Is Law Just Social Science in Robes?

It seems that Pennsylvania is about to adopt a new (well, not entirely new) approach to criminal sentencing: Those convicted of crimes will receive longer or shorter sentences depending on whether they are deemed likely to commit crimes in the future. The relative likelihood will be determined by the social science measure of risk assessment, where researchers assemble huge data banks made up of innumerable facts drawn from the life records of former prisoners; the data is then searched for patterns of behavior and characteristics (including sex, age, neighborhood residence, family history, prior criminal behavior, social attitudes, level of education, etc.) that seem to be correlated in a predictive way with future law breaking.

From these patterns and characteristics one can derive a checklist of factors, and if a convicted defendant displays a sufficient number of those factors, he will receive a longer sentence. The idea (I guess) is to keep potential recidivists in jail as long as possible; since they'll probably be back in a little while for a repeat offense, let's be efficient and keep them in. (For a fuller account see Barry-Jester, Casselman and Goldstein at FiveThirtyEight.com, "Should Prison Sentences Be Based On Crimes That Haven't Been Committed Yet?")

There have been some predictable objections to this plan: It will skew negatively for residents of urban areas who are likely to be minorities; and it can involve a subjectivity of judgment that might be masked by banks of statistics. But the real objection, I would say, is that penalizing a defendant for a crime he or she has not yet committed is to introduce into the law a set of concerns that belongs elsewhere. It is perfectly appropriate for social scientists to investigate the relationship between any number of variables and criminal activity; the results of their investigation might then be used to argue for changes in the delivery of K-12 education or for an increase in the funds allotted to inner cities. That's the general program of social science: Here is a problem caused by social factors; let's see what we can do to ameliorate those factors.

But the law is not future oriented in that way; its concern is not with what will or should happen, but with what has already happened, and its focus is narrowly on the fixing of responsibility and the assigning of blame for an act that is proscribed by statute. That narrowness, as every viewer of Law and Order knows, extends to a willed ignorance of a defendant's past deeds. Prior bad acts cannot be introduced into evidence under a rationale articulated in rule 404 of the Federal Rules of Evidence: "Evidence of a person's character or a character trait is not admissible to prove that on a particular occasion the person acted in accordance with the character trait." The prosecution must prove that a defendant did it; an argument that he's the kind of person likely to have done it is not to the legal point, although it might have a place in the discipline of psychology. An argument that he is the kind of person likely to do it in the future is even more tenuously related to the issues of guilt and punishment, and should not be a part of the process. Allowing risk assessment statistics to determine the length of sentences is a category mistake.

It is analogous in its inappositeness to another category mistake that is now regularly made when "victim impact statements" are heard in the sentencing phase. Ohio's victim impact statute reads in part, "Before imposing sentence upon... a defendant... for the commission of a crime,... the court shall permit the victim of the crime... to make a statement." The justifications for victim impact statements, law professor Paul Cassell explains, include (1) a therapeutic benefit to the victim who gets to be heard and to express his or her pain; (2) an educational benefit to the defendant who is confronted with the consequences of his act; and (3) an information benefit to the court that has the task of assessing the seriousness of the crime. (1) and (2) are obvious goods -- it is good for victims to heal and it is good for perpetrators to understand the suffering they have caused. But these are psychological and sociological goods rather than legal ones and, as such, they are goods judges and courts are ill equipped to engineer; it is not the business they are in. As for (3), information is what trials are designed to elicit -- if they are properly conducted there is no need for the crowd-pleasing drama of victim impact statements.

Risk assessment sentencing and victim impact statements are the residue of Legal Realism, an early twentieth century movement that derided so-called "legal reasoning" because it was entirely internal and filled with abstract terms that spoke only to each other, never making contact with the real world where real men and women live real lives. Legal Realists proposed to jettison the empty and self-referencing vocabulary of traditional legal discourse and replace it with observations of what people -- including judges -- actually do.

In a legendary essay in the Columbia Law Review, Felix Cohen urged the abandonment of the "ghost-world of supernatural legal entities" and the putting in its place of an understanding of "legal concepts as patterns of judicial behavior which affects human lives." If a term like due process "cannot pay up in the currency of [social] fact upon demand, it is to be declared bankrupt" and is no longer to be used. Every part of the judicial process must be presented and described in "non-legal" terms. Obviously, if this program is prosecuted to its fullest, there will be no distinctively legal vocabulary and, consequently, there will be no distinctive thing called law. Instead law will be the name -- a place marker rather than an essence -- of one more location where the truths of psychology and sociology work themselves out.

So the question (an old one) is, Do we regard law and legal reasoning as essentially no different from the reasoning we deploy in ordinary commercial and domestic transactions, or do we regard law as autonomous, that is, distinguished by concepts, rules and procedures that belong peculiarly to it? If it is the first, risk assessment sentencing and victim impact statements will seem perfectly appropriate because predictive formulae and calculations of effect are familiar ingredients of our everyday practices. But if it is the second -- if there is actually something called law that has its own integrity -- risk assessment sentencing and victim impact statements will seem like strangers from another land who have wandered into a territory they don't belong in.