Bar News - September 16, 2015
Opinion: Applying the Automation Paradox to Criminal Justice
By: Tony Sculimbrene
There are two companies that make the vast majority of large commercial airplanes Ė Boeing and Airbus. These two companies could not have more diametrically opposed views of automation in the cockpit. For Airbus, computers do the majority of the flying, other than takeoffs and landings. Its sophisticated fly-by-wire system does almost everything on its own. Pilots are mostly monitoring instrument panels. Boeing also has sophisticated flight automation, but its approach includes a more active role for the pilot. He or she is the hub, the decision-maker for all of the machines that aid in flying Ė this is called piloted-centered automation. Both make incredibly safe planes. Accidents are literally one in a million.
But something has happened as science and computers take over. In a recent Airbus crash, the automated systems failed, and the system monitoring the automated systems also failed. The result was a crash with all lives lost because the pilots, so removed from the act of flying, couldnít save the plane in time. This is the automation paradox. The more automated a plane is, the less the pilots do. The less that the pilots do, the more likely they are to not be able to save the plane when the automated systems fail.
In many ways, this automation paradox parallels what is happening in criminal cases with the use of science. The advent of DNA evidence, the importation of the scientific method into the forensic lab, and the increasing reliance on scientific experts heightens the expectation that criminal convictions should be more reliably correct. However, as in flying, something similar to the automation paradox is at work. Scientific evidence presents two different problems, assuming of course, it is legitimate science (which is not always whatís presented in court).
First, sometimes science seems to say one thing when actually it says something else. For example, offenders in sexually violent predator (SVP) cases are evaluated using questionnaires, which are scored using massive databases. These scores generate a profile of risk, placing the potential SVP in a group. The groupís risk profile is well-known, having been tracked for decades. The size of the samples used and the longitudinal tracking of these groups imbues these scores with an appearance of incontestable certainty. The group a person is placed in might have a 75 percent re-offense rate. Case closed; he is too risky.
Not so fast.
If you have an understanding of probability and statistics, you can see the problem very quickly. While the group has a 75 percent re-offense rate, the rate for any individual in that group is completely unknown. Compare it to team batting average in baseball. We know that as a team, the players bat .280, but we do not know any individual playerís batting average.
The problem with scoring in the SVP context is this very issue Ė the evaluators offer certainty in their methods, but their conclusions arenít exactly what the court wants. Courts want individual rates of re-offense, but we donít and probably never will have the ability to predict that. Confusion about conclusions is one way science can lead to bad results.
Another problem with over-reliance on science is more straightforward. In many cases, the scientific evidence is unassailable. DNA found at the scene confirms that person X stabbed person Y. That sort of conclusion gives us a tremendous amount of confidence in what happened and seems like it could close the book on a case. Itís so powerful a conclusion that it might even blind prosecutors, defense lawyers, judges, and juries to the truth.
What if this was a case of self-defense? The law is not just about what happened, but why it happened, and that is something science cannot perfectly explain. Motive is not a scientific concept; it is, at its root, a moral one, outside the realm of science. But when we have certainty regarding actus reus, we often assume it is certainty with mens rea as well.
Science is an indispensable tool in the courtroom, but it is not the final arbiter of guilt or innocence. The entire justice system needs to guard against the legal equivalent of the automation paradox. In the end, the law is about judgment, and judgment is something about which science remains silent.
Tony Sculimbrene is a public defender at the NH Public Defender office in Nashua. His views do not represent those of his employer.