PROFILES, PROBABILITIES, AND STEREOTYPES

Vol. 16 No. 10 (October, 2006) pp.811-816

PROFILES, PROBABILITIES, AND STEREOTYPES, by Frederick Schauer. Cambridge: The Belknap Press of Harvard University Press, 2006. 359pp. Cloth (2003). $29.95 / £19.95 / €27.70. ISBN: 0-674-01186-4. Paper (2006). $18.95/£12.95/€17.50. ISBN: 0-674-02118-5.

Reviewed by Sawyer Sylvester, Department of Sociology, Bates College. Email: ssylvest [at] bates.edu.

Frederick Schauer begins his book by acknowledging that it is based on an unfashionable idea: namely, that justice may be done more often by legal generalization than by an attempt to tailor outcomes to particular cases. So popular is the assumption that we must take into account the special nature of individuals or situations that we often refer pejoratively to its opposite as “stereotyping” or “profiling.” However, Schauer points out that the practice of “thinking like an actuary” or acting on the basis of generalizations is far more common in life and in the law than its opposite. Indeed, a sociologist might make the even stronger argument that behavior on the basis of constructed types is part of the human condition.

Schauer also acknowledges that a preference for making decisions based on generalizations does involve some caveats. First, generalizations, especially those used for policy decisions, must have some adequate empirical basis. Second, even statistically sound generalizations ought not to be acted on if that action would be deemed morally wrong. This, in fact, is a major theme of Schauer’s book. He uses two examples to illustrate the difference between cases wherein it might be wrong and where it might not be: eastern European Jewish women who are denied health insurance because of a generally greater propensity to certain forms of cancer, and English soccer fans who as a class have been denied admission to Continental games because of a greater history of hooliganism. There may be other distinguishing features, but Schauer chooses to distinguish the two cases morally on the basis of the far more serious consequences and cost to individual exceptions in the former than in the latter. However, he states that exclusion in neither case can be deemed morally wrong based on the mere use of generalizations or even generalizations that might be empirically invalid.

But if generalization is not only pervasive but permitted even in the law, is the law still a unique case of social policy because of the consequences for non-conforming cases? The law’s generalizations, like all others, inevitably produce mistakes. But must the law, because of the potentially greater harm of those mistakes, take special pains to provide a system for correction in individual cases? Schauer suggests that many would think it must philosophically, and has institutionally. For example, he believes that Aristotle’s concept of aequitas requires such a corrective, assuming that the law achieves justice only when it is prepared to rectify its errors. He also sees such a principle institutionalized in the mediating authority of the Roman [*812] Praetor and in the rise of the courts of Chancery in England (although, in the latter case, the system may have been as much inspired by the Crown’s desire to hold on to some of its prerogative authority in the face of the growing power of the law courts as it was any benevolent concern for the law’s harsh bite).

The mistakes to which some believe the law is given often arise because laws tend to be over-inclusive, and perhaps some system of equitable adjustment for the individual hard case is needed to provide a remedy. Yet, there are other ways to look at the problem. Schauer refers to the case of the pit bull. Apparently there are data from which one may infer that pit bulls, as a breed, are likely to be more vicious than other breeds of dogs. On this basis, some jurisdictions have simply made it illegal to own a pit bull. However, pit bull owners say that such laws are overly inclusive and that there may be many pit bulls that are not vicious. The solution, they say, would be to prohibit ownership of only those dogs that have injured people – an example of the venerable legal maxim that every dog gets one bite. Should society not wish to bear the cost of multiples of the “one bite,” all pit bulls, or all dogs for that matter, could be tested for vicious traits, and only those testing positive would be prohibited. In the second and third cases, the legislation is tailored more narrowly, but, Schauer points out that there is no essential difference among these policies in terms of generalization. They all involve the use of some class characteristic to proxy for predicted behavior, and all are capable of making mistakes.

Schauer claims that there is also an evidentiary side to the generality/particularity debate, in the distinction between “direct” evidence and evidence from which conclusions can be drawn only by inference – more frequently, these days – statistical inference. Schauer suggests that judges often see the law as part of the “all or nothing world” we conventionally inhabit (guilty or not guilty, for plaintiff or defendant, admissible or inadmissible, contract or no contract), even though the decision-making process in law is often more a matter of inference than commonly realized. In part, this is because of the courts’ distrust of “naked statistical evidence” and a preference for such direct evidence as eyewitness testimony – even when the latter is shown to involve its own process of inference (often faulty) from visual cues (often misapprehended). Courts will likely prefer the expert willing to say, “I have compared the two fingerprints, and I am convinced there is a match,” to one who is only able to say, “I have compared the two fingerprints and studies have shown that in cases where there are this many points of similarity, there is an eighty-seven per cent chance of a match.” The dark note is, as Schauer points out, that even though decision-making based on individualization is more intuitively appealing because decision-making by generalization is thought to have the greater capacity for error, more actual errors have been shown to occur by the former than the latter. [*813]

Schauer uses two more extended examples of his initial observation that decision-making by rule or generalization is not in itself morally questionable, but only questionable on other grounds. Take the case of airline pilots forced to retire by the Age-Sixty Rule, and women being denied admission to a military school on the basis of gender. In the first case, the Age-Sixty Rule forced pilots to retire because it was believed that age was valid proxy for a decline in some of the skills necessary for safe flying. Although the relationship between age and skills was not spurious, it was claimed by the pilots to be arbitrary by being overly inclusive. One solution raised was to test all pilots at a certain age for the necessary skills. But this only shifts the problem of arbitrariness from testing to selection for testing. Actually, using age as a proxy would always be arbitrary. But we are quite willing to use it in other areas such as voting age, driving age, drinking age, and the like. If not spurious and not arbitrary, then age may simply be a category, the use of which is inherently unjust. Schauer thinks not, and again, perhaps because of a cost/benefit ratio (the cost of a few undeserved retirements each year against the cost of undiscovered age-related deficiencies in the cockpit) or, possibly, because its use just does not shock the conscience in the way other forms of discrimination do.

Contrast this with the opinion by Justice Ginsberg in UNITED STATES v. VIRGINIA (1996). This case involved an effort by Virginia Military Institute to exclude women from its program based on the claim that women lacked, among other things, the necessary “adversarial” traits to succeed in the program at VMI and that gender was a valid proxy for that deficit. The Court found that VMI had engaged in unlawful discrimination, and Schauer suggests that Justice Ginsberg reasoned as follows: that the relevant variable – here, “adversarial” traits – might well have been a product of discrimination itself and came about through acculturation rather than being genetically determined. In addition, Schauer sees in the Court’s opinion reluctance to accept a rule which is overly inclusive in this case where there was no similar reluctance with pilots, and a willingness to mandate that any woman wishing to apply to VMI should be allowed to demonstrate individually that she had the traits necessary for the program. He suggests that this difference in approach by the Court was occasioned by the fact that the use of gender as a proxy has historically been “routinely exaggerated and routinely overused.” It has thus been seen as a more suspect category. Once again, it is not the use of generalizations per se that is the issue, but the uses to which they are put.

Perhaps the best known use of generalization in law enforcement, and one of the most controversial, is profiling. Psychological profiling is used in the investigation of serial murders and rapes. Geographical profiling is used in connection with a routine activities approach to other predatory crimes. And for even a longer time, profiling of a sort has been the basis of parole and probation prediction schemes. But controversy seems to have centered on the sort of profiling that picks out subjects to be stopped and questioned by [*814] law enforcement officers. Schauer discusses the two examples: drug courier profiles and profiling used to justify traffic stops.

A drug courier profile was involved in UNITED STATES v. SOKOLOW (1989). Andrew Sokolow was stopped and questioned at the airport because his appearance and demeanor fit the multi-faceted written profile in use by the DEA. A majority of the US Supreme Court upheld his subsequent conviction, concluding that fitting the profile gave rise to “reasonable suspicion” necessary for a stop-and-frisk. Justice Marshall, however, dissented on the basis of a preference for reasonable suspicion based on the experience of individual agents rather than the collective experience reflected in profiles. Schauer again points out the fallacy in concluding that these two types of profiling differ in that one is based on generalization and the other on particularity. In fact, they both generalize, but one does so on the basis of multiple items, each having some proven contribution to the entire profile, as against generalization based solely on accumulated individual experience. Not only does the former process generally prove to be more accurate, but the latter easily leads to abuse in the hands of those whose experience includes the exaggerated use of suspect categories in profiling, such as race or ethnicity. Schauer’s example of the latter is the scandal in the New Jersey State Police where African-American drivers were stopped disproportionately based on a profile which seemed to credit race as a proxy for the likelihood that there was contraband in the driver’s vehicle. The profile, so called, was probably no more than the idiosyncratic and unconstrained judgments of the officers involved.

Schauer asks, what if race or ethnicity, however suspect, were merely one of a number of factors, each of which had empirically verifiable predictive value? He suggests that, in such a case, we should make the same cost/benefit analysis as before. How much predictive accuracy will be lost by excluding race or ethnicity because of the special burden placed on persons in those categories? And, if there is a loss, is there some way to make up for it even if there are extra burdens created but more generally shared? Race, for example, might be removed from an airport profile, but all passengers would then be given greater scrutiny, even at the cost of more time waiting in line for everyone.

Generalization, as a whole, in human behavior accomplishes the necessary task of simplifying a complex world of perceptions by forming classes of things and allowing one to behave toward others based on class characteristics with a fair approximation of correctness toward the individual. In social behavior, we respond to people, not with a full appreciation of their individuality but in accordance with some stereotypic image we have created of them. There is a basic and necessary economy in all this. And it is, in part, the same economy that drives the law to make demands and create remedies that fit broad classes of persons and behaviors rather than to tailor individual remedies for each person or situation. [*815]

But Schauer asserts that the law does something more by the use of generalization, that it can exemplify certain values – equality, for example. Treating like cases alike is seen by many as an element of justice. But Schauer notes that treating unlike cases alike is probably closer to core values of the law. When the Declaration of Independence states, “We hold these truths to be self-evident, that all men are created equal,” the statement is descriptively invalid but proscriptively necessary to the implementation of individual rights. It is a rule which states that certain distinctions among human beings, however relevant, will be treated as irrelevant.

Schauer offers another, and more controversial, example of cases in which the law treats presumptively relevant conditions as legally irrelevant. This would be mandatory sentencing schemes established by legislatures to act as a brake on judicial discretion. The most famous of these schemes (or infamous, depending on your point of view) is the Federal Sentencing Guidelines established by the Sentencing Reform Act of 1984. These guidelines severely restrict the power of Federal judges to construct sentences individualized for persons or circumstances. Some differences may be taken into account, but only within the constraints and categories of a sentencing grid.

Many judges have complained about the Guidelines, not only because they correctly see a major portion of their authority as judges being hijacked, but also out of a recognition – probably gained from experience – that there are some cases where justice, fairness, mercy, and simple good sense demand an outcome not available under the Guidelines.

What, then, can be said in support of such legislation in the face of the hard cases it produces?

Schauer answers that, much like the proven superiority of actuarial prediction over clinical prediction, sentencing by the Guidelines may, in the long run, simply make fewer mistakes than a less-than-perfect judiciary. And he reminds us that judicial discretion has itself been known to produce results which are unjust, unfair, merciless, and simply nonsensical. Perhaps “Equal Justice” needs to be “Under Law.” Schauer also points out that uniformity by itself my be intrinsically valuable in a legal system and may also serve as a reminder that “at a moment of enormous significance – determining how much of the state’s power will be brought to bear against those who transgress its laws – we are all in this together” (p.261).

PROFILES, PROBABILITIES, AND STEREOTYPES is an excellent book. Although it seems fashionable these days to mine an otherwise creditable work for something with which to find fault, I can find nothing. The writing throughout is clear and felicitous, the logic tight, and the historical illustrations apt. It is a book which could easily contribute to a course in legal philosophy or sociology of law, or be a focal point around which one might arrange other readings to build a seminar on the subject of the book alone. It is highly recommended. [*816]

CASE REFERENCES:

U. S. v. VIRGINIA, 518 U. S. 515 (1996).

U. S. v. SOKOLOW, 490 U. S. 1 (1989).

*************************************************