to assist inexperienced public defenders assess the worth of plea bargain offers. The creators want to take offense conduct, criminal history and other relevant factors, collect them over a period of time, and weigh their relative worth. They would then present this in the form of a software program that can opine whether an offer made to a particular client should be considered or rejected.I received a proposal to help in the creation of an algorithm to assist inexperienced public defenders assess the worth of plea bargain offers. The creators want to take offense conduct, criminal history and other relevant factors, collect them over a period of time, and weigh their relative worth. They would then present this in the form of a software program that can opine whether an offer made to a particular client should be considered or rejected.
In defense of the proposal, I do not believe the creators think this will replace the independent judgment of trained and experienced lawyers. They offer it merely as a tool. Additionally, the creators are qualified social scientists whose previous studies of criminal justice are recognized and admired.
I declined to participate. It is one thing to examine trends and come to conclusions on a broad scale, and quite another to then apply those conclusions to individual complex cases. The problem is that the proposed software reduces complicated decisions to a simple equation. The federal sentencing guidelines were also designed to reduce disparity, but they merely institutionalized it.
I have been a criminal defense lawyer for 30 years. I have established and managed federal and state public defender offices for the last 20 years. Every attempt I have seen to make outcomes fairer, by trying to make them equivalent, has failed. There are simply too many variables and they change in each jurisdiction, each court and each case. The relevant criteria to examine a plea offer are a long list of subtle details. Even if one could somehow create the perfect algorithm for one jurisdiction, it would not be applicable in thousands of different jurisdictions across the United States.
However, leaving aside the science, there is a more basic reason why I could not participate. The failure of public defense in the United States generally arises from the acceptance that criminal defendants may be treated as less than human. See Amy Bach, Ordinary Injustice: How America Holds Court (Metropolitan Books 2009). When brand new public defenders are taught to move criminal defendants through the system as efficiently as possible, innocent persons are convicted and guilty persons are punished more severely. Any public defender office that does not have the time to have its lawyers fully investigate its cases as to proof of guilt and punishment simply has too many cases. Software for plea bargaining sends the wrong message.
I should also mention that I support the utilization of the Laura and John Arnold Foundation's pretrial risk assessment here in Harris County. That is because it attempts to improve release decisions by expanding the criteria beyond merely the offense charged. It does not try to take a much more complex and permanent decision and reduce it to uniform criteria. In other words, bail decisions can be changed, convictions and sentences cannot. Bail decisions are relatively simple calculations, convictions and sentences are not.
Lawyers need to learn more about science. That is true for forensic science, where new protocols are introduced and old methods are debunked. It is also true for social science, in which statistics can both support and undermine criminal justice policies. Lawyers need to know how far science can go. Here, it attempts to reach too far.