Criminal Careers and "Career Criminals,": Volume II (1986)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

8 Purblind Justice Normative Issues in the Use of Prediction in the Criminal Justice System Mark H. Moore In the workaday world of criminal jus consider the risk that a defendant will flee (or commit additional crimes) in setting lice, predictions are commonplace (Der- we ~ , showitz, 1974:1~0, 781~46; Wilson, bait (Dawson, 1969:80; N. Morris, 1974: 1983a: 157, 1983b:2791. Moreover, they are consequential for clefendants: they affect the magnitude of the criminal lia bilities that defenclants confront. Judges Mark Moore, Guggenheim Professor of Criminal Justice Policy and Management, Harvard Univer sity, notes: "In producing this paper, I am princi pally indebted to Susan Estrich, Daniel McGillis, and William Spelman, my collaborators on the Harvard Project on Dangerous Offenders and coau thors win me of Dangerous Offenders: The Elusive Target ofJustice (Cambridge, Mass.: Harvard Ur~i versity Press, 1984). Indeed, some material on the use of predictions in sentencing and bail is repro duced here exactly as it appears in Dangerous Of fenders, and a great deal else is borrowed less directly from that analysis. I am also indebted to those on the Panel on Research on Criminal Careers who read and commented on this work, specifically, John Kaplan, Norval Morris, and James Q. Wilson. I am also greatly indebted to Andrew von Hirsch and Michael Tonry, whose sharp disagreements may have improved the quality of my arguments and whose patience in instructing me has been ex~aor dinary. Win such great assistance, it is hard to believe errors could be made. But no doubt there are many, and they are mine alone." 374 28-57; Roth and Wice, 1980; Gaynes, 1982; Blumstein et al., 1983a,b; N. Morris en cl Miller, 1985: 12) and the prospects for rehabilitation in imposing sentences. Prosecutors weigh the gravity of the threat posed by accused offenders in de- cicling how much effort to put into prepar- ing their cases and in setting the mini- mum acceptable plea bargains (Kaplan, 1965: 174; Forst and Brosi, 1977: 177- 191~. Police stucly the moclus operancli of offenders to thwart future crimes and to help them identify likely suspects in cur- rent cases (Moore et al., 1983a,b). The widespread, consequential use of predictions in the criminal justice system prompts normative questions. If the jus- tice of the system rests on the notion that  punishment should be for past acts, not guesses about future behavior, it is wrong to impose criminal liabilities on the basis of predictions. It would be wrong even if the predictions were perfectly accurate. If they are inaccurate, however (as they in- evitably will be), additional objections

PURBLIND JUSTICE come to the fore. Offenders incorrectly predicted] to commit crimes in the future wouIc] be exposed to criminal liabilities that are doubly undeservecI: once be- cause they were based on predictions rather than past creeds, and twice because the predictions were inaccurate. And, to the extent that the predictions were based on characteristics of offenders that lie out- side the ordinary purview of the criminal justice system or are imperfectly mea- sured, defendants wouIct be exposed to more intrusive investigations and greater risk of errors than would be the case if the focus of the system remainec! on past crimes. In short, predictions undermine the rigorous discipline essential to crimi- nal justice in a free society. On the other hancI, criminal justice of- ficials now rely on predictions because they seem to add to the overall justice and performance of the criminal justice sys- tem. The most obvious virtue of predic- tions is that, by focusing the attention of the system on those offenders who are most likely to commit crimes in the fu- ture, they allow the community to main- tain tolerable levels of crime with less extensive use of imprisonment than would be possible without them. Given that it is desirable to reduce criminal victimization and to be economical in the use of the state's moral and financial re- sources in cloing so, it seems clesirable to exploit the focus on dangerous offenders that predictions make possible. Many view this apparent virtue as a dangerous temptation-one that will lure the community into increasing its overall demands for security at the expense ofthe rights and liberty interests of alleged criminal offenders and, therefore, at the risk of the overall quality of justice. But even viewed from the special perspective of protecting the rights and interests of accused offenders from the community's demancis for order, the use of predictions has virtues, for predictions can justify 3~5 more lenient treatment for some offencI- ers than their acts alone wouIcl justify. If an offencler's crimes seem uncharacteris- tic (anc! therefore unlikely to be repeated in the future), the current system (which is tolerant of predictions) can be lenient. This opportunity would be denied if pre- dictions of future conduct were excluded from criminal justice decision making. Finally, if interests in individual justice and aggregate efficiency continue to mo- tivate and sanction the widespread use of predictions in the criminal justice system, it would be valuable to recognize the  practice explicitly. That way, the society could guarantee that the predictions were made consistently, accurately, and use- fully rather Wan on an ad hoc basis. So, the question of whether conse- quential predictions are tolerable in our criminal justice system might not have a general answer. Some moral intuitions and ethical standards might exclude them entirely, while others would countenance them. For the ethical systems that tole- rate predictions, the particular form of the prediction may matter a great deal: some predictions may be more just than others. The central purpose of this paper is to develop moral intuitions about whether consequential predictions are tolerable in the criminal justice system, and, if they are, to establish what sorts and for what purposes. This requires an examination from several vantage points: from the per- spective of moral intuitions about the fun- damental values that animate the crimi- nal justice system and their connections to different systems of ethical theory; from an analysis of the tension between ideal standards and the implicit sanction granted to current practices by virtue of their traditional acceptability; from a de- tailed consideration of aspects of predic- tions that seem to have normative signif- icance; and from an inquiry into how the moral issues involving predictions differ

316 at different stages of the criminal justice system. THE ANATOMY OF PREDICTION IN THE CRIMINAL JUSTICE SYSTEM To fix conceptions, it is used! to clelin- eate the basic elements of prediction in the criminal justice system. Essentially, there are four: an offense, an associated offender, a predictive rule that links char- acteristics ofthe offender to predictions of future conduct, ant! a discretionary deci- sion to be made by a criminal justice official that could be influenced by pre- dictions of future criminal activity and that affects the criminal liability of the offender. The Offense A criminal offense is central because it is the thing that occasions the interest of the criminal justice system. Without a crime, there are no decisions to be made. Not even the most enthusiastic advocates of predictions in the criminal justice system would advocate the imposition of criminal sanctions without a criminal offense. The crime is also important because it constrains the decisions that will be made. A minor offense cannot be used to justify a major intrusion into the offencl- er's life even if the predictions are very ominous. Exactly how tightly the charac- teristics of the crime should bind the decisions of criminal justice officials is one of the major controversies surround- ing the use of predictions. Those who think that the justice of the system rests entirely on proportional and consistent responses to criminal acts seek to bind the decisions very closely to the act ant! to leave little room for consideration of the offender's characteristics and predictions of his or her future concluct (von Hirsch, 19851. Those who think that the overall CRIMINAL CAREERS AND CAREER CRIMINALS justice of the system requires some con- sideration ofthe character and future con- cluct of the offender will leave more room for these aspects to be considered in crim- inal justice decision making (Monahan, 1982~. But no one thinks that the nature of the offense is irrelevant to the decisions of criminal justice officials. Obviously, the offenses can vary along many dimensions. One is the gravity of the offense. It can be murder or petit larceny. It can involve serious injury to victims, threats of serious injury, or only minor property Tosses. A second is the certainty with which the criminal justice system has establishecl that an offense has occurred and that a particular offender  did it. This connection may have been definitively established through a crimi- nal conviction or persuasively alleged in a criminal indictment or simply sus- pected as a guide to investigative activity. In general, the more serious the offense and the stronger the established connec- tion to an offender, the greater the license criminal justice officials have to impose liabilities on offenders. Whether this in- cludes a greater right to make and use predictions about future criminal con- duct, however, remains unclear. The Offender The offender is also central to preclic- tions in the criminal justice system. With- out him, there is little of consequence for criminal justice officials to decide. It is most natural to think of the offender as someone who has just been convicted! and is awaiting sentencing. But the of- fender couIcl be at earlier stages of crim- inal justice system processing. He couIct be someone who has been indicted and is waiting to have bait set. He could be someone who has a strong evidentiary case against him and is awaiting a formal charge and indictment. Or, he couIc3 be a

PURBLIND JUSTICE leading suspect who is the focus of a police investigation. From the point of view of the criminal justice system, the most important at- tribute ofthe offender is the connection to a current offense. That is what makes him the subject of criminal justice action. But the offender has other characteristics as well. One of the most important is a criminal record. The criminal record may be nonexistent, or it may be quite exten- sive; it may involve only minor offenses or may include serious offenses; it may be a record of nothing more than arrests, or it may include convictions; it may be an adult record, or it may include crimes committed as a juvenile. In addition to a criminal record, the offender has such other characteristics as levels of alcohol and drug use, neighborhood ties, employ- ment status and experience, age, race, religion, political beliefs, favorite foods, and tastes in music. These characteristics differ from one another in several ways. One concerns their moral and legal status. Some characteristics, such as prior crimi- nal conduct and current illegal drug use, are themselves crimes and therefore of direct interest to the criminal justice sys- tem. Others, such as race, religion, and political beliefs, are the opposite: they are specially protected against being used by criminal justice officials in making deci- sions. Some characteristics, such as prior crimes, drug use, and perhaps employ- ment, are thought to be under the control ofthe offenders and therefore expressions of their inclinations and values. Other characteristics, such as age or race, are not under the control of the offenders and consequently are of little moral signifi- cance: they cannot be expressions of a person's character although they might be good predictors of future conduct. These offender characteristics also dif- fer from one another in terms of how accurately they can be determined for 317 individuals ant] how conveniently they can be observed. Some characteristics, such as employment history, are rela- tively objective and can be established and verified for inclividuals through intu- itively obvious, if laborious, methods. Others, such as psychopathic tendencies, may be relatively objective, but the meth- ocis used to validate them are special and arcane. Still others, such as community ties, are quite subjective and hard to establish, although one can develop operational measures of an intangible characteristic that can be objectively cle- termined. Similarly, some of the charac-  teristics of offenders are already known en cl recorded in files available to criminal justice agencies. Some can be inexpen- sively learned because they are recorded elsewhere or because the information is not carefully guarded by the defendant. But some characteristics can only be dis- covered through expensive and intrusive investigative efforts. A Predictive Rule The characteristics of offenders are im- portant for they form the basis for all predictions. All predictive tests have the same structural form: if an offender has a certain specified set of characteristics, that offender is predicted to be more (or less) likely to engage in future criminal activity than offenders with different characteristics. Any particular predictive rule has certain properties that are norma- lively significant. One important feature of the predictive rule is exactly which characteristics of offenders are selected to serve as preclic- tors. As noted above, the characteristics included in the test may be acts over which the individual has a great deal of control and are themselves criminal, or they can be characteristics over which the indiviclual has no control and, far from

318 being criminal, are given special protec- tion. The characteristics included in the test may be more or less convenient and accurate to observe for individual offenc3- ers. Somewhat less obviously significant is the fact that the test can include many or few characteristics. The more charac- teristics included, the greater the oppor- tunity to accommodate important individ- ual differences among offenders. But the more characteristics, the more comoli- cated the rule, and the greater the chance of errors. Beyond the properties of the set of characteristics incorporated in the test, the test has other features that are norma- tively significant. It is more or less accu- rate in terms of its capacity to identify all those offenders who will, in fact, commit crimes in the future and to exclude those who will, in fact, not commit crimes in the future. The test may be designed to iden- tify small (and therefore more unusual) segments of the offending population, or it may be less discriminating. The vaTid- ity of the test may be based on common sense, elaborate statistical investigations, or clinical theories and judgments. The test can be explicitly promulgated or im- plicitly used. It can be authorized by a legislature, established through ac3minis- trative guidelines, or sanctioned by com- mon professional practice. A Consequential Decision In adclition to an offense, an offender, and a predictive rule, the consequential use of prediction in the criminal justice system requires that an action be taken by a criminal justice official with respect to the offender. A sentence must be im- posed; bait must be set; a plea bargain offered; or an allegation pursued win more or less zeal by prosecutors and police. The consequences of these deci- sions register in three quite different c30- mains through different causal and evalua CRIMINAL CAREERS AND CAREER CRIMINaLS five systems. The decisions obviously affect the rights and liberty interests of the individuals who are affected. These may be either enhanced or climinished by the ef- fects of predictions. The decisions also af- fect overall levels of crime in the commu- nity through the mechanisms of deterrence (both specific and general), incapacitation, and rehabilitation. Ancl, finally, Me deci- signs affect the community's overall per- ception that justice is being done in terms of striking the right balance between the community's interests in security and the offencler's interests in freedom, and be- tween the (lesire to treat cases with consis-  tency and at the same time acknowledge important particular differences. The Central Ethical Issues The central ethical question raised by the use of predictive rules in the criminal justice system is whether an offender may be exposed to additional criminal liabfli- ties in the form of a longer sentence, higher bait, more determined prosecu- tion, or closer police scrutiny because of characteristics indicating that he is more likely than others to commit crimes in the future. A less fundamental but equally important question given the widespread current use of predictions is what kinds of predictions are better than others. These questions can be answered clirectly. But it seems that views about these questions are Inked to much broacler and more general notions of justice and of what constitutes a worthwhile improvement in the performance of We criminal justice system. It is as though the subject of prediction in the criminal justice system raises general moral connotations as well as specific normative issues. It is worth addressing these general ideas before ex- amining closely the specific ethical issues raised by prediction lest the influence of the moral connotations be decisive but unexaminecl.

PURBLIND JUSTICE GENERAL NOTIONS OF JUSTICE AND STANDARDS OF CRIMINAL JUSTICE SYSTEM PERFORMANCE Most normative discussions of the criminal law and the operations of the criminal justice system proceed on the basis of s here cl intuitions about the vir- tues of these social enterprises. The shared intuitions are captures! in a few words that stand for whole clusters of more particular icleas. Moral Intuitions of Criminal Justice One key virtue of criminal justice is its "fairness." At the center of the concept of fairness are notions such as the following: that citizens shouIc3 know in advance what actions will be punished and how alleged offenses will be investigated (Packer, 1968:801; that the system shouIc3 be consistent, i.e., treat like cases alike (Hart, 1968a:36-37, 1968b:24-25; Packer, 1968:139-145; Winston, 1974:1-39; von Hirsch, 1976:77-83; N. Morris,1982:179- 2091; that criminal liability for conduct should be ctistributec3 across possible acts according to the seriousness of the of- fense, not the social status or power of potential offenders (von Hirsch, 1976: 77~31; that people shouIc3 be held re- sponsible for things they can control and not for things they cannot control (Hart, 1968a: 158-185, esp. 174, 1968b:24-251; and that the actual operations of the sys- tem in imposing criminal liability should be unbiased with respect to race, social class, and other social variables (McNeely and Pope, 1981; Blumstein et al., 1983a: 8, 13-21; Klepper et al., 1983:5~128; Petersilia, 19831. At the edges, the concept of fairness shacles into the concept of justice. In- deecI, the concept of justice seems to incorporate all the particular ideas associ- ated with faimess. However, while the concept of fairness seems to emphasize 379 the distribution of criminal liability through the society, the concept of justice seems equally concerned about the amount of criminal liability and the intru- siveness of the means used to impose it. In a free society the concept of justice implies restraint a sense of proportion and frugality in using the coercive power and moral indignation associated with criminal sanctions (Packer, 1968:24 260~. Thus, particular ideas central to the notion of justice are those that give citi- zens significant rights against the state and against those who accuse them: for example, the right of citizens to be free from unwarranted searches and seizures  (McNamara, 1982:26-541; to confront their accusers in open trial (McNamara, 1982: 169-177~; and to have adequate time to prepare a defense (McNamara, 1982:21~2301. By establishing such rights for individuals, society constrains the amount en c] nature of state power that can be exercised against individuals on behalf of the community. To some, the notion of justice is not restricted to concern for the rights of de- fenclants. Arguably, justice is equally con- cemecl with protecting the moral stand- ing of the law and with guaranteeing that those who deserve punishment receive it (van den Haag, 1975:24-50; Weinreb, 1979:5; Carrington, 1983: 1~191. Some- times this position is described as one that protects "victims' rights" as well as "offenders' rights" (Bedau, 1977; Reiff, 1979; Carrington, 1983: 1(}12~. Insofar as the victim is interested in righting the wrong clone through retribution, this is an appropriate characterization. The crimi- nal justice system has many practical as well as moral reasons to accommodate the victim's interests in its proceedings (Greenwoo(l, Chaiken, and Petersilia, 1977; Blumstein et al., 1983a:41~. In our system of justice, however, it seems much more accurate to describe the obligation to administer just punishment as belong

320 ing to the state rather than the victim. As Aeschylus portrays in The Eumenictes, a great moment in Western history is the moment when the concept of justice changed from private vengeance to pub- lic retribution. In that moment the state took from victims the right to punish those who had offencled against them in the interests of ensuring an accurate de- termination of guilt or innocence and fair- ness and moderation in the imposition of penalties. In a free society the desire to mete out deserved punishment must be tempered. Punishments should fit crimes, not be excessive. The process of deciding whether a person is guilty or innocent shouIct be sufficiently cleliberate to pre- vent passions from overwhelming evi- clence. And the standard of proof should be set very high to ensure that those judged to be guilty are in fact guilty, even if that means that many guilty people are found innocent. It is all of these features that ctistinguish public justice in a demo- cratic state from either private vengeance in primitive societies or totalitarianism. Still, it is important to keep in minc3 that like private vengeance, public justice has passion ant! moral indignation as key in- gredients. Indeed, without these features, it is almost impossible to distinguish criminal sanctions from civil sanctions (von Hirsch, 1976:48; von Hirsch and Gottfredson, 1983-1984:341. Stan(ling somewhat apart from these traditional notions of fairness ant] justice is the notion that the system should be useful and effective as well as fair and just (Blumstein, Cohen, and Nagin, 1978~. To most people, this means that the system should succeed! in reducing crime and should do so at the lowest possible cost (Nagel ant] Neef, 1977; Blumstein, Co- hen, an(1 Nagin, 1978; Silberman, 1978; Anclreano and Siegfried, 1980:411-426; Wilson, 1983b). Some would acl(1 recluc- ing fear to the utilitarian purposes of the CRIMINaL CAREERS AND CAREER CRIMINaLS criminal justice system (Moore et al., 1984:~22, esp. 1~191. Most would also probably recognize the interests in fair- ness and justice as important constraints on the practical pursuit of reclucec! victim- ization and fear (Blumstein, Cohen, and Nagin, 1978; Sherman and Hawkins, 1981: 106~. And perhaps everyone would quickly agree that, to be effective, a crim- inal justice system must command the active support of the community, and that that, in turn, might depend on how fair and just it seemed (Weinreb, 1979:~12; AncTreano and Siegfriecl, 1980:8~921.  At the edges, an interest in an effective criminal justice system thus leads one back toward a system that imposes crim- inal sanctions with fairness anct restraint. Nonetheless, most people still see an im- portant distinction between a criminal justice system that is animated by a con- cern for justice anct fairness and one that is preoccupied with effectiveness. Specif- ically, it seems that the interest in effec- tiveness elevates the community interest in security over the interest in protecting the rights ofthe accusecl, allows estimates of aggregate social consequences to guide decisions that profoundly affect inclividu- als, an(1 leaves more room for social sci- ence and technology to be used to en- hance the efficiency or effectiveness of the system's operations, even at the cost of procedures honored by Tong tradition. All this makes the general idea of an effective criminal justice system quite clif- ferent from one animated by justice and fairness. Ethical Theory and Moral Intuitions About Criminal Justice The different intuitions about the vir- tues of a criminal justice system in a free society correspond to important differ- ences in modern systems of moral rea- soning. Modern ethical theory establishes a sharp distinction between "deonto

PURBLIND JUSTICE logical" and "utilitarian" philosophies (Frankena, 1973:12~601. The difference between them is that deontological theo- ries assert that an act is right or wrong in itself, regardless of its consequences. Utilitarian theories, on the other hand, assert that acts can be judgecI to be good or bad only in terms of their conse quences. The moral intuitions that lie behind the concepts of fairness and justice seem closer to the spirit of cleontological than to utilitarian reasoning. These intuitions see virtue in the criminal justice system inso- far as it acts properly with respect to accused citizens and ignores the practical consequences of its actions both for the defendant and the broacler community. The intuitions that prompt a commitment to effectiveness seem much more utflitar- ian in spirit. The concern for effective- ness finds the virtue of acts by the crimi- nal justice system in terms of their consequences for the future of the of- fencler and the future security ofthe com- munity. The link between moral intuitions about criminal justice and the different modes of ethical reasoning means that the normative standing of the different moral intuitions about criminal justice is inex- tricably linked to the general standing of these different modes of reasoning in eth- ical discourse. In general, it seems that the deontological systems have greater standing. Why this should be true re- mains unclear since philosophers have not as yet reached a decisive conclusion in favor of cleontological systems. The dominance of deontological moral sys- tems seems to reflect a general expecta- tion, rooted in tradition, that ethical pro- nouncements should take the form of rules prescribing conduct rather than ends that must be pursued. This, in turn, may be based on the notion that rules honor GocT or human traditions more re- liably than particular calculations, which 32] clepend so heavily on the qualities of the individual calculator; or on a prudential judgment that reliance on rules would avoid many temptations and errors that would otherwise corrupt the particular calculations; or simply on the intellectual appeal of reasoning from principles rather than concrete instances. NiVhat~ver the reasons, the general preference for deontological systems makes it hard for utilitarian arguments to be taken seri- ously in ethical arguments. Particularly in the criminal justice system, where the stakes for individuals seem so high and  where so much of the work involves the application of substantive rules to incli- vidual cases, utilitarian arguments seem a bit shabby. The sharp distinction between deonto- logical and utilitarian systems of reason- ing is unfortunate, for the challenge fac- ing those who guide the operations of the values shared criminal justice system is to integrate the values and concerns of each system of thought. In principle, this shouIcl not be difficult since our moral intuitions about the criminal justice system commingle deontological and utilitarian principles. As we have seen, fairness and justice are often clefenclec3 not simply as virtues in themselves, but also as qualities that en- hance the overall effectiveness ofthe sys- tem by drawing broad support from the community. Similarly, one can argue that the notion that the criminal justice system should be effective and economical in the use of state power anti money is not simply a shabby interest of the society, but a fundamental duty of those who gui(le, and operate within, criminal jus- tice institutions. It might be possible, then, to have a criminal justice system that successfully integrates the particular that are contained within and among our general moral intui tions. The clifficulty is that the schism be- tween (leontological en cl utilitarian sys

322 terns may make it difficult for us to see when a successful integration has been achieved. The integration will always look a little too unprincipled to a cleontologist and a little too ineffective to a utilitarian. And each will feel free to complain about the apparent corruption of the system viewer] from his or her vantage point. There may be no strong philosophical voice to step forward and say that the successful integration repre- sents a coherent view, because the suc- cessfuT integration will not fit wholly within either of the two systems of thought that have become familiar. Standards of Criminal Justice System Performance The general notions of fairness and justice, on the one hand, and economy and effectiveness, on the other, offer al- temative conceptions of the directions in which improvements in criminal justice system performance might lie: i.e., toward more consistent treatment of of- fenders, toward a more refiner] balance of community and incliviclual interests, or toward less use of the state's limited fi- nancial and moral authority to achieve the same amount of community security. They do not in themselves define tolera- ble levels of criminal justice performance in the pursuit of one or the other ideals. And yet the extent to which the current system realizes any of these idealized notions may be as important in judging the overall quality of the system as which icleal it is approximating. The most demanding standard for the criminal justice system is that it be an exact expression of an ideal system: that it be perfectly fair, perfectly just, or per- fectly economical. Although no one would really hold any human institution to these exalted stanclarcis, when one is talking about the criminal justice system, one is tempted to set the minimal stan CRIMINAL CAREERS AND CAREER CRIMINALS cards of performance very high and to be impatient with mere improvements in a basically corrupt system. The reason is that the decisions of the criminal justice system are so consequential for incliviclu- als (and for the overall character of the community) that the obligation to express the community's highest ideals is very strong. This is particularly true when one is talking about fairness and justice, for these qualities do not seem to exist in degrees. In common parlance, people conclude that the system is tolerably fair and just, or it is not. And any system of criminal justice that is unfair or unjust is  intolerable. So, our moral intuitions push us toward idealism in setting standards for criminal justice, and particularly so in the areas of fairness and justice. A different standard of justice would be whether the operations of the criminal justice system meet constitutional re- quirements. Often, this standard is con- fused with the first standard because many observers of the criminal justice system would like to believe that their idealized notions of justice are not only sanctioned by the Constitution but also required by it. Moreover, the room to make this claim often exists because the Supreme Court decisions that establish constitutional principles are rarer and less definitive than is necessary to banish ambiguity about constitutional issues. Nonetheless, one can distinguish what is clearly unconstitutional from something that is conceivably acceptable, and this provides a second standard of criminal justice performance. A third standard is simply whether a proposed policy or program constitutes an improvement in one or another dimen- sion of performance compared with cur- rent operations. Inevitably, all real sys- tems of justice fall short of idealized notions. They may also sometimes fall short of constitutional standards. Conse- quently, it may be important to know

PURBLIND JUSTICE whether proposer] changes in criminal justice system operations are moving us toward or away from an ideaTizecl concept of the criminal justice system. If a pro- posal promises to smooth the rough jus- tice that is actually meted out in the system even a little bit, it may be worth adopting even though the proposal fails to usher in a heretofore unattainable ideal. To a degree, these stanclarcis form a hierarchy for evaluating proposed changes in criminal justice policy. The most demanding is whether the proposed change is the final step in establishing an ideal system of justice. Much less cle- manding is that the proposed change fur- ther the aims expressed in the Constitu- tion. Less demanding still is that the proposed change be an improvement over current practices. These different stanciarcts of what con- stitutes a valuable or worthwhile im- provement in criminal justice operations are as important to keep in minct as the different intuitions about the substantive values that shouIc! guide the criminal jus- tice system, for they, too, become part of our discussions about whether predic- tions are tolerable in the criminal justice system. There are many proposals that might enhance the justice, fairness, and efficiency of the criminal justice system but that could be rejected because they fait to establish perfect justice. The cru- cial question is whether such proposals would be worth adopting. Once again, this important issue is af- fected by the difference between deonto- Togical and utilitarian systems of reason- ing. In principle, both schools have their "idealists" and their "realists." But the spirit of deontological systems is more given to idealism and exacting stanclarc3s. The spirit of utilitarianism, on the other hand, is quite tolerant of practical reali- ties ant] keenly interested in marginal improvements wherever they can be made. This means that those who want to 323 hold the moral high ground by sticking to the spirit of deontological systems will tend to establish very high standards across the board. Those who are inter- ested in encouraging small improve- ments in current operations might well be tarred with the brush of utilitarianism, even if the improvements they seek are in the areas of fairness and justice. My own position is that we all have a fundamental duty to encourage improve- ments in criminal justice system opera- tions in the directions of justice, fairness, and efficiency and to do so regardless of  how large or small the changes. That may seem far too utilitarian, or realistic, or pragmatic to have much standing in moral discourse. Ant! it is certainly true that this position wouIcl not only countenance but also enthusiastically embrace many pro- posals that seem shabby against the back- clrop of an i(leaTizecl system. But the weight of the duty to make improvements where they can be found can be mea- sured by asking what we would think of a criminal justice official who know- ingly abandoned some opportunity to im- prove the fairness, justice, or efficiency of the system without significant loss to so- ciety. ETHICAL ISSUES IN PREDICTION IN TtIE CRIMINAL JUSTICE SYSTEM The most fundamental objection to the use of prediction in the criminal justice system is that it is unjust and that any explicit or implicit use of prediction dis- graces our system of justice (Dershowitz, 1973:1277-1324; N. Morris, 1974:62~73; N. Morris and Miller, 1985:651. This po- sition is held by those who think that the most fundamental quality of the criminal justice system is justice (rather than effec- tiveness) and that a just system is one that holds people accountable only for acts committed in the past. This position has come to be callecl the "retributivist" or

324 "just deserts" position (von Hirsch, 1976: 124-127; von Hirsch and Gottiredson, 1983-1984:34~37; von Hirsch, 1984b: 17~194, esp. 1791. Re~ibutivist Attacks on Prediction A second objection to predictions of criminal activity is that they are inaccu- rate ant] that inaccuracy results in injus- tice to those offenders who are mistak- enly predicted to commit crimes in the future and are thereby exposed to unwar- rantec] penalties ant! liabilities within the criminal justice system (N. Morris, 1974:62-73; von Hirsch, 1976:21~36; N. Morris and Miller, 1985:24~36; von Hirsch and Gottfredson, 1983-1984: 1771. This view is often embraced also by retributivists, since it reflects their gen- eral suspicion of the reliability of social science technology, is consistent with their strong concern for individual justice for criminal offenders, and, in any case, offers an additional line of attack on pro- posals to use predictive methods more extensively. Nonetheless, this objection can be sharply distinguished from the first objection. The ctifference is that this second view does not say that it would be wrong to impose adclitional liabilities on those preclictec3 to commit offenses-only that it would be wrong to do this inaccu- rately. "False positives" are the prob- lem not liability being placed on incli- viduals for acts in the future. The first and second objections would effectively rule out the current use of prediction in the criminal justice system: the first because prediction is ruled out absolutely; the second because current predictive techniques cannot measure up to required levels of accuracy. Conse- quently, unless these objections can be overcome, the discussion of prediction is at an end. A third objection is less funclamental because it focuses on the characteristics CRIMINAL CAREERS AND CAREER CRIMINALS inclucled in the predictive rule rather than the appropriateness of prediction in general or the requirement that the pre- dictive rule meet a high stanciarc! of accu- racy to prevent injustice to individuals. By this standard, only certain characteris- tics of offenses and offenders may be incluclect in the predictive test. Appropri- ate characteristics are those that the indi- viclual controls and that themselves re- flect criminal concluct (such as prior offending and drug abuse). These are ap- propriate because they establish "morally relevant differences" among offenders (von Hirsch, 1976:212-213, 1981b). Inap-  propriate characteristics are those over which the individual has only limited or no control, that are not in themselves criminal conduct, ant] that are correlated with deprived social status, such as em- ployment status, race, or poverty. Also inappropriate are such variables as reli- gion or political views. Incleed, it may be so important that the criminal justice sys- tem avoid any taint of bias with respect to race, income, religion, or political views that it not only resist using these variables explicitly but also avoid variables that are correlatecl with these especially sensitive variables (N. Morris and Miller, 1985~. At the extremes, concern about the characteristics incorporates! in the predic- tive tests may make it impossible to con- struct any useful and decent test. This is particularly true if the tests must meet a high standard of accuracy and be neutral with respect to sensitive variables on a He facto as well as a He jure basis. So, scru- ples about the characteristics used in pre- clictive tests may not only reduce the practical value of the tests but also rule them out completely on normative grounds (Moore et al., 1984:7~791. Thus, from a retributivist position, the whole notion of predictions and particu- larly predictions that establish (liffer- ences among individuals on the basis of morally irrelevant characteristics is fun

PURBLIND JUSTICE damentally objectionable. Moreover, the objections work together. Restrictions on the characteristics to be used in tests render the tests less accurate. Less accu- rate tests exacerbate the problem of false positives. Both together create more in- justice and reduce the practical value of the proposals. And besides, it is always wrong to have any criminal liability at- tached to predictions of future conduct. Justifications fiom the "Modified Just Deserts" Position The common attack on the "retributiv- ist" position objects to its exclusive con- cern for a particular vision of justice and its idealism (Walker, 1982:27~2891. Es- sentially, the argument is that the retri- butivist position is fine in theory, but wrong in practice (N. Morris and Miller, 1985:21. Although it is commonsensical to base a system of justice on past acts, it also makes sense to exploit opportunities to control crime more effectively when they come along. Moreover, since the commu- nity expects the system to be effective, and since criminal justice officials will respond to this demancI by making pre- clictions, it is in the interests of justice to make sure that the predictions are made as accurately and decently as possible. In effect, while it might be wrong in theory to use predictions in the criminal justice system, the system can be made to per- form more effectively and more fairly than it does by explicitly introducing and managing predictive techniques (N. Morris and Miller, 19851. Although this hardly sounds like an ethical argument for the use of prediction, one can reason- ably argue that we have an obligation to make improvements in current perform- ance even if the improvements clo not usher in the ideal. This position which has been callecI "modified just deserts" honors retribu- tive principles and keeps them in the 325 dominant position in its conception of the virtues of criminal justice policies, but it also leaves some room for utilitarian con- cems and practical opportunities. And this, in turn, allows room for some kinds of prediction as part of criminal justice system operations. But the problem of inaccuracy in predictions and the inevitable injustice to inclividuals wrongly precticte(1 to commit offenses in the future remain. To this difficulty, those who take a "mollified just deserts" position seem to have two answers. One is that predictions of future criminality should be the basis for leniency and mercy but not for en-  tranced punishment (N. Morris, 1974:75, 1982:179-209, esp. 203~. Benign precTic- tions can mitigate, but a(lverse predic- tions shouIcl not aggravate, criminal pen- alties. Thus, no offender's liability would be increaser! by the use of prediction. This position not only alleviates concerns that offenders might be unjustly punished but also ensures the aggregate conse- quence that the scope of social control would not be widened by the broader use of preclictions. Unfortunately, this answer creates other problems. To the extent that justice establishes an affirmative obligation to punish criminal acts, being merciful to those who are predicted to be safe seems no more just than being harsh to those pre(lictecl to be dangerous. Obviously, the frailties of human institutions and judgments always counsel one to err on the sicle of leniency when moral juclg- ments are being ma(le and penalties ex- actecl, and this is what generally makes mercy a virtue (N. Morris, 1974:521. The point, however, is not that leniency and mercy are not virtues, but that they must be justified in in.clividual cases, and there is no guarantee that the characteristics of the offense or offender that might incline one toward mercy are those that distin- guish those who will be safe in the future (Moore et al., 1984:1011.

326 Similarly, granting mercy in individual cases on the basis of predictions of future criminal activity can create problems of equity ant] fairness. In principle, there is no problem if our notion of just penalties remains precise anct fixed, and if the bases for granting mercy are well estab- lishecI. But if (a) our iclea of just penalties is based on current average practices, (b) we think of penalties above the average practice as being unfair, and (c) there are many people in the system who seem to deserve mercy, many offenders who re- ceive penalties that are just when com- pared with the initial standard of justice will appear to be unjustly treated because they received penalties that were above the average. We may try to explain why they did not qualify for mercy but many others did ancI, in any case, why their punishment was cleserved. But this argu- ment still leaves them with the question of"Why me?" a question that has some force given the apparent inequity. So, the iclea of using predictive techniques to lessen but not enhance punishments may alleviate the concern that individual of- fenders will be excessively punished, but it increases concerns that some of those who deserve punishment will not receive it and that clefencIants who have commit- ted similar offenses might be treated clif- ferently. The second answer to the problem of false positives offered by those who hold a "mollified just deserts" position is that most people have misconstrued the na- ture of prediction. In this view, predict- ing that a person is dangerous floes not necessarily imply that a person will com- mit a criminal act; it implies only that a criminal act is more likely (Flout and Young, 1981:4~49; N. Morris and Miller, 1985:24-28~. Therefore, if an offender predicted to be dangerous does not, in fact, commit an offense, the prediction was not necessarily wrong. In this view: "a prediction of dangerousness . . . is the .b CRIMINAL CAREERS AND CAREER CRIMINALS statement of a condition (membership in a defined group with . . . certain at- tributes) and not the prediction of a re- suZt (of future violent acts in each indivicl- ual case)" (N. Morris and Miller, 1985:241. Thus, the crucial factual ques- tion in making predictions is not whether a person will commit offenses but whether that person sloes or does not have the attributes that qualify for mem- bership in a particularly high-risk or par- ticularly Tow-risk group. Since that factual  question can typically be answered with great accuracy, there are very few errors ~ . . . .. In prec action. This explanation sloes away with wor- ries about mistaken predictions but raises a new question: namely, why people who unambiguously have the attributes that make them members of a group predictecI to be active offenders in the future shouIcI be exposed to adclitional penalties and liabilities from the criminal justice sys- tem. The answer to that is sometimes cast in the language of "just deserts." As Flouct and Young (1981:48) observe: "the fact that if we were to set (those offenders preclictec3 to be dangerous) at liberty, only half of those we are at any time detaining as dangerous would clo further serious harm, does not mean that the other half are all in this sense innocent [emphasis arlcled]." They have the qualities that make them risky and (langerous in the same sense that all unexploded bombs are dangerous even though most never explode (N. Morris anct Miller, 1985:25; von Hirsch, 1985:2801. And it is those qualities that justify different treatment in both moral and pragmatic terms. While this, at first, seems appealing, and while it is not hard to imagine the practical inter- ests we have in treating such offenders lifferently, it is clif~icult to answer the question of exactly why they are not in- nocent. To put the matter differently, it is hard to say of exactly what offense such offenders are guilty.

PURBLIND JUSTICE To avoid this problem, the argument for treating clifferently offenders who have qualities that place them in high- risk groups quickly shifts to practical in- terests. Such offenders are risks that the society must bear. The crucial question is who will bear the costs of that risk: the offenders who must give up their freedom to protect the society or the society that must live with the risk of having cianger- ous people at liberty (N. Morris and Miller, 1985:24-36, esp. 281. This is con- siclerec! a policy problem to be resolver! by legislatures balancing social interests rather than either a constitutional ques- tion or a matter of incTiviclual justice to be cleciclecI by judges, prosecutors, or police (N. Morris and Miller, 1985:351. Once the legislature strikes the balance ancT defines the groups, the system can implement the . . . po lcles In gooc conscience. While this argument has some aspects that make it seem a principlec] position, the fact of the matter is that this argument pushes the justification of prediction far from the retributive perspective anc! puts it at the center of utilitarian concerns. In essence, the argument is the following: There is a social problem to be managed that consists of people who are inclinecI, or at least willing, to hurt or threaten other citizens or to take their property. This makes social life unpleasant. The society has a right to take action to protect itself from the risks. Because the risks come from other citizens, however, ef- forts to manage the risks must take ac- count of their rights ant! interests as well. One way to accomplish that goal is to limit social control to those who have qualities that indicate they are much greater risks than others. To protect the rights and interests of those who repre- sent greater risks, it is crucial that there be an evidentiary hearing on whether the offenders do or do not have the requisite qualities and that the scope of the state's penalties and controls be commensurate 327 with the magnitude of the risks such of- fenclers represent. Ideally, legislatures rather than indiviclual administrative offi- cials shouIc3 balance these competing in- terests. Obviously, this position simply drapes the utilitarian argument for pre- cliction in the clothing of legislatively balanced risks and clue process protec- tions for those who are about to be ex- posecl to enhanced criminal liabilities on largely utilitarian grounds. Justifications from a Utilitarian  Perspective The most direct counterargument to retributive objections is simply to assert not only the relevance but also the dom- inance of utilitarian concerns in the cle- sign of the criminal justice system. In this conception the criminal justice system has not only an interest but also an af- firmative obligation to use whatever is available to reduce crime and promote security (Blumstein, Cohen, and Nagin, 1978:~14~. Because it is important to enlist the support of the community, and the community is as worried about exces- sive state power as about criminal of- fenclers, it is prudent for the criminal justice system to be restrained, to protect rights to clue process, and to operate con- sistently with the community's moral in- terests in a just and decent criminal jus- tice system. But the touchstone for all innovations in the criminal justice system is to enhance effectiveness in reducing crime and promoting security. In this utilitarian conception the con- cern about punishing people for pre- dicted future offenses disappears en- tirely unless the means seem so grossly unjust and so bizarre as to be repugnant to the community. Similarly, the concern for "false positives" fades, but does not en- tirely disappear. It stays partly because even the most pragmatic utilitarian might see some moral virtue in punishing only

328 those who deserve it. Besides, economy in the use of both the moral and financial resources of the government would be considered a virtue by utilitarians, and there is no reason to waste those re- sources on people who are unlikely to commit offenses in any case. So, if one were to adopt a predominantly utilitarian position, the most fundamental objection to making criminal justice system actions contingent on predictions of future crim- inal conduct would be overcome. Some scruples might remain about the charac- teristics user! in making predictions. For example, race, religion, and political views might be excluded to enhance other social purposes. But much greater room would be created for the use of predictions. VIRTUES OF PREDICTIVE METHODS AND RULES So far, the basic objections to the use of predictive rules in the criminal justice system have been examined from the vantage points of retributive, modified just deserts, and utilitarian ethical posi- tions. As one moves through these clif- ferent positions, more scope for precTic- tion is creates! largely because concerns for"effectiveness" and some aspects of "fairness" gain in relative importance to concern for "justice." This suggests that one way of deciding whether prediction is ethically acceptable is to decide first on one's general ethical position and then see whether it allows prediction. A slightly different way of thinking about the ethical issues raised by predic- tion is simply to imagine what the virtues of predictive methods might be. Obvi- ously, if one is a strict retributivist, this exercise holds lithe interest since even the most virtuous system of prediction would be ruled out as unjust. Similarly, if one is a basic utilitarian, the exercise CRIMINAL CAREERS AND CAREER CRIMINALS hoIcts little interest since the best system must be the one that produces the great- est reduction in crime for the smallest use of the moral authority and financial re- sources of the state. But if one is a practi- cally minded retributivist or a principled utilitarian, identifying the virtues of a system of prediction has some appeal because it not only identifies the qualities that would make the system acceptable but also indicates where and how im- provements might be made. Attractive Qualities of Predictive Rules In thinking about the qualities that  make systems of prediction more or less acceptable or virtuous, it is worth clistin- guishing features of the predictive rule itself and the circumstances under which the rule is appliect. At least five important qualities of the predictive test or rule can be examined. Focus of the Predictive Test The first important question focuses on the behavior that the rule is trying to predict and the distinctiveness of the population that is being singlecI out by the rule when it works well. In general the more important the behavior that is being predictecl and the smaller and more distinctive the population that is being singled out, the more appropriate seems the use of the predictive rule. This principle applies in both retribu- tive and utilitarian ethical systems. In the retributive conception a tolerable iclea might be to single out the most wicked offenders (those who are most callous and show the fewest signs of remorse) and expose them to special punishments. Moreover, this seems much more appro- priate if the rule singles out only a few who are outliers in the distribution of all

PURBLIND JUSTICE offenders, rather than a substantial minor- ity of offenders who are much closer to the center of the distribution in terms of "wickedness." Similarly, in a utilitarian framework the value of a predictive rule goes up if it isolates the worst kinds of criminal offenses and if it identifies the worst 5 to 10 percent of offenders, rather than more ordinary offenses and offend ers. The narrow focus is clesirable in both systems partly as a matter of economy in the utilization of the state's limited moral and financial capacity to punish and partly because the more serious the con- cluct and the smaller the population that is identified, the more plausible the argu- ment that the offenders are at least quan- titatively and perhaps qualitatively cliffer- ent from ordinary offenders and therefore deserving of special treatment. Thus, a nar- row, discriminating focus is to be preferred on both retributive and utilitarian grouncts to predictive rules that place more offencl- ers in special categories. Accuracy of the Predictive Test A second important feature of the pre- dictive rule is its accuracy. As noted ear- lier, two quite different notions of accu- racy exist. One is the idea that the rule does in fact predict who will commit serious offenses in the future: the actual conduct of the offenders determines the truth or falsity of the prediction. The sec- onc! is that an accurate assessment is made of whether an offender does or does not have the characteristics that qualify for membership in a group preclictec] to engage in unusually high levels of crimi- nal activity. In this conception the actual conduct ofthe offenders is not considered relevant: they are "dangerous" if they have the proper characteristics. The first conception of accuracy is central to some notions of"justice" and to all utilitarian 329 concerns. The second is fundamental to notions of justice as fairness. This discussion of accuracy is limited to the first notion: that the rule predicts accurately who will commit serious of- fenses at high rates. Obviously, accuracy in a predictive rule is a virtue in retribu- tive systems because it minimizes the problem of exposing people who are not in fact dangerous to whatever special lia- bilities attach to this designation. It is a virtue in utilitarian systems because it economizes on the use of the state's re- sources in producing crime-recluction benefits. The difficulties with this concept arise  because a predictive rule can be inaccu- rate in two ways: it can incorrectly iden- tify as dangerous offenders who are not dangerous (so-callecl "false positives"), an(1 it can incorrectly identify offenders who are in fact dangerous as not danger- ous (so-callecl "false negatives". Liberal democratic societies, acutely aware ofthe frailties of human institutions, have typi- cally treate(1 "false positives" as much worse than "false negatives." So, a test that is attractive is not only one that makes few errors of both types but also one that distributes the errors in an appro- priate way i.e., makes many fewer er- rors of inclusion in the category of clan- gerous offenders than of exclusion. People's views differ about the proper tradeoffs between reducing errors of all types an(1 reducing errors of one type at the expense of increasing the total num- ber of errors of both types. So do their views about the rates at which they will trade one kind of error for the other in a world in which the total number of errors of both kinds remains constant. But ev- eryone agrees on the directions in which improvements lie: fewer errors are better than more; errors of inclusion are worse than errors of exclusion when the of- fencler is to be given special penalties or

330 control by virtue of inclusion in a group predicted to be dangerous. Basis of the Predictive Rule A third important quality of the predic- tive rule is the basis on which it is estab- lished. The standard distinctions in this area are made between rules established on the basis of"statistical" or "actuarial" methods and those established on the basis of"clinical" methods (Monahan, 1981:4~93~. Norval Morris and Marc Miller (1985:18-19) have added a third kind of prediction, which they call "anamnestic," that is, a predictive rule is developed for an individual on the notion that individual behavior is repetitive and thus becomes predictable to those who know the individual very well without necessarily being generalizable to others. At first it might seem odd that the basis of a predictive rule would hold much ethical or normative interest. Of course, we might assume some connection be- tween the basis of the rule and its accu- racy. And, to the extent that we thought accuracy was important and had views about which basis produced the most ac- curate rules, the basis of the rule would assume normative significance. But the significance would be exhausted by an examination ofthe rule's accuracy regard- less of its basis. Yet many commentators seem to attach significance to the basis of the rule beyond its implications for the accuracy of prediction (Monahan, l9S1: 9~101). On reflection, this concern seems to be tied to three features of the predictive rule that are linked to moral intuitions about the just construction of such rules. One notion is that, if predictions are to be made, they should emerge from a unique consideration and understanding of the individual (N. Morris and Miller, 1985: 201. This honors the principle of individ- ualized justice (but it sometimes jeopar CRIMINAL CAREERS AND CAREER CRIMINALS dizes, or at least complicates, the princi- ple of like cases being treated alike). By this standard, "anamnestic" and "clini- cal" predictions, both of which are based on detailed case information, might be preferred to "statistical" methods, which concern aggregates and abstract from in- dividual circumstance. A second notion is that it should be possible to state the predictive rule sim- ply and to have it conform with ordinary common sense. This is consistent with aspirations for "fairness" in the system and for mobilizing community support for the operations of the criminal justice sys-  tem. By this standard, "anamnestic" rules are once again dominant, "actuarial" rules are close behind (depending on how commonsensical they appear), and "clinical" rules appear the least attrac- tive. The third notion- closely related to the second is that the development and in- terpretation of the rules should minimize the use of specialized professionals. This is primarily to protect the connection of the criminal justice system to the commu- nity and to tradition but also perhaps to maintain the professional dominance of lawyers over other professionals in the criminal justice system. By this standard, anamnestic predictions once again seem the best; actuarial and clinical predictions are far behind because both involve ar- cane methods and different kinds of pro- fessionals. So, the basis of predictive rules seems to be important, independent of their prospects for accuracy. When all charac- teristics associated with the basis of rules are considered, most commentators seem to prefer statistical methods (Meehl, 1954; Floud and Young, 1981:26; Mona- han, 1981:97-98; N. Morris and Miller, 1985:20~. Rules established by such methods have the virtues of calibrated accuracy, simplicity of form, and consis- tency of application. They have the liabil

PURBLIND JUSTICE ities of being indifferent to most inclivid- ual characteristics (which is the opposite side of the coin to simplicity), of being rooted in aggregate rather than individual experience, and of engaging unfamiliar techniques. Anamnestic rules, on the other hand, have the virtues of being rooted in indiviclual experience, respon- sive to incliviclual circumstances, and commonsensical. They have the liabili- ties of unproven accuracy and uneven application. Clinical rules have the virtue of being responsive to individual circum- stance and the liabilities of being inaccu- rate, of being complicated to state and to apply, and of surrendering some of the powers of the criminal justice system to a (suspect) group of professionals (Stone, 1975~. Characteristics Used to Make Predictions A fourth quality of predictive rules that bears on their fairness, justice, or efficacy is the character of the variables that are used to assign people to groups and to make predictions. This point has already been discussed. From a retributive per- spective, the only appropriate variables are those that an incliviclual can control and are themselves reflective of criminal conduct although not necessarily the most serious forms of criminal conduct. From a utilitarian perspective, variables are appropriate to include if they are successful in predicting criminal conduct. From a mixed perspective, the challenge is to balance interests in having the char- acteristics used in the test be just and in predicting reliably. There is some consensus about what variables may properly be included. Ev- eryone agrees that the seriousness of the current offense is proper to consider in 1974:73; von Hirsch, 1976:Chapters 8 and 9; Blum- stein et al., 1983a: 11-12, 83 841. The sentencing (N. Morris, 337 main reason is that since it is the offense that justifies the punishment, the serious- ness of the offense must determine the seriousness of the punishment. The seri- ousness of the offense is often judged not only on the objective harm (lone by the offender, however, but also on the state of mince of the offender (Hart, 1968b:11 135; von Hirsch, 1976:80~. If the violence was particularly wanton or if the offender behaved very recklessly with respect to life and property, the penalty (anc] per- haps future suspicion) will be greater than if the offense was more moderate  (Vera Institute of Justice, 19771. In short, the offense itselfmay indicate the danger- ousness ofthe offender as well as produce the objective harm to victims that justifies intervention by the state. Nearly everyone also agrees that the adult record of the offender may properly be included (von Hirsch, 1976:84-94, 1981a). The only people who disagree with this position are the most strict retributivists, who think the right (anc! the obligation) to punish is tied strictly to acts and that punishment is meted out to bal- ance the wrongs clone. In their view each act deserves a discrete penalty ant] to enhance the penalty for a third or fourth offense is to be unjust (von Hirsch, 1976:172; Fletcher, 1978:46~466; Sing- er, 1979:67-74~. Other retributivists think that it might be just to enhance penalties for those with criminal records not be- cause criminal records necessarily pre- clict well but because they reveal the offender as unusually persistent and therefore unusually (reserving of punish- ment (von Hirsch, 1976:84-944. Thus, while these reasons for considering crim- inal record are different from those held by the "morlifiecl retributivists" and the "utilitarians," many retributivists would allow criminal record to influence the extent of punishment and control as- serted by the system. The utilitarians ap- prove of the use of criminal record be

332 cause it is correlated with future criminal offending (Monahan, 1981 :10~105~. And the "modified retributivists" accept the iclea because this variable fits within the principle that the variables used to pre- dict should be uncler the control of the offender and consist of conduct that is itself criminal (Monahan, 1981: 10~105; N. Morris and Miller, 1985~. Virtually everyone but the hardest-core utilitarians also agree that there are some variables that should clearly be excluded from any predictive test. Such variables wouIc3 be those that define groups that have special protection under the Consti- tution, such as religious groups, political organizations, and groups that have his- torically been the object of discrimination (e.g., racial groups anal, perhaps, age groups). It is quite clear that the de jure use of characteristics such as religion, political beliefs, and race are ruler! out (Wilson, 1983b: 1581. It is more controver- sial whether to countenance the use of . that the characteristics used in predictive tests should be as far removed from any taint of political, cultural, or racial bias as possible. These points mark out areas of agree- ment. The field of contention is wicler, however. Some of the disagreement fo- cuses on the degree of certainty one must have about whether an offender actually has a certain attribute to be able to use it in making predictions. This arises most sharply and obviously in the use of crim- inal record; the issue is whether the pre- dictive test should be restricted to convic- tions or whether it might also inclucle indictments and arrests. There is a strong argument for relying only on convictions: since they are the only criminal acts that have been confidently attributer! to an CRIMINAL CAREERS AND CAREER CRIMINALS offender, they are the only acts that could justify any additional penalty or control. The argument for allowing indictments and convictions is weaker and relies much more heavily on a utilitarian justi fication: since indictments and arrests can only be made on the basis of enough evidence to establish "probable cause" to believe that an offense occurred and that the particular suspect committed the of fense, since inclusion of information on indictments and arrests seems to improve the accuracy of predictions of future crim inal activity, and since this information is aIreacly widely used in the criminal jus tice system, it is tolerably just to use this  information. Indeed, it may be much bet ter to rely on indictments and arrests, which have the virtues of having some relationship to criminal conduct and of being recorder] relatively accurately, than to rely on characteristics (e.g., drug use or employment status) that do not necessar ily reflect serious criminal conduct, are variables that might themselves be unreliably measured, and may be only proper but are sufficiently correlated with imperfectly uncler the control of the of other characteristics to result in de facto fencler. discrimination if utilized (Moore et al., This raises the second main area of 1984:731. At any rate, everyone agrees disagreement: how close to criminal con duct must the behavior be and how con ficlent must one be that the behavior was under the control of the individual (Hart, 1968b: 174; Underwood, 1447~. These issues arise most a~rec~y when we consider the appropriateness of incorporating variables such as drug use and employment status. Drug use seems closer to acceptability than employment status because it is closer to criminal con cluct and much more under the control of the individual than employment status. But one can reasonably argue that drug use in itself is only criminal by virtue of laws that make it so; that many drug users have lost control over their use; and that, in any case, it is hard to measure accu rately for individuals (Wish et al., 1981~. Hence, it wouIc! be unjust to make levels 1979: 1432~ ~ - . 1

PURBLIND JUSTICE of punishment and supervision concli- tional on drug use. And, if this argument is accepted for drug use, it would also exclucle use of employment status for there is nothing remotely criminal about being unemployed (although it is cer- tainly a virtue to be employecI) and peo- ple may have relatively little control over this status (although there may be oppor- tunities for them to work or to make investments in themselves so that they will be qualified for employment). Perhaps the most interesting area of disagreement involves the use of juvenile records. On one hand, from a retributive perspective, a juvenile record of criminal offenses seems appropriate to use be- cause it is part of a criminal record incli- cating persistent criminal activity. On the other hand, we tend to view juvenile offenses as less under the control of the offenders-and therefore less indicative of intent and character than aclult of- fenses (Institute for Judicial Administra- tion, 1977:11. Moreover, we have institu- tionaTized this conception of diminished criminal responsibility by establishing ju- venile courts, which clo not finct juveniles "guilty" of specific offenses, but instead find them "delinquent" or "nonclelin- quent," and do so through relatively in- formal systems in which records are deliberately kept spare to avoid future stigmatization and labeling (Institute for Judicial Administration, 1977:25(~252; Zimring, 1978:4~49, 66 691. Since juve- nile offenses are conceived to be less under the control of individuals and since they are measured imperfectly, it wouIc3 be unjust and unfair to use them in pre- dictive tests. But there is an additional part of this issue that is emphasizer] by utilitarian interests and concerns. Much criminolog- ical research indicates that rates of of- fending peak for individuals between the ages of 18 and 25 (Collins, 1978; Moore et al., 1983b). Moreover, those who are very 333 active and violent offenders in this period tend to have accumulated serious juve- nile records (Moore et al., 1983b). Hence, if juvenile records were used as part of the predictive tests, the tests would iden- tify the most clangorous offenders not only more accurately but also earlier than they otherwise wouIc3 be. In fact, they couIc3 be identified during their peak years of offending. If, on the other hancI, juvenile records are excluclec] from the predictive tests, the system will identify people as dangerous offenders less accu- rately and later in the individual careers  of the offenders. This means that some important crime-controT potential is lost (BolancI and Wilson, 1978:22-351. As in other areas of normative debate, the question of which variables are proper to use in predictive tests comes (town to the balance between retributive en c] utilitarian principles. It seems clear that current offense and prior adult con- victions can be used. It also seems clear that race, political views, and religious beliefs may not properly be usecI. After that, a great clear is contested. The Harvard Project on Dangerous Offenders concluclecl that indictments for adult offenses could properly be included since they dic3 represent evidence of criminal conduct and were routinely used in sentencing anyway but that employ- ment status and history should not be included (Moore et al., 1983a: 132, 1984: 74-751. The Harvard project also concluder] that juvenile records of serious criminal offenses could be included if the offender committed an a(lditional serious offense shortly after graduating from the juvenile system (Moore et al., 1983b:32~327, 1984:173-1761. This position was justi- fied with both a retributive and a utilitar- ian argument. The retributive argument was that, while there was a presumption that juvenile offenders were not respon- sible for their offenses in the same way

4 334 that aclults were because they clid not will them, this presumption was vitiated by those few juveniles who persisted in criminal offending as adults, because their persistence gave a clifferent mean- ing to their juvenile offenses. Viewed in retrospect, the youthful offenses were not indiscretions occasioned by the conflu- ence of circumstances, peer pressure, and transient recklessness but, instead, were early signs of cleterminec3 criminality. The utilitarian argument is the following: since the main reason to seal juvenile records is to relieve youthful offenders of the stigmatizing burden of past offenses and make it easier for them to stop of- fending, ant] since those youthful offend- ers who continue committing serious crimes have aIreacly failer] to take ac3van- tage of that opportunity, no practical pur- pose is served by continuing to protect their juvenile records and some practical purpose is lost by not exposing their rec- orcis of serious offending as juveniles. So, both retributive and utilitarian arguments line up in favor of including records of serious offenses committed as juveniles for those offenders who continue to com- mit crimes as adults. The basic logic that leads to these con- clusions on particular characteristics is the desire to keep the predictive rules close to retributive principles, and per- haps even to improve the justice and fair- ness of current operations, while at the same time exploiting some of the crime- contro! benefits that might come from improved preclictions. Thus, characteris- tics involving criminal conduct were treated as more acceptable than variables describing noncriminal conduct or sta- tuses, and concessions to utilitarian inter- ests in crime control were macle on the basis of the accuracy with which such variables were measured rather than in terms of the nature of the variables them- seIves. This position undoubtecITy goes too far for retributivists and not far enough for utilitarians, but those features CRIMINAL CAREERS AND CAREER CRIMINALS may be the virtues rather than the vices of the position. Auspices of the Prediction Rule The fifth quality of a predictive rule that affects its acceptability is the aus- pices that establish it as a guide for crim- inal justice decision making: i.e., whether the rule is promulgated by a legislative body, a court, or an administrative agency. The rule has different kinds and  degrees of legitimacy clepencling on the source that established it and the process that lay behind its establishment. In general, we think of legislative bocI- ies as having the broadest kinds of re- sponsibility and legitimacy. As represen- tatives of the people, they are competent to assess current problems, weigh alterna- tive solutions, and balance competing so- cial values at stake in alternative policy responses. Moreover, in reaching conclu- sions, they are free to consult wiclely- inclucling specialists in legal reasoning, in statistical methods, and in psychiatry. Thus, in principle, when they reach a decision about a proper predictive rule, that decision should carry great weight. It can be changed only by a successful ar- gument that an important constitutional principle was violatecl-a judgment that is fairly rare. In practice, though, we often worry that legislatures are too responsive to transient passions of the majority; that important traditions or rights of minori- ties ant! individuals might be over- whelmed; and that important scientific and technical issues might not be well enough understood. In effect, legislatures might have the un(leniable virtue of re- flecting the people's will but might fait to take advantage of institutions that em- bocly other virtues. Courts might be a better author of pre- dictive rules. They typically lack the close connections to the political commu- nity that legislatures have and might well be as incompetent as legislatures in acl

PURBLIND JUSTICE dressing the technical issues of statistics and psychiatry that could arise in formu- lating a predictive rule. But they have the virtue of representing tradition and a creep concern for the rights of minorities en c! individuals. Ant! since these are im- portantly at stake in the design and use of predictive rules, perhaps the courts are the most legitimate authors. Alternatively, administrative agencies- either correctional systems, criminal jus- tice planning agencies, or specially estab- lishec] commissions-might be the proper authors of predictive rules. While they lack the close connection that courts have to tradition ant! individual rights and the close connection that legislatures have to the popular wilt, they are assumed to have the virtue of being able to command technical expertise. And since there are a great many technical issues to be clis- cussed and resolved in formulating a pre- dictive rule, and since neither courts nor legislatures provide an appropriate forum for this debate, perhaps the rules should be formulated by administrative agen- cies. The ideal wouIcI be a legislatively es- tablished rule, formulated through a leg- islative process that effectively synthe- sized the perspectives and expertise of professional criminal justice administra- tors, judges, lawyers, statisticians, and psychiatrists. Anything short of this would be distinctly inferior. But probably the worst situation is one in which either courts or administrative agencies formu- late their own predictive rules without the benefit of connections to the political community or to the knowledge of tech- nical experts. And that seems to be the most common contemporary source of predictive rules. Attractive Qualities in Applying Predictive Rules Obviously, predictive rules are vuIner- able to a great many vices: they can be too 335 indiscriminate, too inaccurate, built from inappropriate methods, based on unjust characteristics, and promulgated by the wrong agencies. Perhaps some of these vices can be overcome by virtues in ap- plication. Some commentators have suggested, for example, that some degree of inaccu- racy or some flawed characteristics or some informality in the construction and promulgation of the rule might be accept- able if the action to be taken by criminal justice officials was relatively insignifi- cant (N. Morris and Miller, 1985:20,  30~331. In effect, it is appropriate to think of a balance to be struck among the im- portance ofthe social objective being pur- sue(1 through the application of the pre- dictive test, the size of the sanction to be applied by criminal justice officials, and the requirements placecl on the test itself. The more important the objective ant! the smaller the infringement on the interests of offenders, the less clemanding the stan- dards for the predictive test. If, for exam- ple, it was plausible that the use of pre- clictive tests might substantially reduce the likelihood of a presidential assassina- tion, and if the predictive test resulted in nothing more than refusing admission to a public speech by the president, a quite imperfect test could be used (N. Morris and Miller, 1985:311. If, on the other hand, the society sought to eliminate "joyricling" and wanted to JO so by plac- ing those teenagers predicted to be active joyriders under close, continuing supervi- sion, even predictive tests that met the strictest possible standards might be un- acceptable. Virtue in application comes in balancing competing interests, and the s ten dark of what sorts of test are accept- able is somewhat elastic. A great clear depends on the size of the harm to be avoided and the magnitude of the penal- ties or controls exerted over those who are the subjects of preclictions. It also seems clear that predictions are more acceptable if they are made after an

336 offender has been convicted of a crime. The preclictions of dangerousness associ- ated with civil commitments have always been suspect (Dershowitz, 19741. The ob- vious deficiencies of the predictive rules in use have been justified by the assertion that civil commitment was therapeutic and in the interests ofthe person who was committed. As it has become apparent that the "treatment" available to those who were civilly committed was virtually inclistinguishable from the "punishment" metecl out to those judged guilty of crimes, however, this justification has worn thin, and the entire iclea of creclic- tion has been tainted with the hypocrisy and overreaching of civil commitment procedures (Dershowitz, 1974; Stone, 19751. On the other hand, predictions have long been tolerates!-even enthusi- astically endorser]-in making sentenc- ing decisions once an offender is con- victed of a crime (Williams v. New York, 337 U.S. 241, 247, 1949~. No doubt some of this enthusiasm for prediction comes from the expectations that predictions will mitigate rather than aggravate sen- tences and that explicit predictions might introduce some consistency into haphaz- arc3 patterns of sentencing. But the wider scope given to the use of prediction in sentencing also has a great deal to do with the fact that the liberty interests of an offender are taken less seriously by the society once he has been convicted of a criminal offense. Our laws ant] moral in- tuitions establish a fairly broach zone of discretion in limiting the freedoms of those who have been convicted of crimi- nal offenses, anct there is less objection to using prediction to fill out this zone of discretion than there is to using preclic- tions as the sole basis for restricting a citizen's liberty even if the purpose is terribly important. These two principles of application- that the quality of the predictive tests or rules shouts] be commensurate with the CRIMINAL CAREERS AND CAREER CRIMINALS significance of the harm to be reduced and the sanction to be imposer] on incti- viduals predicted to be dangerous and that prediction is more justified when it involves people who have been con- victed of criminal offenses have impor- tant and subtle implications for the appro- priateness of using predictive methods at different stages of the criminal justice system. These will be discussed in cletai] in the next section, but it is worm begin- ning the discussion here. The principle that predictive tests are more (or only) appropriate when they  involve convicted offenders seems to im- ply that predictions are appropriate at the sentencing stage and inappropriate at any stage prior to sentencing (e.g., investiga- lion, prosecution, or bail). After all, to use predictions before that stage is to violate the presumption of innocence and to ex- pose innocent people to heightened state interest and control- including the loss of liberty in pretrial detention. Such actions lack even the thin justification of "treat- ment" available within civil commitment procedures. To a degree, one can argue against this view by insisting that the "presumption of innocence" is an important principle to be used in criminal trials but certainly not as a guicle for the agencies that investi- gate crimes, prosecute offenders, or seek to ensure appearance at trial. In fact, their task is generally the opposite: to clevelop evidence that constitutes a case showing that a given person is very likely to have committe(1 an offense. In allocating re- sources and pursuing their objectives, it might be proper for them to make preclic- tions of whether an offender is likely to be an active offencler. While this argument has some merit, it sounds a bit ten(lentious even to those who support proposals to use pre(lictions of dangerousness at the stages of investi- gation, prosecution, and pretrial decision making. What gives the argument adcle

PURBLIND JUSTICE weight, however, is the observation that predictive methods wouIcl most com- monly be used at these stages for people who had been convicted for serious of- fenses at some time in their careers and generally quite recently (at least in terms of"street timed. This raises the question of whether convictions for offenses in the past couIc3 justify exposing offenders to the use of predictive methods to vary levels of investigative ant] prosecutorial effort even though the current offenses of which they are suspected are so far un- proven. Strict retributivists, who view the offencler's liabilities as exhausted when they complete their punishment for ore- vious acts, might object to this iclea. So wouIc3 some utilitarians who are inter- ested in maximizing offenders' chances for rehabilitation ant! who would regard the heightened police and prosecutorial interest as antitherapeutic. But to many, the idea that a person convicted of previ- ous crimes and plausibly accused of more recent crimes shouIc3 receive greater at- tention from investigators and prosecu- tors, might even face higher bait, seems both just ant] commonsensical. In effect, the society reserves the right to be a little more suspicious of those who have been convicted of previous offenses ant] seem to be persisting in committing offenses. So, it is by no means clear that the broacler scope grantee] to prediction once a person has been convicted of an offense 337 on offenders also affects judgments about the appropriateness of using predictions at different stages of the criminal justice system. What the implications are de- penc3s on how one regards the signifi- cance of the actions taken at different stages of the criminal justice system for the offender. If one regards the sentenc- ing decision as relatively insignificant- an anticlimax to the drama ofthe trial and its crucial judgment of guilt or inno- cencc one would grant relatively wide discretion to the use of predictions in sentencing anct much less discretion to those parts of the system that affect the judgment of guilt or innocence at trial. On the other hand, if one regards the sen- tencing decision as very significant be- cause it directly affects the length of time an offender will be imprisoned ant] con- siclers decisions regarding levels of inves- tigation and prosecutorial effort as much less significant because they have, at most, only a minor effect on the possibil- ity of a guilty judgment at trial, one wouIc3 grant much more latitucle for predictions to police and prosecutors than to sentenc-  ing judges. It is not clear which position is correct. Observers may make much of the impor- tance of guilt or innocence, but I suspect offenders are much more concerned about sentence length than the level of police scrutiny and prosecutorial zeal they must endure. Moreover I suspect rules out predictions at earlier stages of ~ ~ ~ ~ ' criminal justice processing. If a person has been convicted of previous offenses recently, and the current case is serious and well supported by evidence, there may be scope in our moral intuitions, law, and current practice to increase levels of investigation, prosecution, and bail guar antees. The principle that the quality of predic- tive tests should be commensurate with the significance of the social harm to be avoided and the sanctions to be imposed , this is particularly true for those who are most likely to be exposed to prediction methods, namely, those who have been convicted of prior offenses. In sum, the application of predictive rules may itself have qualities that en- hance or detract from the overall fairness, justice, or efficacy of the predictive rules. In fact, judiciousness in application might compensate to some degree for defects in the construction, promulgation, or charac- ter of the rule itself. In general, predictive rules are more acceptable if they are used

338 to manage important social problems, if the liabilities contingent on predictive rules are relatively minor, and if the pre- clictions are macle with respect to people who have already been convicted of re- cent criminal offenses. The implications of these principles for use at different stages of the criminal justice system are a little subtle since they turn on judgments about whether convictions for past of- fenses sanction enhanced investigative and prosecutorial attention, and about the relative significance of sentencing cleci- sions versus investigative, prosecutorial, and bait decisions for the interests of accused offenders. These questions de- serve further treatment, but at the outset it should be clear that predictions are not clearly excluclecI at investigative, prosecu- torial, or pretrial stages. PREDICTION AT DIFFERENT STAGES OF THE CRIMINAL JUSTICE PROCESS Discussions of prediction in the crimi- nal justice system have a certain same- ness about them. As cTiscussecI, from a retributive perspective, pre(liction of any sort seems unethical ant] illicit. Yet, there are often practical reasons to make pre- cTictions particularly if they can be macle with clecency and reasonable accuracy. And the explicit use of well-developed prediction methods might well enhance the quality of justice if the practical alter- native is to have biased and impression- istic predictions bootlegger] into the sys- tem by the thousands of criminal justice officials who are doing jobs that seem to require predictions. Besides, it is by no means clear that the best criminal justice system wouIcT be one that honoree] retrib- utive principles to the exclusion of utili- tarian interests in overall crime control effectiveness and in making incremental improvements in criminal justice system operations. CRIMINAL CAREERS AND CAREER CRIMINALS In discussing issues of prediction at different stages of criminal justice system processing, we cannot escape from the general shape of this argument. But the use of prediction at different stages does raise different normative issues-partly because the relevant constitutional and statutory laws are different and partly be- cause current operational practices have accommociated the interest in prediction in different ways. The approach here will be to examine the justice of making pre- dictions of dangerousness at the sentenc- ing stage, in setting bait, and in clevelop- ing prosecutorial strategies, and to do so  from the vantage point of current prac- tices, constitutional law, and moral intui- tions. Sentencing For the past 30 to 40 years, the domi- nant philosophy ant] practice of sentenc- ing has been "rehabilitative sentencing" (Blumstein et al., 1983a:60 61). The aim has been to use the process of sentencing to encourage the rehabilitation of crimi- nal offenders. The legal authority to pur- sue this goal lay in "indeterminate sen- tenc~ng" laws. The principal agents who operates! this system were judges (aider] by probation officers), who set the initial sentence, ant! parole and corrections offi- cials, who decicled whether a person couIc3 be released earlier than the maxi- mum limit on his sentence and, if so, exactly when. Although by no means widely adver- tisecl, this system was built around a core of prediction. When judges sentenced de- fenciants under indeterminate sentencing laws and when parole boards chose to grant or deny inmates' requests for parole or early release, they were implicitly or explicitly making predictions about fu- ture offenses (N. Morris and Miller, 1985:1~12). As an operational matter, that is what it meant to gauge the rehabil , .

PURBLIND JUSTICE itative achievements and potential of a given offender. Moreover, they made these predictions not on the basis of the offenses committee} by the offender (as would be recommended by retributivists) or on the basis of well-developecT preclic- tion methods (as would be recommended by utilitarians), but instead! on the basis of a detailed but essentially discretionary examination of the offender's back- ground, characteristics, and behavior while in prison. Professional opinion among judges and correctional officials supported this form of sentencing. So did legislatures. And so did the Supreme Court. As summarized in Moore et al. (1984:107-108~: In deciding the case of Williams v. New York (1949), for example, the Supreme Court found that criminal sentences should be based "on the fullest information possible concerning the defendant's life and characteristics." Sim- ilarly, in Pennsylvania v. Ashe [302 U.S. S1, 55, 1937], the Court decided that "for the determination of sentences, justice generally requires consideration of more than the partic- ular acts by which the crime was committed and that there be taken into account the cir- cumstances of the offense together with the character and propensities of the offender." LIn understanding the Court's view of sen- tencing,] the Williams case is particularly in- structive. The trial judge overruled a jury recommendation of life imprisonment and im- posed the death penalty on the basis not only ofthe shocking details ofthe crime, which had been revealed, of course, to the jury, but also on the information in the presentence investi- gation. According to the Supreme Court's ac- count, the trial judge "referred to the experi- ence appellant 'had had on thirty other burglaries in and about the same vicinity' where the murder had been committed. The appellant had not been convicted of these burglaries although the judge had information that he had confessed to some and had been identified as the perpetrator of some of the others. The judge also referred to certain ac- tivities of appellant as shown by the probation report that indicated the appellant possessed a 339 'morbid sexuality' and classified him as a 'menace to society'." The Supreme Court up- held the imposition of the death penalty on this basis against a due process challenge. Noting that the "New York statutes emphasize prevalent modern philosophy of penology that the punishment should fit the offender and not merely the crime," the Court reasoned that strict adherence to evidentiary rules limiting the basis for sentencing to testimony given in open court by witnesses subject to cross-exam-  ination would undermine the ability of judges to individualize sentences on the basis of the best available information f United States v. Grayson, 438 U.S. 41, 19781. This sentencing philosophy also tapped an important moral current: the notion that justice must recognize that crimes emerged not simply from evil in- tentions of offenders but also from social circumstances; that there must, therefore, be mitigating facts behind many criminal offenses; and that the best form of justice would be one that tailored social re- s~onses to the guilt of the offender and gave the offender the best chance for rehabilitation. This idea drew on both retributive and utilitarian icleas. The idea that guilt might be mitigated by social circumstances is essentially an idea of justice, since it finds the agency of a crime somewhere outside the minc3 or con- scious will ofthe defendant. The idea that we might clo better to control crimes by rehabilitating offenders rather than sim- ply imprisoning them is essentially a util- itarian idea. For a generation, individual- ized, rehabilitative sentencing was sanctioned by practice, law, and moral aspirations. It had prediction at its core. The dominance of this philosophy was eroded by attacks by both the retributiv- ists and the utilitarians. The retributivists attacking from the left focused on the broad discretion granted to sentencing judges and parole boards, the resulting disparities in sentences for similar of- fenses, and the room left for racial dis- crimination and other forms of unfairness

340 (American Friends Service Committee, 1971; FrankeT, 19731. The retributivists attacking from the right focused on the degradation of both the criminal law and the broact presumption of indiviclual re- sponsibility (van den Haag, 1975; Szasz, 1977~. The utilitarians, attacking from both left and right, focused on the appar- ent ineffectiveness of rehabilitative pro- grams for offenders (Martinson, 1974:22- 54~. Together, these attacks weakened the popular and professional support for the concept of rehabilitative sentencing. The current debate on criminal sentenc- ing seeks an elusive balance of retributive and utilitarian principles. The crucial ques- tions are not about the features of pure systems but on what terms the integration of the systems will proceed. And it is here that retributivists, utilitarians, and "mixec3- system" advocates do battle. The strongest retributive position has been advanced by von Hirsch (1976:77- 88; 1981a:591-634; von Hirsch and Gott- fredson, 1983-1984:341. In his concep- tion, three crucial principles must guide criminal justice sentencing. The first is that the reprehensibility of the criminal offense for which the offender has been convicted is the essential factor that must be recognized in setting criminal penal- ties not the likelihood of committing fu- ture acts or the general deterrent value of punishing the offender (von Hirsch, 1976:6~94; 1981a:5921. The second is that, while exactly how much punishment is deserved by a given criminal offense is somewhat indetermi- nate, it is possible to establish a rank ordering of the reprehensibility of crimi- nal acts ant] that rank ordering must be rigorously preserved in the ordering of punishments meted out (von Hirsch, 1983:21~214, 221-230, 1984a: 10971. Thus, someone convicted of a burglary should never receive a sentence longer than that given to someone else convicted CRIMINAL CAREERS AND CAREER CRIMINALS of a robbery, and a second-(legree burglar shouIcI never receive a punishment more severe than a first-degree burglar. While this principle floes not necessarily deter- mine the size of the bands of punishment surrounding a given offense, the require- ment to preserve ordinal relationships across a great many offenses within the constraints established by the ordinary lengths of human life may, in fact, require that the bands around the offenses be quite narrow. The third principle is that people con- victec] of the same act should receive the same punishment unless some "morally  relevant difference can be established be- tween the offenders" (von Hirsch, 1983: 212-213, 22~2271. The likelihood of committing future crimes wouIcI not be considered morally relevant, although the fact of past crimes might be (von Hirsch, 1976:84~8; 1981a:591~341. Taken together, these principles leave little room for utilitarian interests in gen- eral deterrence or incapacitation or reha- bilitation to come into play. These inter- ests and objectives gain a purchase only within the bands established around of- fenses (which are narrowed by the re- quirement that ordinal relationships be preserved in a limited space of possible punishments), and only insofar as the dif- ferences among offenders may be made "morally relevant" to our judgment of them. A general social interest in reduc- ing crime through (leterrence or rehabil- itation or incapacitation is not sufficient for treating an offender differently. A weaker position in retributive terms, but stronger in utilitarian terms, has been advocated by such scholars as John Monahan (1982: 103-113) and Norval Morris (1974:7~77, 1982: 179-209; N. Morris and Miller, 1985~. Monahan (1982) has called this position a "modified just deserts" position. In this conception the outer limits of punishment for given of

PURBLIND JUSTICE Lenses are once again established by re- tributive concerns. On the question of how broad the range of punishments sur- rounding a given offense could be with- out doing injustice to the offender, the authors are silent. Moreover, there is no strict principle requiring the preservation of ordinal relationships to establish a sense that the bands must be tight. So. it seems that a --just punishment in this conception may be broader than in the von Hirsch conception. Similarly, there is no rigorous statement that similar cases (definer] in terms of offenses) must be treated alike. So, there is a great deal of room for utilitarian concerns to come into play. Obviously, if the bands around offenses are sufficiently wide, and if there is no rigorous principle requiring that similar offenses be treated similarly, utilitarian concerns could determine everything within the hollow shell of retributive principles. And it is this that focuses von Hirsch's criticisms (1981b:772-789, esp. 784 - 785~. On the other hand, the mocli- fied just deserts position has the virtue of allowing the criminal justice system to fit criminal liability to the varied forms of human conduct and misery that appear in the system and to the limited capacities of the system to punish (N. Morris, 1982: 190), and to do so in a way that preserves some of the society s interest in a valu- able and useful criminal justice system, as well as a just one. The strongest utilitarian position has been adopted by Peter Greenwood (19821. His argument is that the society has an interest in both minimizing crime and reducing its reliance on prisons. In a world in which rehabilitation seems to have failed, the efficacy of general deter- rence remains uncertain, and general in- capacitation costs too much in terms of liberty and money per unit of crime re- duction achievecl, it is valuable to focus . . ~ . . 341 scarce prison capacity on those who are likely to commit the most crimes. This is particularly true when it seems that the differences among offenders in terms of the seriousness and rate of offending are quite substantial, and when some capac- ity exists to distinguish the high-rate, se- rious offenders from the lower-rate, less serious offenders. If the opportunity cre- ated by this situation was exploited, the society could have both less crime and fewer people in prison than it now has (N. Morris, 1974:63; N. Morris ant] Miller,  1985:6; Wilson, 1983b: 15~156; von Hirsch and GottEredson, 1983-1984: 22~31, 44 45; Moore et al., 1984:79 891. Obviously, these different positions balance retributive and utilitarian con- cerns in quite different ways. In particu- lar, they come to radically different con- clusions about how great a role the society should grant to predictions of in- cliviclual conduct in imposing criminal sentences and about how just distinctions among people convicted! of similar of- fenses might be made. But this brief ac- count of the history of sentencing policy inclicates that there must be room for predictions in our normative conception of sentencing. It has been, and is now, sanctioned by current practice and by statutory and constitutional law. More- over, all but the most stringent retributiv- ists would accept predictions based on some characteristics of offenders as part of sentencing policies. There may well be limits on the use of predictions with re- spect to the magnitudes of the sentence increases that could be meted out and the sorts of variables that could be includecI. It might also be important to establish procedural devices to ensure that the characteristics of offenders relevant to sentencing were accurately assessed. But it seems strange to insist that there is no room for predictions of future criminality in sentencing.

342 Bail and Pretrial Decisions* To many, the notion of jailing someone not yet convicted of a crime on the basis of uncertain judgments about the danger he presents to the community seems an- tithetical to our most fundamental legal traditions. And although pretrial deten- tion is not the explicit goal of guidelines that increase bait for offenders estimated to be dangerous, that is the frequent and unlamented result. Two objections to both preventive de- tention and risk-adjusted bait are com- monly voiced. It is wrong to jail- and therefore punish people who have not been convicted of crimes. And it is partic- ularly unjust to detain them on the basis of predictions about future crimes. Stated affirmatively rather than negatively, the argument is that the state's only proper interest is to guarantee that accused indi- viduals appear for trial. The amount of bail should be determined with this pur- pose in mind, and bail can hardly ever be denied on that basis. It is especially inap- propriate to detain people solely to pro- mote community security. While compelling in principle, this po- sition is undercut by three observations. First, the actual operations of the existing system reveal the bankruptcy of the guid- ing principles. The defendants who are detained are not those whose appearance at trial is of greatest concern to the state, but instead those whose financial re- sources are most limited. Some critics urge the release of more clefendants on their own recognizance; others propose substitution of community sureties for money bail on the grounds that these would be more equally available to all *This section, in its entirety, originally appeared in Mark H. Moore, Susan R. Estrich, Daniel McGil- lis, and William Spelman, 1984, Dangerous Of- fenders: The Elusive Target ofJustice, pp. 122~125. (D 1984, Harvard University Press, Cambridge, Mass. Reprinted with permission. CRIMINAL CAREERS AND CAREER CRIMINALS defendants (Freed, 1982). Such reforms might well lead to less pretrial detention without harming the state's interest in guaranteeing appearance at trial. But the most important implication of the present system is that we are apparently willing to detain people without a finding of guilt simply to guarantee their appearance at trial. If the right to be free before trial can be overwhelmed by the state's limited interest in guaranteeing future appear- ance, then the right cannot be so funda- mental, and it occasionally might be over- whelmed by the state's interest in  reducing crime as well. Second, many deny that the state's in- terest is limited to guaranteeing appear- ance at trial. Some legal scholars have argued that bad! and sureties were also designed to promote community security (GoldLamp, 1979:15~1; Freed, 19821. And as a practical matter, both citizens and judges clearly think it is not only appropriate but crucially important that the citizens' interests in security be re- flected in pretrial detention decisions. Finally, the Supreme Court has so far refused to establish an unlimited right to bail, nor has it been willing to limit the state's interest to guaranteeing the defen- dant's appearance at trial. True, the Court has not yet heard a case on the constitu- tionaTity of preventive detention because all such cases have become moot before the Court could take them up `~`Pretrial Reporter 6 (March 1982) :131. And in the leading bait case, Stack v. Boyle, the Supreme Court did indicate that guaran- teeing appearance should be the most important factor `(Stack v. Boyle, 342 U.S. 1 (19511:51. But the constitutional right of an individual to be set free on bait based solely or primarily on the need to guaran- tee appearance at trial has not been estab- lished. To many, the Court's reluctance in this area seems inexplicable, for the constitu- tional language seems clear and straight

PURBLIND JUSTICE forward. The Eighth Amendment to the U.S. Constitution asserts flatly that "ex- cessive bad] shall not be required." Unfor- tunately, this simple assertion can be given at least three interpretations (GolUkamp, 1979: 16-17~. One is that de- fenciants have a right to reasonable bait, and the Supreme Court will determine what is reasonable. That interpretation, which would establish a right to be re- leased on reasonable bail, has been sup- ported by a historical analysis of bad] in England (Foote, 1965:959-999; Fabric- ant, 1968: 303-315~ . A s e cond interpreta- tion restricts the amount of bait to reason- able levels but leaves it to the states to pass laws indicating what is reasonable. No conception of a constitutional right is envisioned in this interpretation. Indeed, the states could clecide that it was reason- able in some cases to deny bail. A third interpretation is that"in the absence of constitutional or statutory discretion . . . judicial discretion determines the ap- propriateness of bait within the bouncis that it shouIc! not be excessive" (Gol~kamp, 1979~. This also rejects the notion of any right to bail, but it allows judges to set bait when statutes do not explicitly authorize it. The District of Columbia enacted a preventive detention statute in 1970 that explicitly allowed offenders who were preclicted to be dangerous to be detained. The constitutionality of the statute was tested in United States v. Edwarcis (No. 80-294 (D.C. App. May 8, 1981), cert. cleniecI, 22 March 1982~. The District of Columbia Court of Appeals held that the statute was constitutional, narrowly re- jecting the interpretation that the Eighth Amendment guarantees a right to bail. The court reviewed the origins of the excessive bait clause and the case law pertaining to it and concluded that the aim of the Eighth Amendment was not to limit the power of Congress to deny pre- trial release for specifier] classes of of 343 fenders or offenses, but rather to limit the discretion of the judiciary in bait setting. The court also ruled that the Fifth Amenc3ment's clue process clause was not violated by the preventive detention stat- ute. Opponents ofthe statute objected on grounds that it permitted punishment of the defendant prior to full adjudication of the case. The court concluded that pre- trial detention is not a form of punish- ment but rather a regulatory action and hence permissible. The case was appealec! to the Supreme Court, but the Court declined to consider  it, perhaps for reasons similar to those justifying its reluctance to consider a pre- vious Nebraska case, Murphy v. Hunt (No. 8(~2165, 30 Cr L 3075, 1982; Parker v. Roth, 278 NW 2d 106, 19791. That case involved the constitutionality of Nebras- ka's constitutional amendment requiring "the denial of bail to defendants charge c3 with forcible sex offenses when the proof is evident or the presumption great" (Murphy v. Hunt). The U.S. Court of Appeals for the Eighth Circuit found the amendment to be an unconstitutional re- striction on the right to bail and asserted that "the constitutional protections in- volved in the grant of pretrial release by bail are too fundamental to foreclose by arbitrary state decree" (Hunt v. Roth, 648 F. Al 1148, 19811. The Supreme Court vacated the Eighth Circuit's decision and found that the case was moot because the clefenclant hac3 already been convicted for rape and sentenced to prison (`Murphy v. Hunt). The Eclwards case might also have been viewed by the court as not present- ing a "live" issue because Ec~warcls en- tered guilty pleas in both cases in which preventive detention was sought. Such a ruling poses an interesting dilemma since "pretrial detention orders will almost surely not outlive the appellate process" (Pretrial Reporter 6 ( March 19821: 13~. The Court could choose to treat a future case as an exception embodying the prin

344 ciple of being "capable of repetition, yet evading review," and this rule was em- ployed by the District of Columbia Court of Appeals in its review of the case (United States v. Edwarcts). As in the case of sentencing, current practice and constitutional law both seem to sanction bail decisions (including a decision to deny bail altogether) based on predictions of dangerousness. This does not necessarily establish any amrmanve reason for doing this, however, and it does seem contrary to our most important legal traditions. One justification for pre- diction is the community interest in con- trolling crimes committed by people on bail. But by most estimates, the practical effect will be small, and there are other ways of controlling crime on bail, such as special penalties or going to trial sooner (Lazar Institute, 1981:481. CRIMINAL CAREERS AND CAREER CRIMINALS Prosecution To some, the prosecutor seems the most powerful criminal justice official partly because his or her decisions are consequential for defendants but even more importantly because the prosecutor has broad discretion to make the choices (Vorenberg, 1981:1521-1573~. The prose cutor can quash charges, make a deal to trade information for a forgone prosecu tion, threaten a defendant with serious charges, and determine when a case will go to trial. Moreover, these choices are neither guided by explicit policies nor commonly reviewed. Despite the wide discretion, profes sional norms and community pressures lend some consistency to prosecutorial decision making. Generally, prosecutors decide how much effort to apply to indi Perhaps the most important reason to vidua1 cases according to the seriousness use prediction in making pretrial deci- ofthe current offense and the strength of signs is not to reduce crime on bail but to limit and rationalize the current system (Moore et al., 1984:125~. Just as judicious use of prediction in sentencing convicted offenders could lead to fewer people be ing imprisoned, pretrial detention of dan gerous offenders might lead to fewer peo ple being detained and to the use of explicit criteria that would be fairer. A system that detained only those few who represented great risks of flight or new crimes, regardless of their financial re sources, would be a welcomer! relief, even if it required making explicit deci sions about who was to be detainee! and who released. Compared with the current system, the only loss to justice would be in the explicit recognition of a community interest in controlling crime committed  on bail, a principle that already seems to have some political and legal vitality cle spite the controversy over whether it is constitutionally recognized. the evidence: serious cases with strong evidence attract a great clear of prosecuto- rial attention; minor cases with weak ev- idence are screened out early or dis- patched to overworked sections of the office that cannot give them anything but negligible attention (Institute for Law and Social Research, 1976a,b, 1977~. The focus on offenses and the strength of the evidence in the case can be under- stood from both retributive and utilitarian perspectives. It makes sense to retributiv- ists because it ensures that prosecutorial attention will be focused on those who are likely to have committed serious crim- inal acts and because it imposes less lia- bility on those whose acts are less serious or whose guilt is less likely. It makes sense to utilitarians since it seems to en- sure that scarce resources will be spent where they will do the most good: in punishing those who seem to cause the worst part of the crime problem.

PURBLIND JUSTICE Although the offense-based focus of prosecutors remains the dominant princi- ple in guiding prosecutorial discretion, in the last clecacle prosecutors have experi- mented with a new principle that would give priority attention to "career crimi- nals" or "major offenders'7 (Harper and McGillis 1977; Moore et al. 1984:1371. , In effect, in deciding how cleterminec3ly to pursue a case, prosecutors have de- cided to consider characteristics of the offender as well as the offenses and the strength of the evidence. The characteris- tics that qualify an offender for special treatment include a history of serious, repetitive, and persistent criminal con- duct although there are important dif- ferences among prosecutors' offices with respect to the relative weights given to the different characteristics or criminal history (Harper and McGillis, 1977; With- comb, 1980; Rhodes et al., 1982~. Some officials think a few serious crimes- even if widely separated in time would qualify an offender for special attention; others pay much closer attention to the rate and persistence of criminality and worry less about the seriousness of the offense. The special treatment to which offenders are exposed includes special efforts to gather, preserve, and protect evidence in the case; charges filed at the highest possible level sustainable at trial; restrictions on plea bargaining; ant! prompt trials. The aim is to increase the likelihood that those with a serious record will be convicted and to extend sentences for those who are convicted. This change in prosecutorial proce- dures can also be understood in both retributive and utilitarian terms. The re- kibutive justification is the same as that for habitual offender sentencing laws: that offenders with long records have shown themselves to be unusually unrepentant and careless of society's val 345 ues and, therefore, unusually (reserving of punishment. The utilitarian justifica- tion is that offenders who have committed crimes repeatedly in the past are particu- larly likely to commit crimes in the fu- ture, anti, therefore, it is particularly valu- able to focus scarce prosecutorial time on ensuring that these unusually dangerous offenders will be punished and incapaci- tated. Obviously, this focus on criminal record and characteristics of the offenders is related to the question of prediction. To the extent that a utilitarian logic motivates the shift from the focus on current of-  fenses to past offenses an(1 to the extent that past offenses predict future conduct well, one can argue that prediction has crept into prosecutorial decision making and is therefore sanctioned by current practice. Nonetheless, it would probably be more accurate to say that predictions of dangerousness have not yet been as sys- tematically or as explicitly introducer! into prosecutorial decision making as they have into sentencing decisions or even bait decisions. So, explicit use of predictions of dangerousness is not yet sanctioned by current prosecutorial prac- tice. The important ethical questions are whether such methods would be consti- tutional and consistent with moral intui- tions about the criminal justice system. At the outset, the idea of selective pros- ecutions focused on those predicte(1 to be (langerous seems to threaten the princi- ples of equal protection and due process. Indeed, it seems even more threatening if dangerous offenders are prosecuted more determinedly for relatively minor of- fenses or for charges in which the evi- clence is relatively weak (Moore et al., 1984:141-142~. As noted above, if prose- cutors organized an overwhelming on- sTaught against a dangerous offender charged with a serious crime, or if they

346 kept prosecuting dangerous offenders for vagrancy or disorderly conduct, or if they kept bringing robbery cases on the basis of trumpecl-up evidence, they wouIc3 have crossed an important line that makes our system of justice fair and restrained. The interesting question, however, is not at the extremes but in the misfile range. ShouIc3 prosecutors give a slightly more vigorous and determined prosecu- tion to cases involving dangerous o~er~cI- ers? A vigorous prosecution conic] mean enhanced effort in cases of serious crime in which the evidence was very strong refusing to accept plea bargains, conduct- ing extensive collateral investigations, or moving very quickly in a case in which there were strong physical evidence and eyewitnesses. It couIcI also mean a greater willingness to prosecute less seri- ous offenses where the evidence was strong-for example, holding out for a felony conviction in a case of gun posses- sion when the testimony of two police officers is corroborated by a witness. Or, it could mean being willing to risk failure in prosecuting a serious crime in which the evidence was well above the constitu- tional standard but much less than the usual prosecutorial stancIarc3 of 90 percent certainty to win at trial for example, a robbery case in which there is no physical evidence ant] the eyewitness testimony is shaky. It is in these areas that a selective focus among prosecutors would operate, and it is the justice of these actions that must be consiclerect. As a constitutional matter, it seems fairly clear that prosecutors c30 have the leeway to establish principles for acljust- ing levels of prosecutorial effort among offenders as Tong as the principle serves some legitimate social purpose, and as Tong as the policies are not based on an unjustifiable standard (such as race, reli- gion, or social classy, the motives of the prosecutor are not vindictive, and the policies are not designed to frustrate cle CRIMINAL CAREERS AND CAREER CRIMINALS fondants in their exercise of constitu- tional rights, such as freedom of speech, assembly, and religion (CarclinaTe and Feldman, 1978:659-692; Vorenberg, 1981~. While there have been a few cases in which the mere exercise of discretion was found objectionable on equal protec- tion grounds (Village of Fairlawn v. Fuller, 8 Ohio Misc. 266, 221 N.E. 2d 851), the dominant court opinion has been that it was not sufficient for a de- fendant to show that offenders escapee! punishment tOyler v. Boyles, 368 U.S. (19621; Washington v. Unitecl States, 401  F.2d 915, 925 (D.C. Cir. 1968~. On the other hand, where prosecutors seem to have been motivated by arbitrary, racially tainted standards, or where they seem to have been guided by vindictiveness, the courts have found constitutional viola- tions Wick Wo v. Hopkins, 118 U.S. 356 (1886~; People v. Utica Daw's Drug Co., 16 A.D.2d 12, 225 N.Y.S.2d 128 (1962); Unitecl States v. Berrios, 501 F.2d 1207 (2d Cir. 197411. But in showing discrimi- natory enforcements the courts have gen- erally placed the burden on defendants (People v. Utica Daw's Drug Co.~. Such cases do not affimnatively establish a li- cense for prosecutors to vary levels of effort according to predictions of future criminality. But\ to the extent that such predictions were accepted by the courts as a legitimate law enforcement purpose and they were formulated in a way that avoided any taint of arbitrariness or racial bias, the courts would probably accept the policies as within the range of prosecutorial discretion. Indeed, what makes the conclusion seem particularly justifiable is not so much that the court countenances predictions as that the court has been extremely reluctant to ex- ercise any control over prosecutorial dis- cretion at all. As fudge (now Chief [us- tice) Burger wrote in Newman v. United States, 382 F.2d 429, 480 (D.C. Cir. 1967~: "Few subjects are less adapted to

PURBLIND JUSTICE juclicial review than the exercise by the Executor of his discretion in clecicTing when anc3 whether to institute criminal . ,, proceec lngs.... If establishec! legal principles are not a bar, what about moral intuitions? Here, one is once again plunger! into the gen- eral discussion of cleontological objec- tions to anc! utilitarian justifications for prediction. The only difference is that here we are talking about "special" prosecutorial efforts, anc] much turns on what is meant by "special." If what is meant is nothing more than special efforts to collect anc! preserve evidence anc] to proceec] quickly to trial, surely there is little objection. Although an interest in "fairness" among clefenclers might be vi- olatecI, one can reasonably argue that of- fenclers clo not have a constitutional or even moral right to the ordinary, sloppy prosecution they receive in tociay's overburclenec! criminal justice system. AncI, since no clue process issue is raised, this kinc! of"special treatment" seems acceptable. Somewhat more worrisome are those concerns relater! to clue process: that cle- fenciants might be overawes! by zealous prosecutors, that the trial process might be contaminated if it was known that a clefenciant was one of those preclictec] to be clangorous, anc] that the balance be- tween the resources available to defense and prosecution might be upset. All of these are important because they affect the substantive findings of guilt or inno- cence ant] do so in a way that violates the defendant's rights to due process and the community's interest in being sure that justice is being done. There are answers to these concerns. Special procedures could be developed to make sure that judges and juries were unaware of the special status of the of- fenclers to avoid the informal introduction of prejudicial information at trial. Special resources couIct be made available to the 347 defense as well as the prosecutor in cases involving those prectictecI to be danger- ous. And by developing prediction crite- ria basecl on an extensive criminal record, we couIcT guarantee that the defendants who were vulnerable to the special pros- ecution were relatively experienced of- fen(lers who would not easily be fright- enec3 by a prosecutor's bluffs and threats. But none of these answers is wholly sat- isfactory. As is generally the case, the decision comes down to a balance among the com-  munity's interest in security, the defen- dant's interest in avoiding criminal lia- bility, and a broad social interest in guar- anteeing certain standards of justice. In striking the balance, many see special prosecutions as particularly threatening to standards of justice since they may have a decisive effect on the question of guilt or innocence. Hence, they judge the defendant's rights and interests to weigh more heavily in this regard than in sen- tencing decisions. And this would clearly be true if a defendant was being prose- cuted for the first time. But the more interesting question is whether special prosecution would be inappropriate when a defendant has already been con- victed of several offenses and when he is predicted to be dangerous. Arguably, this is more acceptable because it makes it less likely that the defendant will be overawed and may, in any event, dimin- ish the defendant's rights in the same way that they seem to be diminished in deter- mining sentences. PREDICTION AND BLAMEWORTHINESS After one has been through the intel- lectual contortions of evaluating predic- tion-based criminal justice decisions from retributive and utilitarian perspectives, and, as a general idea and in particular applications, one longs for a simpler

348 view. The very complexity of the analysis weakens its credibility. In my view this complexity is unneces- sary. It is forced on us by a recently found sophistication in reasoning about this is- sue. Incleed, the sharp distinction drawn between retributive and utilitarian posi- tions that is the cornerstone of much con- temporary analysis obscures a far simpler and more coherent view. This simpler view depends on seeing what is common to retributive ant! utilitarian views rather than what is different. The iclea that emerges is unfamiliar and unconven- tional in today's debates, but I think it might be treated as commonplace and obvious in a world in which the current distinctions were less firmly ant! sharply drawn. The contemporary view of retributive theories is that they properly focus the attention of the criminal justice system on current acts rather than the character of offenders. It is the criminal act that pro- vides the justification for punishment. The more serious the act, the more seri- ous the punishment. There is much to commencI this posi- tion. It connects to more primitive icleas of justice as vengeance without being hostage to the excessive passions and penalties that might characterize private vengeance. The offense is against the community and the state-not a private individual. The response is regulated by concerns for equal protection and due process-not the strength of the victim's comrades. It also turns out to be a position that limits the state's interest and surveil- lance to narrow areas marked by actual criminal offenses (Moore, 1983:17~21. This not only protects much of social life from government scrutiny but also guar- antees that, when the state's interest is engaged, it is focused on an area in which it can do some good rather than mischief. And the focus on acts prevents the society from developing any permanent view of CRIMINAL CAREERS AND CAREER CRIMINALS the character and status of criminal of- fenclers (von Hirsch, 1981a:599~. All this seems to strike a nice balance between the community's interests in simulta- neously engaging state power to protect a limited number of community values and preventing the state itself from becoming too powerful and intrusive. Attractive as the focus on acts seems to be, however, it produces some curious anomalies when used to explain our cur- rent criminal laws. The most glaring is the importance that the criminal law at- taches to the mental state of the offender  at the time he committed the offense. If the act itself is so important to criminal punishment, one might expect many criminal statutes to establish strict liabil- ity for criminal offenses. In fact, however, strict liability is very rare in criminal statutes (Packer, 1968:121-1311. It is gen- erally important that some demonstration be made that the offender willed an act as well as that the act occurred. Similarly, there are many climinishect-competence defenses and statuses (including mental illness, compulsion, and youth) that miti- gate blameworthiness by casting doubt on whether the offender was in fact the author of the act in the sense that the outcome ofthe act was a complete expres- sion of what the offender wanted. Finally, under some circumstances (defined in the law of"entrapment"), government com- plicity in a crime can absolve an offender. Thus, anything that drives a wedge be- tween a criminal act and the intention of the offender tends to mitigate guilt be- cause it confuses our capacity to infer criminal intentions from criminal acts. So, the act alone is not sufficient for criminal responsibility. The intention to do the crime- to cleny the values of the soci- ety must be shown, as well as the act. What is even more surprising is that a harmful act is not even necessary for criminal responsibility. Laws that make attempts or conspiracies to commit

PURBLIND JUSTICE crimes vulnerable to criminal prosecution essentially make a durable, visible inten- tion to do a crime worthy of punishment even if the substantive offense never oc- curs (Packer, 1968:100-1011. True, these laws typically carry less severe penalties than the completed offenses would jus- tify. And true, some overt acts are neces- sary to trigger the investigation and pro- vide proof of a durable criminal intention. But the point is that the acts are important not in themselves but only as they afford insight into the intentions of offenders, and it is the intentions alone that justify punishment. So, even though we are accustomed to thinking of acts as the most essential fo- cus of the criminal justice system, a harm- fuT act is neither sufficient nor even nec- essary for findings of"blameworthiness." Intention, on the other hand, which seems less essential, is not only necessary for criminal responsibility, but sufficient itself! One possible implication of these observations is that it is criminal inten- tion- the willing rejection of society's values, including that obligation to re- spect the life, liberty, and property of others-that justifies the punishment. The act is important not only in itself but also and most fundamentally as an objec- tive piece of evidence about the inten- tions, values, and character of citizens. If this interpretation were accepted, it would also help to explain why most peopIc- inclucling many retributivists- believe that it is appropriate to adjust the severity of criminal justice sanctions in response to prior criminal acts as well as to the seriousness of current criminal of- fenses. This is true regardless of whether the criminal justice sanction in question involves sentencing and is established through statutes (as in habitual offender statutes) or involves prosecution and is establishecI by administrative fiat (as in the establishment of "career criminal united. This position is problematic, 349 however, to a strict retributive position that ties criminal liability only to acts. The inconsistency can be resolved in three ways. One is to point out that the series of offenses indicates that an of- fender is unusually resistant to learning from punishment and therefore more punishment is called for. This may make sense, but it is a utilitarian rather than a retributive argument. A second argument is that the fourth robbery is somehow worse than the first and therefore is more deserving of punishment. But that is sim- ply an assertion. The obvious question  that is unanswered is exactly what makes the fourth robbery worse. A third argument, which seems more satisfactory, is that the criminal law ad- justs penalties for offenses on the basis of what can be discerned about intention and character and that a series of offenses reveals an offender as clearly more will- ing to commit crimes than others and, therefore, as more deserving of punish- ment. We all understancl that criminal offenses can be caused by circumstance and transient passion as well as by clear intention. When we Took at first offenders, it is quite possible that their values and character their commitment to the society's values are much like every- body else's and that they were simply unlucky enough to stumble into a situa- tion that proclucec3 an uncharacteristic offense. When we look at someone who has committee] many offenses, however, the hypothesis that the offender is much like everyone else in terms of his values must yielcl to the alternative hy- pothesis that the values are different: the offender is less solicitous of and more willing to attack the lives, liberty, and property of fellow citizens. It is this in- creasecl certainty about the offencler's values that justifies enhanced punish ment. So, there is a certain coherence in thinking of retributive conceptions of jus

3so lice as being concerned about the inten- tions, values, and character of offenders as well as their acts. This is important, for if intentions and character are durable (i.e., if people have guiding values that last for at least a little while), past actions of offenders might well predict future ac- tions. Consequently, a policy that sanc- tioned extra punishment for past repeated criminal acts would produce about the same results as a policy that adjusted penalties on the basis of predictions of future criminal acts. Thus, retributive and utilitarian justifications coalesce in a fo- cus on those who have revealed their intentions ant! capacity to commit crimi- nal acts through a pattern of past offenses. What ties these principles together is the argument that character relatively clura- ble values ant! intentions is fundamen- tal to both retributive and utilitarian jus- tifications for punishment. Note that to accept the idea that char- acter is durable and funciamental to both retributive and utilitarian justifications for punishment is not to accept the idea that it is permanent. People's values and in- tentions can change. Even the most cyn- ical might excuse offenders who had aged ant! matured before their just penalties were served and be reluctant to exact the maximum penalties from those 20-year- olds who committee! many offenses, on the grounds that such offenders might change. So, we need not decicle that char- acter is permanent to decide that it is somewhat durable and relevant to crimi- nal justice decisions. If this interest in character provides the basis for a synthesis of retributive and utilitarian principles, why is it an unfa- miTiar idea? My answer is that this idea runs counter to a dominant icleology guid- ing criminal justice policy. Central to that ideology is the idea that moraTism must be kept out of the criminal law because the passions that would be released if it were invited in are uncontrollable (Gil CRIMINAL CAREERS AND CAREER CRIMINALS lers, 1983:4021. The focus on the inten- tions and values of offenders incleec] the argument that it is wrong values as re- veaTed by acts that justify punishment puts values at the center of the criminal justice system and thus runs directly counter to the dominant icleology. Per- haps equally important, we have been guided by a hopeful view of human na- ture: human character is transient, changeable, and influenceable; guilt for current offenses is therefore always miti- gated; and bright hopes for rehabilitation are reasonable. The focus on durable  character treats the role of outside influ- ence as morally irrelevant and is less optimistic about the rate at which impor- tant changes in values can occur. So, the focus on character flies in the face of icleologies that have been central to our contemporary jurispruclence. Obviously, no one is interested in un- leashing a new age of moral oppression. We value our freedom, our mobility, our ability to experiment with different val- ues far too highly for this. But it floes seem valuable to remind ourselves of some simple principles we seem to have forgotten: that the criminal law is a moral statement about the values that bind our society together by imposing minimal ob- ligations on one another; that the society insists that people honor those laws and the values that lie behind them; and that, when a person clearly shows an incliffer- ence to those obligations through his or her actions, the society has a right to respond with indignation moderated by concerns for clue process and equal pro- tection. This set of principles sanctions an interest in character in those who have committed offenses in the past and will (lo so in the future. At the same time, it limits the reach of the system to those who have committed offenses in the past. It does not try to reach for extra state control through improved techniques of prediction that provide less satisfactory

PURBLIND JUSTICE ways of exploring character than prior criminal conduct. SUMMARY AND CONCLUSIONS Our sharer! vision of the worIc3 of crim- inal offenses and criminal justice policy has become a great clear more compli- cated than it once was. We now think of criminal offenses as the result of acci- clents and transient passions as well as considered intentions. We think of crim- inal justice decision making as cliscretion- ary and relying on professional knowI- edge and expertise rawer than automatic application of well-established princi- ples. Simple notions of justice that com- bined concern for justice with ordinary prudence have become elaborate, sharply differentiated ethical theories emphasiz- ing retributive or utilitarian aims of crim- inal justice policy. So, it is hard to fins! the thread of decency and justice in propos . . 1- - ~ng cr~m~na Justice policies. Into this tangled and overburdened world come proposals to make wider use of improved prediction techniques in tar- geting offenders for investigation and prosecution, in setting bail, and in impos- ing sentences. The appeal of such tech- niques comes from their apparent poten- tial to produce greater community security from the financially (and morally) limited capacities of the state to punish, and to impose some rational order on what is otherwise a crazy-quiTt pattern of discretionary decision making that leaves great room for injustice. But there are problems with the iclea of relying on predictive tests. To retributiv- ists, it seems wrong to impose criminal liabilities on the basis of predictions of further criminal acts. To many others, it seems wrong to impose liabilities on peo- ple who are falsely predicted to commit crimes in the future. Still others worry about the characteristics that will be used in the predictive tests, thinking that it 357 would be wrong to use characteristics that were not under the control ofthe offender and were not themselves criminal in nature. And there are always the ques- tions of exactly at what point in the crimi- nal justice process the tests would be applied and what consequences the use of the tests would have for criminal of- fenders. One can wrestle with these questions at many levels. It seems to me, however, that the easiest way through this tangle is to be guides] by two principles: First, the best guide to both blameworthiness and  future criminal conduct is prior criminal offenses. Second, it is a virtue to be eco- nomical in the use ofthe state's moral and financial capacity to punish and control. If accepted, these principles would have the following implications: · That predictive or discriminating tests should be clesignecT to identify a small and distinctive element of the of- fending population. · That the tests should be based pre- dominantly on prior criminal conduct. · That no one should be identified as, or predicted to be, dangerous who does not have repeated adult criminal convic- tions on his or her record. · That juvenile records of serious of- fenses couch be used for purposes of dis- cerning dangerousness or predicting fu- ture crimes if a person committed serious offenses soon after graduating from the juvenile justice system. · That the use of information on indict- ments and arrests in addition to convic- tions can be used in the tests and is probably to be preferred to the use of employment or marital data. · That the required accuracy of the tests shouIcl be consistent with the size of the practical benefits of the test and with the size ofthe burdens imposed on clefen- ciants. · That the tests couIcl be used not only

352 for sentencing, but also for targeting in- vestigations and prosecutions. · That the additional liability at sen- tencing should be limiter] by Me serious- ness of the offense for which We person was convicted. · That the additional liability at inves- tigation and prosecution stages be expo- sure to more vigorous investigation and prosecution but within clue process pro- tections. · That Me principal justification for us- ing improved prediction techniques at Me ball stage would be to reduce Me use of pretrial detention, guarantee Mat de- tention is focused on Me most dangerous offenders, and rationalize the current cha- otic system. · That Me tests be thought of less as prediction techniques and more as a way of focusing attention on those offenders who have revealed tendencies to be un- usually dangerous through their past acts. These proposals may have the effect of dampening some of the technocratic en- ~usiasm for prediction. But in my view Mat is their virtue rawer than their vice. REFERENCES AND BIBLIOGRAPHY American Friends Service Committee 1971 StruggleforJustice: A Report on Crime and Punishment in America. New York: Hill and Wang. Andreano, Ralph, and Siegfried, John J., eds. 1980 The Economics of Crime. New York: John Wiley & Sons. Attorney General's Task Force on Violent Crime 1981 Final Report. Washington, D.C.: U.S. De- partment of Justice. Bedau, H. 1977 Concessions to retribution in punishment. In J. Cederblom and W. Blizek, eds.,Justice and Punishment. Cambridge, Mass.: Ballinger. Blumstein, Alfred, Cohen, Jacqueline, and Nagin, Daniel, eds. 1978 Deterrence and Incapacitation: Estimating the Effects of Criminal Sanctions on Crime Rates. Washington, D.C.: National Academy of Sciences. CRIMINAL CAREERS AND CAREER CRIMINALS Blumstein, Alfred, Cohen, Jacqueline, Martin, Susan E., and Tonry, Michael, eds. 1983a Research on Sentencing: The Searchfor Re- form. Vol. I. Washington, D.C.: National Academy Press. 1983b Research on Sentencing: The Search for Re- form. Vol. II. Washington, D.C.: National Academy Press. Boland, Barbara 1983 Identifying serious offenders. In Mark [I. Moore, Susan Estrich, Daniel McGillis, and  William Spelman, eds. Dealing With Dan- gerous Offenders, Vol. 11: Selected Papers. John F. Kennedy School of Government. Cambridge, Mass.: Harvard University. Boland, Barbara, and Wilson, James Q. 1978 Age, crime and punishment. Public Interest 51:22-35. Brown, Peter 1981 Assessing officials. In Joel L. Fleischman, Public Duties. Cambridge, Mass.: Harvard University Press. Caplan, Gerald M., ed. 1983 ABSCAM Ethics: Moral Issues and Decep- tion in Law Enforcement. Washington, D.C.: The Police Foundation. Cardinale, Philip J., and Feldman, Steven 1978 The federal courts and the right of nondis- criminatory administration of the criminal law: a critical view. Syracuse Law Review 29:659 692. Carrington, Frank 1983 Crime and Justice: A Conservative Strategy. Washington, D.C.: The Heritage Foundation. Cederblom, J., and Blizek, W., eds. 1977 Justice and Punishment. Cambridge, Mass.: Ballinger. Chelimsky, Eleanor, and Dahmann, Judith 1981 Career Criminal Program: National Evalua- tion: Final Report. Washington, D.C.; U.S. Government Printing Office. Collins, James, J. 1978 Offender Career and Restraints: Probabili- ties and Policy Implications. Philadelphia, Pa.: University of Philadelphia Press. Curtis, Lynn A., ed. 1985 American Violence and Public Policy. New Haven, Conn.: Yale University Press. Dawson, Robert O. 1969 Sentencing. Boston, Mass.: Little, Brown. Dershowitz, Alan 1973 Preventive confinement: a suggested frame- work for constitutional analysis. Texas Law Review 51:1277-1324. 1974 The origins of preventive confinement in Anglo American law. University of Cincin- nati Law Review 43: 1~0, 781-846.

PURBLIND JUSTICE 1976 Background paper. In Twentieth Century Fund Task Force on Criminal Sentencing. Fair and Certain Punishment. New York: McGraw-Hill. Eck, John E. 1983 Investigative strategies for identifying dan- gerous repeat offenders. In Mark H. Moore, Susan Estrich, Daniel McGillis, and William Spelman, eds. Dealing with Dangerous Of- fenders, Vol. II: Selected Papers. John F. Kennedy School of Government. Carn- bridge, Mass.: Harvard University. Fabricant, Neil 1968 Bail as preferred freedom and the failures of New York's revision. Buffalo Law Review 18:303 ;315. Fleischman, Joel L., ed. 1981 Public Duties. Cambridge, Mass.: Harvard University Press. Fletcher, George 1978 Rethinking Criminal Law. Boston, Mass.: Little, Brown. Floud, Jean, and Young, Warren 1981 Dangerousness and Criminal Justice. Lon- don, England: Heinemann. Foote, Caleb 1965 The coming constitutional crisis in bail: I. University of Pennsylvania Law Review 113:95~999. Forst, Brian, and Brosi, Kathleen B. 1977 A theoretical and empirical analysis of the prosecutor. Journal of Legal Studies 6:177-191. Frankel, Marvin E. 1973 Criminal Sentences: Law Without Order. New York: Hill and Wang. Frankena, William K. 1973 Ethics. 2nd ed. Englewood Cliffs, N.J.: Pren- tice Hall. Freed, Daniel 1982 Dangerous Offenders and the Bail Process: Protecting Public Safety Without Preventive Detention. Unpublished paper. Yale Univer- sity, New Haven, Conn. Gaynes, Elizabeth 1982 Typology of State Lauds Which Permit the Consideration of Danger in the Pretrial Re- lease Decision. Washington, D.C.: Pretrial Services Resources Center. Gillers, Stephen 1983 Selective incapacitation: does it offer more or less? Record of the Association of the Bar of the City of New York 38 (41. Golding, Martin P., ed. 1966 The Nature of Law. New York: Random House. 353 GoldLamp, John 1979 Two Classes of Accused: A Study of Bail and Detention in American Justice. Cambridge, Mass.: Ballinger. Greenwood, Peter, with Abrahamse, Allan 1982 Selective Incapacitation. Santa Monica,  Calif.: Rand. Greenwood, Peter, Chaiken, Jan, and Petersilia, Joan 1977 The Criminal Investigation Process. Lexing- ton, Mass.: Heath. Harper, Andrew, and McGillis, Daniel 1977 The Major Offense Bureau: An Exemplary Project. Washingon, D.C.: U.S. Government Printing Office. Hart, H. L. A. 1968a Law, Liberty and Morality. Stanford, Calif.: Stanford University Press. 1968b Punishment and Responsibility. New York: Oxford University Press. Institute for Judicial Administration and American Bar Association 1917 Standards for Juvenile Justice. Cambridge, Mass.: Ballinger. Institute for Law and Social Research 1976a Case Screening. Briefing Paper No. 2. Wash- ingon, D.C.: Institute for Law and Social Research. 1976b Uniform Case Evaluation and Rating. Brief- ing Paper No. 3. Washington, D.C.: Institute for Law and Social Research. 1977 Curbing the Repeat Offender: A Strategyfor Prosecutors.Washingon, D.C.: U.S. Govern- ment Printing Office. Kaplan, John 1965 The prosecutorial discretion. Northwestern Law Review 60:174-193. Klepper, Steven, Nagin, Daniel, and Tierney, Luke Jon 1983 Discrimination in the criminal justice sys- tem. Pp. 55-128 in Alfred Blumstein, Jac- queline Cohen, Susan E. Martin, and Michael Tonry, eds., Research on Sentenc- ing: The Search for Reform. Vol. II. Wash- ingon, D.C.: National Academy Press. Lazar Institute 1981 Pretrial Release: A National Evaluation of Practices and Outcomes. Washington, D.C.: Lazar Institute. Lindsey, Edward 1925 Historical sketch of the indeterminate sen- tence and parole system.Journal of Criminal La~v and Criminology 16. Martinson, Robert 1974 What works: questions and answers about prison reform. Public Interest 35:22~4.

354 McNamara, Richard B. 1982 Constitutional Limitations on Criminal Pro- cedure. New York: McGraw-Hill. McNeely, R. L., and Pope, Carl E., eds. 1981 Race, Crime and Criminal Justice. Beverly Hills, Calif.: Sage Publications. Meehl, Paul E. 1954 Clinical Versus Statistical Prediction: A Theoretical Analysis and a Review of the Evidence. Minneapolis: University of Min- nesota Press. Mitford, J. 1973 Kind and Usual Punishment: The Prisoner's Business. New York: Knopf. Monahan, John 1981 Predicting Violent Behavior. Beverly Hills, Calif.: Sage Publications. 1982 The case for prediction in the modified des- ert model of criminal sentencing. Interna- tional Journal of Law and Psychiatry 5: 10~113. Monahan, John, and Steadman, Henry 1983 Crime and mental disorder: an epidemio- logical approach. Pp. 14~189 in Michael Tonry and Norval Morris, eds., Crime and Justice: An Annual Review of Research. Vol. 4. Chicago, Ill.: University of Chicago Press. Moore, Mark H. 1983 Invisible offenses: a challenge to minimally intrusive law enforcement. Pp. 17~2 in Gerald M. Caplan, ea., ABSCAM Ethics: Moral Issues and Deception in Law Enforce- ment. Washington, D.C.: The Police Foun- dation. Moore, Mark H., Estrich, Susan R., McGillis, Daniel, and Spelman, William 1983a Dealing With Dangerous Offenders, Vol. 1: Final Report. John F. Kennedy School of Government. Cambridge, Mass.: Harvard University. 1983b Dealing With Dangerous Offenders, Vol. 11: Selected Papers. John F. Kennedy School of Government. Cambridge, Mass.: Harvard University. 1984 Dangerous Offenders: The Elusive Target of Justice. Cambridge, Mass.: Harvard Univer- sity Press. Morris, Herbert 1976 Persons and Punishment. Pp. 31~3 in Her- bert Morris, ea., On Guilt and Innocence: Essays in Legal Philosophy and Moral Psy- chology. Berkeley: University of California Press. Morris, Norval 1974 The Future of Imprisonment. Chicago, Ill.: University of Chicago Press. CRIMINAL CAREERS AND CAREER CRIMINALS 1982 Madness and the Criminal Law. Chicago, Ill.: University of Chicago Press. Morris, Norval, and Miller, Marc 1985 Predictions of Dangerousness. Pp. 1-50 in Michael Tonry and Norval Morris, eds.,  Crime and Justice: An Annual Review of Research, Vol. 6. Chicago, Ill.: University of Chicago Press. Morris, Norval, and Tonry, Michael, eds. 1983 Crime and Justice: An Annual Review of Research. Vol. 3. Chicago, Ill.: University of Chicago Press. Nagel, Stuart, and Neef, Marian G. 1977 The Legal Process. Beverly Hills, Calif: Sage Publications. National Institute of Justice 1981 Pretrial Release: A National Evaluation of Practices and Outcomes. Washington, D.C.: U.S. Department of Justice. Packer, Herbert 1968 The Limits of the Criminal Sanction. Stanford, Calif.: Stanford University Press. Petersilia, Joan 1983 Racial Disparities in the Criminal Justice System. Santa Monica, Calif.: Rand. President's Commission on Law Enforcement and the Administration of Justice 1967a The Challenge of Crime in a Free Society. Washington, D.C.: U.S. Government Print- ing Office. 1967b Task Force Report: Juvenile Delinquency. Washington, D.C.: U.S. Government Print- ing Office. Rawls, John 1972 A Theory of Justice. Cambridge, Mass.: Harvard University Press. Reid, Robert 1979 The Invisible Victim. New York: Basic Books. Rhodes, William, Tyson, Herbert, Weekly, James, Conly, Catherine, and Powell, Gustave 1982 Developing Criteria for Identifying Career Criminals. Unpublished paper. Institute for Law and Social Research, Washington, D.C. Roth, Jeffery A., and Wice, Paul B. 1980 Pretrial Release and Misconduct in the Dis- trict of Columbia. Washington, D.C.: Insti- tute for Law and Social Research. Sellin, Thorsten, and Wolfgang, Marvin E. 1964 The Measurement of Delinquency. New York: John Wiley & Sons. Sherman, Michael, and Hawkins, Gordon 1981 Imprisonment in America. Chicago, Ill.: University of Chicago Press. Silberman, Charles E. 1978 Criminal Justice, Criminal Violence. New York: Random House.

PURBLIND JUSTICE Singer, Richard G. 1919 Just Deserts: Sentencing Based on Equality and Desert. Cambridge, Mass.: Ballinger. Steffel, Linda 1977 The Law and the Dangerous Criminal. Lexington, Mass.: Lexington Books. Stone, Alan 1975 Mental Health and the Law: A System in Transition. Washington, D.C.: U.S. Govern- ment Printing Office. Szasz, Thomas 1977 Psychiatric Slavery. New York: Free Press. Twentieth Centur,v Fund 1978 Confronting Youth Crime. New York: Holmes and Meter. Twentieth Century Fund Task Force on Criminal Sentencing 1976 Fair and Certain Punishment. New York: McGraw-Hill. Underwood, Barbara 1979 Law and the crystal ball: predicting behavior with statistical inference and individualized justice. Yale Law Journal 88:1432-1447. van den Haag, Ernest 1975 Punishing Criminals: Concerning a Very Old and Painful Question. New York: Basic Books. Vera Institute of Justice 1976 Impact Evaluation of the VictimtWitness As- sistance Project's Appearance Management Activities. New York: Vera Institute of Justice. 1977 Felony Arrests: Their Prosecution and Dis- vosition in New York Citu Courts. New - York: Vera Institute of Justice. von Hirsch, Andrew 1976 Doing Justice: The Choice of Punishments. New York: Hill and Wang. 1981a Desert and previous convictions in sentenc- ing. Minnesota Law Review 65:591~34. 1981bUtilitarian sentencing suscitated: the Ameri- can Bar Association's second report on criminal sentencing. Rutgers Law Review 33:772-789. 1983 Commensurability and crime prevention. Journal of Criminal Law and Criminology 74:20~248. 1984a"Equality," "anisonomy," and justice: a re- view of"Madness and the Criminal Law." Michigan Law Review 82:109~1112. 1984b The eithics of selective incapacitation: obser- vations on the contemporary debate. Crime and Delinquency 30:17~194. 355 1985 Past and Future Crimes. New Brunswick, N.J.: Rutgers University Press. von Hirsch, Andrew, and GottlGredson, Don M. 198~Selective incapacitation: some queries about 1984 research design and equity. New York Uni- versity Review of Law and Social Change VII: 11~5. Vorenberg, James 1981 Decent restraint of prosecutorial power. Harvard Law Review 94:1521-1573.  Walker, Nigel 1982 Unscientific, unwise, unprofitable or unjust? British Journal of Criminology 22:27 289. Weinreb, Lloyc} L. 1979 Denial ofJustice. New York: Macmillan. Wilson, James Q. 1983a Crime and Public Policy. Cambridge, Mass.: Harvard University Press. 1983b Thinking About Crime. 2nd rev. ed. New York: Basic Books. Winston, Kenneth 1974 On treating like cases alike. California Law Review 62:1~39. Wish, Eric D., Klumpp, Kandace, Moorer, Amy, Brady, Elizabeth, and Williams, Kristen 1981 An Analysis of Drugs and Crime Among Arrestees in the District of Columbia. Wash- ington, D.C.: U.S. Department of Justice. Withcomb, Debra 1980 Major Violator Unit, San Diego California: An Exemplary Project. Washington, D.C.: U.S. Government Printing Office. Wolfgang, Marvin E., and Tracy, Paul E. 1983 The 1945 and 1958 birth cohorts: a compar- ison of the prevalence, incidence and sever- it~v of delinquent behavior. In Mark H. Moore, Susan R. Estrich, Daniel McGillis, and William Spelman, eds., Dealing With Dangerous Offenders, Vol. Il.: Selected Pa- pers. John F. Kennedy School of Govern- ment. Cambridge, Mass.: Harvard Univer- sit~v. Zimring, Franklin L. 1977 Making the punishment fit the crime: a con- sumer's guide to sentencing reform. Univer- sity of Chicago Law School Occasional Pa- pers No. 12. 1978 Background paper. In Twentieth Century Fund, Confronting Youth Crime. New York: Holmes and Meier.