The Elusive Concept of Dangerousness: The State of the Art in Criminal Legal Theory and the Necessity of Further Research

Preventing future crime has become an increasingly dominant function of the criminal law of many liberal democracies. This “preventive turn” has led to a profound debate on the legal and ethical boundaries of the “preventive state.” However, the concept at the core of preventive justice—the dangerousness of the offender—has attracted relatively little attention in the current debate. This is remarkable, as the legal establishment of dangerousness permits intrusive preventive measures, such as preventive detention for an indeterminate period of time. In the past, various concepts of dangerousness have been developed by criminal law scholars. We discuss these concepts in a chronological order to demonstrate how the meaning of dangerousness has evolved over time, and how it has been shaped by concurrent developments in forensic psychiatry and penology. Our description of the state of the art of legal scholarship on the concept of dangerousness also shows the lack of a fully developed theory of dangerousness, and therefore the necessity of further research. We identify five “aspects” of the concept of dangerousness on which scholars have widely diverging views. These five aspects are intended to guide further research on the concept of dangerousness in preventive criminal law.


I. Introduction
During the past decades, the criminal law of many liberal democracies has taken a "preventive turn." This term was coined to describe the increasing importance that is attached to the preventive function of criminal law, i.e., preventing crime before having to react to it. 1 The preventive turn manifests itself-among other things-in the fact that preventive measures play an increasingly important role in the criminal law. 2 These measures vary widely in nature, but have in common that they-to different degrees-restrict the liberty of offenders or deprive them of it in order to reduce risks that are associated with them, most prominently the risk of recidivism. 3 The imposition (or extension) of preventive measures often requires that the offender is determined to be "dangerous," i.e., that the offender poses a sufficiently high risk of the kind that is to be reduced. 4 In recent decades, the number of potential preventive measures, as well as the number of offenders that is subjected to these measures has increased significantly in many liberal democracies. 5 This development is a manifestation of what Lucia Zedner and Andrew Ashworth, inspired by Carol Steiker, 6 portray as the "rise of the preventive state." 7 While acknowledging that providing security has always been a core function of the state and criminal law a means to serve that end, Zedner and Ashworth perceive: a shift that entails both the greater prevalence of risk and security and a consequent increase in preventive laws and measures. Designating these as traits of the Preventive State is not intended to ground grandiose epochal claims about change nor to ignore the continuing variety of modes, means, and aims of crime control. Rather it aims to illuminate salient and important shifts worthy of closer criminological attention and to make possible greater analytical clarity about the ways in which states seek to govern. 8 While canvassing causes and manifestations of the rise of the preventive state is a criminological and sociological enterprise, Zedner and Ashworth, again following Steiker, point at the relative neglect in the legal doctrine on the preventive part of criminal law compared to the doctrinal scrutiny applied to the punitive, blame-based criminal law (the "punitive" state). 9 Steiker put the preventive state on the research agenda in the 1990s. The key question was, and still is today, "What constitutional and/or policy limits are there on the non-punitive preventive state?" 10 More than two decades later, some progress has been made in carving out normative limits on the preventive state. Most of these limits are procedural (e.g., the proposal of a presumption of harmlessness 11 ), concern the law of evidence (e.g., the required standard of proof for dangerousness 12 ), deal with substantive criminal law (e.g., limits on criminalization of inchoate offences 13 ), or regard specific forms of liberty deprivation, (e.g., preventive detention 14 ). What has attracted far less attention in the present debate, however, is a concept at the core of the preventive state: the dangerousness of the offender that is required to impose (or extend) many preventive measures in the first place.
Legal scholarship has not always lacked interest in the concept of dangerousness in criminal law. As we show in this article, several scholars, almost all from the United Kingdom, have attempted to conceptualize dangerousness in criminal law. These concepts are relatively old, however: most date back from before the preventive state as such attracted scholarly attention. This The Elusive Concept of Dangerousness 143 could explain why these concepts seem to be overlooked in current research. For this reason, we believe it is fruitful to give an account of the state of the art on the concept of dangerousness in criminal legal theory. In this article, we examine ten concepts of dangerousness that have been proposed since the 1970s (Sections II-VI). We discuss these concepts in a more or less chronological order to demonstrate how the meaning of dangerousness has evolved over time, and how it has been influenced by concurrent developments in forensic psychiatry and penology. Our description of the state of the art of the legal scholarship on the concept of dangerousness also shows that this concept is not yet fully developed, and therefore merits further research. In this regard, we identify five "aspects" of the concept of dangerousness on which the scholars that we discuss have widely diverging views (Section VII). These aspects are (1) the nature of the harm, (2) the probability of the harm, (3) the extent to which dangerousness entails a fact and/or a value judgement, (4) whether dangerousness is a personal characteristic or is based on abstract risk factors, and (5) the temporality of dangerousness. These five aspects are intended to guide further research on the concept of dangerousness. We end this article with a short conclusion (Section VIII).

II. Dangerousness Until Its Renaissance in the 1970s
The starting point of our chronological inquiry into the concept of dangerousness in criminal legal theory is what Bottoms dubbed in 1977 the "renaissance of dangerousness." 15 "Renaissance" refers to a renewed interest in the United Kingdom in the 1970s in dangerousness as a ground for preventive measures. In this decade, several British expert commissions attempted to conceptualize dangerousness; we discuss these concepts in Section III. 16 At that time, dangerousness was not a new concept in criminal law, but was given a different meaning than in the past. Before turning to the renaissance of dangerousness, we briefly discuss these earlier meanings of dangerousness.
Until the turn of the twentieth century, dangerousness was, unlike now, associated with dangerous groups, rather than with dangerous individual offenders. For example, for the greatest part of the nineteenth century, in criminal law certain classes were deemed to be dangerous. Concerns about social unrest made the middle and upper classes suspicious of the working class and of particular groups of social outcasts, such as common criminals, beggars, and vagrants. The crimes that members of these dangerous classes committed were regarded as acts of resistance against the social and political order. In response, governments of many liberal democracies expanded the criminal law and, more generally, the powers of the state to "discipline" 17 these dangerous classes. 18 By the late nineteenth century, however, the perceived threat posed by the dangerous classes had largely dissipated. At this point, the meaning of dangerousness in criminal law shifted from dangerous groups to dangerous individuals, in Max de Vries and Johannes Bijlsma particular individual repeat offenders. 19 The public viewed repeat offenders of property crimes as particularly dangerous, more so than offenders of violent and sex crimes. Property crimes, like burglary, fraud, theft, and robbery, were feared as lost property was usually uninsured and often irreplaceable for the owner. 20 In the early twentieth century, many governments introduced dangerous offender laws, also known as habitual offender laws. In general, these laws increased the minimum and/or maximum prison sentences in case of recidivism to incapacitate repeat offenders for a prolonged period of time. 21 By the mid-twentieth century, however, society no longer feared property crime as much as before. At the same time, serious violent and sexual crimes caused more anxiety to the public, particularly if committed by offenders with a mental disorder. 22 Since then, dangerousness became mostly associated with (mentally disordered) violent and sex offenders. 23 Particularly the image of the "stranger-predator" personifies the dangerous offender. 24 Predominantly in the second half of the twentieth century, many governments introduced preventive measures to prevent recidivism of these types of offenders. These new measures came alongside existing laws against repeat offenders. 25 These developments preceded the "renaissance of dangerousness" in the 1970s. Bottoms coined the term in response to the reports of the Butler Committee and the Scottish Council on Crime. On request of respectively the British and Scottish governments, these expert groups explored possible preventive measures against dangerous offenders. Shortly after Bottoms wrote about the renaissance of dangerousness, another report was published, this time of the Advisory Council on the Penal System. All three reports offer a concept of dangerousness. The reason for this development could be that it was advocated at that time that precise definitions were needed to prevent arbitrary judgements on the dangerousness of offenders. 26 We discuss these concepts of dangerousness below, starting with the report of the Butler Committee.
III. Renaissance of Dangerousness in the 1970s: The Butler Committee, Advisory Council on the Penal System, and Scottish Council on Crime Reports

Butler Committee
The Committee on Mentally Abnormal Offenders (in this section: the committee), known after its chairman as the Butler Committee, was established by the British government in 1972 to investigate the interaction between offenders with a mental disorder and criminal justice. 27 In its 1975 report, the committee observed that dangerous and mentally disordered offenders that were treated by mental health services could be held in care indefinitely. These offenders were only discharged when their dangerousness was sufficiently reduced. Dangerous and mentally disordered offenders with a determinate prison sentence, however, were always released upon completing The Elusive Concept of Dangerousness 145 their sentence. The release of these "probably still dangerous [offenders] who are not acceptable for treatment in hospital" presented, according to the committee, "a serious defect in society's defenses which ought to be made good." 28 Therefore, the committee proposed the introduction of the "reviewable prison sentence": an indeterminate prison sentence for dangerous and mentally disordered offenders that would detain the offender until "his progress under treatment … allow[s] him to be released under supervision without serious risk to the public." 29 For the purpose of the imposition and periodical review of the reviewable prison sentence, the committee defined dangerousness as the "propensity to cause serious physical injury or lasting psychological harm." In the committee's view, dangerousness concerns danger of violent crime, since "physical violence is … what the public are most worried about." Non-violent crime that causes lasting mental harm is also dangerous according to the committee, since "the psychological damage which may be suffered by some victims of other crimes is not to be underrated." 30 The committee did not explain what it meant by a "propensity" to do harm, but it did provide a hint. According to the committee, dangerousness lies in the majority of cases in the personality of the offender and the social circumstances in which the offender finds him-or herself. Therefore, the experts that determine the offender's dangerousness should consider both. 31 Although the reviewable prison sentence should only be imposed on mentally disordered offenders, the committee noted that the mental illness of the offender does not necessarily have to be the cause of his or her dangerousness. The committee pointed out that offenders can remain dangerous after successful treatment of their disorder. 32 This raises the question why the committee stressed that the presence of a mental disorder should be a separate requirement for the imposition of the reviewable prison sentence. The committee did not explain this decision. A consequence of the committee's position is that dangerous offenders without a mental disorder cannot be taken into preventive detention. 33

Scottish Council on Crime
Meanwhile, in Scotland, the Scottish government established the Scottish Council on Crime in 1972 (in this section: the council) to review the prevention of crime and the treatment of offenders. 34 The reason for this review was the upsurge in crime, particularly violent crime, in the 1950s and 1960s. 35 In its 1975 report, the council proposed the introduction of the "public protection order" to incapacitate violent offenders. This order would secure "the continued detention of a violenceprone offender until it is safe for him to be released." 36 The council intended the order both for offenders with and without a mental disorder. 37 In the council's view, a "violenceprone offender" is a repeat violent offender. The council expressed this in the requirements for the imposition of the public protection order, which are dangerousness of the offender and "evidence of past acts suggesting a persistent tendency to violence." 38 The council defined dangerousness as "the probability that

Max de Vries and Johannes Bijlsma
[the offender] will inflict serious and irremediable personal injury in the future." 39 The term "serious and irremediable personal injury" suggests that the danger must regard serious violent crimes. According to the council, the dangerousness of an offender depends on the probability that he or she commits such a serious violent crime in the future. Therefore, the imposition of the public protection order "must have a basis in an opinion regarding the likelihood of future serious personal harm." 40 Thus, the extent to which an offender is dangerous depends on the probability of future harm. The council was the first to argue that dangerousness is a gradual concept, and that the determination of dangerousness requires an assessment of a probability. This preludes to the development of actuarial concepts of dangerousness that will be discussed in Section V.
The council was ambiguous about which probability of future violent crime is required for the imposition of the public protection order. 41 According to the council, this probability should in any case be sufficiently high to justify the preventive detention of the offender: "orders will be made only in … whose cases the risk … of future serious violence is so high that their detention is justified for the protection of others." 42 The council made conflicting statements in this respect: it variously spoke of "those offenders who, on all the evidence, are very likely indeed, if at liberty, to commit further serious crimes of violence," "the risk, though not necessarily the certainty, of future serious violence," "the Court could make the Order only if it was satisfied of the risk of further violence," 43 and "that there is substantial likelihood that the offender is the sort of person who, having regard to all the circumstances, will, if set at large, commit acts causing or threatening physical harm to others." 44 Bottoms criticized the council's ambiguity about the probability of future violent crime that is required for the imposition of the public protection order. He argued that such imprecise criteria would be inconsistently applied if they were to become law. 45

Advisory Council on the Penal System
The 1978 report of the Advisory Council on the Penal System (in this section: the advisory council), which had been set up by the British government in 1975, echoed the report of the Butler Committee discussed in Section III.1. The advisory council proposed the introduction of the "exceptional sentence" for dangerous offenders, which would have similar terms as the reviewable prison sentence proposed by the Butler Committee. 46 Unlike the Butler Committee, however, the advisory council intended the exceptional sentence both for offenders with and without a mental disorder.
To impose the exceptional sentence, the offender must have caused serious harm in the past and be judged dangerous, that is, he or she must be "likely to commit serious harm." In its report, the advisory council avoided using the word "dangerousness," even though "likely to commit serious harm" amounts to the same thing. 47 The term "likely"-when compared to highly likely-seems to suggest that a rather low probability of serious harm would suffice. The advisory The Elusive Concept of Dangerousness council defined "serious harm" as "serious physical injury, serious psychological effects … ; exceptional personal hardship … ; and damage to the security of the State, or to the general fabric of society." 48 It is also not required that the crime that the offender is likely to commit would actually result in serious harm: "an attempt, a threat or a conspiracy to do serious harm" would be sufficient in the eye of the advisory council. 49 Thus, it appears that the conditions to impose the exceptional sentence are rather low, and definitely lower than the conditions of the reviewable prison sentence. This is reinforced by the advisory council's suggestion that about ten percent of the offenders that were sentenced to imprisonment at that time might be subject to the exceptional sentence. 50 For this reason, the advisory council's proposal was heavily criticized, for example by Leon Radzinowicz and Roger Hood, who argued that it "would be received with open arms by any authoritarian state." 51

IV. The 1981 Floud Report
The 1981 report of the Working Party on the Dangerous Offender of the Howard League for Penal Reform, known as the Floud Report, formulated the most comprehensive concept of dangerousness until then. Unlike the previously discussed reports, the concept that the Floud Report described is not related to a proposal for a preventive measure, 52 as Jean Floud and Warren Young, the report's principal authors, intended to give a "broader, social perspective on the problem." 53 Key to Floud and Young's concept of dangerousness is the distinction between "danger" and "risk." According to them, the difference between danger and risk is that "dangers are unacceptable risks." 54 Thus, they saw risks as indifferent as to whether they should or should not be managed; risks may or may not be taken. Dangers, however, should be avoided, meaning that dangerous offenders should in principle be subjected to preventive measures. 55 Whether a risk is unacceptable and therefore a danger, depends on how people perceive this risk: "we speak of danger when we judge the risk unacceptable and call for preventive measures." As perceptions of risks differ among people, there are conflicting views on which offenders are dangerous. 56 In Floud and Young's view, an individual offender's dangerousness should be determined with a "predictive judgement." 57 This judgement should, in any case, be individualized. Floud and Young therefore argued that a predictive judgement cannot simply consist of an actuarial assessment of the offender's risk of recidivism, for they found it "not just to take preventive measures against an offender solely on the strength of his being a member of a statistical class of high-risk offenders." 58 However, as we will see, they did not entirely exclude actuarial risk assessment from the determination of dangerousness. In Floud and Young's view, a predictive judgement consists of two elements: an evaluation of the offender's character, and the offender's risk of recidivism. 59 We discuss both elements of a predictive judgement below.

Max de Vries and Johannes Bijlsma
Regarding the character evaluation, Floud and Young argued that dangerousness lies in a disposition of the offender to do willful harm to others. This disposition is essentially a character trait, a "pathological attribute of character," as they called it. This character trait involves, according to Floud and Young, "a propensity to inflict harm on others in disregard or defiance of the usual social and legal constraints." 60 Floud and Young based this view on earlier work of the psychiatrists Harry Kozol, Richard Boucher, and Ralph Garofalo, 61 who defined dangerousness as "a pathological selfserving potential for violence." 62 Floud and Young added to this definition that the offender must inflict this harm willfully. In this regard, willfulness means that the offender not simply violates, but also shows contempt for social and legal norms. 63 The reason that Floud and Young included willfulness in their concept of dangerousness is that intentionally committed crimes are seemingly feared more than recklessly or negligently committed crimes. According to Floud and Young, "the prospect of death or injury suffered at the hands of another person arouses greater alarm than death or injury suffered as the direct result of their dangerous or irresponsible behaviour." 64 Floud and Young argued that a predictive judgement additionally requires an assessment of the risk of recidivism of the offender, since "if there is little likelihood of his actually doing harm in the future, the case for preventive measures is weakened no matter how likely it is that we are right to believe the assessment of his character or intentions." According to them, whether an offender "will actually do harm is very much a matter of chance." 65 They argued that experts can objectively ascertain this chance, for example by means of (actuarial) risk assessment. 66 However, whether the risk of recidivism of an offender makes him or her a dangerous offender depends on whether people perceive this risk as a danger. This reflects their position that dangers are risks that people find unacceptable. 67 According to Floud and Young, how an offender's recidivism risk is perceived, is influenced by the intensity of the fear of people of the crimes that are predicted. In turn, they argued, the vulnerability of people to crime determines how much they fear it. 68 Thus, as Floud and Young famously put it, "fear converts risk into danger." 69 An issue with fear of crime, however, is that it can be irrational and biased. 70 People fear crime, even though the chance of becoming a victim of crime is relatively small. 71 Floud and Young acknowledged this fact, although they also claimed that "public judgments of danger do not seem as inherently irrational and inconsistent as is sometimes suggested." They argued that "fear is a function of personal vulnerability and, at a given level of risk of serious harm, it varies inversely with the time and distance that separates the prospective victim from the predicted harmful event." 72 Nonetheless, Floud and Young attempted to pre-empt potential criticism 73 by insisting that only rational fears may be considered in predictive judgements of dangerousness, meaning that "comparable risks must be consistently evaluated." According to Floud and Young, this additional In the 1980s, a more conservative penal ideology that placed the risk of crime on the foreground of penal policy emerged. Malcolm Feeley and Jonathan Simon coined the term "New Penology" for this ideology. 75 What possibly caused this development was that modern society had become increasingly preoccupied with risk and safety, 76 constituting, as Ulrich Beck famously suggested in the 1990s, a risk society. 77 Since the tolerance of society towards risks had decreased over time, the identification and management of offenders with an elevated risk of reoffending had become more important. 78 Along with the implementation of the new penology came an increasing use of actuarial risk assessment instruments to determine the risk of recidivism of offenders, a trend also referred to as actuarialism or actuarial justice. 79 Against this background, several authors developed actuarial approaches to dangerousness. Distinctive of actuarial dangerousness, the name we give to these approaches, is its premise that the dangerousness of offenders can be empirically measured and quantified. This premise is similar to the assumption that actuarial risk assessment instruments can empirically measure and quantify the recidivism risk of offenders, hence the term. In this section, we discuss three actuarial concepts of dangerousness that were developed in the 1980s, namely those of Anthony Bottoms and Roger Brownsword, Norval Morris and Marc Miller, and Nigel Walker. We start with the most comprehensive concept, which is the one of Bottoms and Brownsword.

Bottoms and Brownsword
Bottoms and Brownsword based their concept of dangerousness on the theory of rights of Ronald Dworkin, in particular his vivid danger test. 80 According to Dworkin, criminal law "should treat a man against his will only when the danger he presents is vivid, not whenever we calculate that it would probably reduce crime if we did." 81 Dworkin's vivid danger test is incomplete as a concept of dangerousness, however, as it does not specify when an offender constitutes a danger and when this danger is sufficiently vivid to justify the imposition of preventive measures.
Bottoms and Brownsword attempted to complete Dworkin's vivid danger test as a concept of dangerousness. According to Bottoms and Brownsword, this test consists of "three main components: seriousness (what type and degree of injury is in contemplation?); temporality, which breaks down into frequency (over a given period, how many injurious acts are expected?) and immediacy (how soon is the next injurious act?); and, certainty (how sure are we that this person will act as predicted?)." 82

Max de Vries and Johannes Bijlsma
To determine the dangerousness of an offender, the court or parole board has to score each of these three components on a scale. The offender is dangerous if he or she scores above certain predetermined levels on these scales. This implies that Bottoms and Brownsword believed that dangerousness is gradual, not binary, and that it can be quantified on an interval scale, which is characteristic of actuarial concepts of dangerousness. 83 According to Bottoms and Brownsword, how high the score on each component of this test should be for the verdict that the offender is dangerous, depends on the preventive measure that is considered. Bottoms and Brownsword did not have a particular measure in mind but discussed measures that involve preventive detention in general. For preventive detention, the vivid danger test should be so strict that it is only met in "exceptional cases." 84 Bottoms and Brownsword argued that "the right not to be detained should not be overridden unless, on an amalgam of these factors [the components of the vivid danger test], the rights of others really are substantially threatenedin other words, that the danger he presents is vivid." 85 Thus, if preventive detention is considered, it appears that Bottoms and Brownsword would require high scores on all three components to pass the vivid danger test: the predicted conduct must be of a serious nature, there must be a substantial risk of the predicted conduct and this risk must be acute. 86 The reason for this is that Bottoms and Brownsword, unlike the Butler Committee, the Advisory Council on the Penal System and the Scottish Council on Crime, who all proposed the introduction of new measures, called for restraint in the expansion of preventive criminal law. 87 According to Bottoms and Brownsword, certainty is the most important component of the vivid danger test. They argued that "if there is a very low score on the certainty factor, then whatever the danger it is hardly vivid. However, as the score increases on the certainty element, the risk becomes increasingly vivid and then we have to look very carefully at the kind of danger threatened." 88 Bottoms and Brownsword explained this with an example. If the probability of violence is, for instance, one in three, then it would be "barely conceivable that the vivid danger test would ever be met," even if there are high scores on seriousness and temporality. 89 However, if the probability is one in two and if there are high scores on seriousness and temporality, then the vivid danger test could be passed according to Bottoms and Brownsword. 90 Thus, although certainty is the most important component, it can never be decisive on its own: "Even in conditions of low certainty there comes a point where vivid danger is arguable by virtue of high scores on seriousness and temporality." 91 Conversely, low scores on the seriousness and temporality scales "militate strongly against the danger being described as vivid." 92

Morris and Miller
In the 1980s, Bottoms and Brownsword were not the only legal scholars who argued that the dangerousness of offenders can be empirically measured and quantified. Morris and Miller, for example, asserted that "measuring the The Elusive Concept of Dangerousness probability and severity of harm … is an empirical, scientific activity." 93 Therefore, the probability and the potential injury at stake should be measured with an empirical, scientific method, that is: actuarial risk assessment. 94 Morris and Miller defined dangerousness as a formula: it is the product of the probability and the severity of the harm. According to Morris and Miller, "the key elements [of dangerousness] are the type and magnitude of harm predicted and the predicted level of risk or the rate of that harm, the product of these variables being a measure of total harm that at some point many in our society would agree constitutes dangerousness." 95 Morris and Miller did not specify when this "total harm" would be sufficiently high to find an offender dangerous. They stated, with reference to Floud and Young, that "defining unacceptable levels of dangerousness [is] a social and political rather than an empirical task." 96 Therefore, Morris and Miller did not elaborate on what they would find a right criterion for dangerousness as requirement for the imposition of preventive measures. They only remarked that "[t]o justify protective sentencing, the level of prediction must be high and the threatened harm severe, whereas a much lower level of risk may properly be relied on to justify a lesser deprivation of liberty." 97 Morris and Miller's concept of dangerousness is in part similar to that of Bottoms and Brownsword, but there are three notable differences. First, temporality is not mentioned as a variable of dangerousness in the concept of Morris and Miller, although it could be that they range the temporality under the variables probability and severity. Second, in Morris and Miller's concept, an offender is dangerous if the total harm is above a predetermined level, whereas in Bottoms and Brownsword's concept, each separate component of dangerousness must be above a particular level, and no multiplication of these components takes place. Third, Morris and Miller limit the dangers that dangerousness is concerned with to "intentional behaviour that is physically dangerous to the person or threatens a person or persons other than the perpetratorin effect, to assaultive criminality." 98 In Morris and Miller's view, assaultive criminality is the type of crime that is commonly understood as dangerous and, in their words, "would colloquially be called the behaviour of a dangerous criminal." This could, according to them, not be said of other types of crimes. 99

Walker
The actuarial concept of dangerousness of Walker is slightly older than the concepts of Bottoms and Brownsword and Morris and Miller but is not as comprehensive and we therefore discuss it last. Like Morris and Miller, Walker attempted to capture dangerousness in a formula: "It is tempting to say, pseudo-scientifically, that danger = seriousness × probability of harm." 100 Walker, however, was not fully convinced of his formula of dangerousness: "This [formula] ignores one fact, however: that if either seriousness or probability is below a certain level we do not think of the situation as dangerous. It also assumes that the relationship is multiplicative, whereas we do not know this. Research might conceivably

Max de Vries and Johannes Bijlsma
show that a more complex mathematical expression would fit our way of thinking about danger: but surely it would merely lend precision to what is already sufficiently obvious." 101 Regardless of his reservations about his formula, Walker asserted that dangerousness can be quantified and represented on a scale. An offender is dangerous if the degree of dangerousness is above a predetermined level on this scale. 102 According to Walker, "[a] dangerous situation, action or activity is one which raises the probability of serious harm above a certain level." 103 Walker did not elaborate much on the level of dangerousness that would be necessary to classify an offender as dangerous and impose preventive measures. In this respect, Walker only stated that "[t]o define that level is not easy, but perhaps not impossible. It is the level which causes a person who is not neurotically or superstitiously anxious to become so apprehensive that he explores the possibility of avoiding the situation, action or activity. Obviously this level varies with the nature of the apprehended harm. The threshold is … higher for a broken neck than a broken leg." 104 In Walker's view, dangerousness requires a prospect of serious harm. In this regard, Walker defined serious harm as crimes that cause "serious and lasting hardship to other individuals, of a kind which, once caused, cannot be remedied." 105 This definition includes crimes that lead to "lasting psychological harm as well as disabling or disfiguring physical injury." Property crimes, however, do not constitute serious harm, according to Walker, "since most loss of or damage to property can be remedied by compensation." 106

VI. Dangerousness from 2000 Onwards: Duff, Slobogin, and Ashworth and Zedner
In the 1990s, the interest in the concept of dangerousness in criminal law waned. In this period, no new concepts of dangerousness were formulated in legal theoretical literature. A possible explanation could be that the vocabulary of law and forensic psychiatry started to diverge during this period. Forensic psychiatry moved away from the legal concept of dangerousness to the allegedly more neutral concept of risk of recidivism. 107 Peter Snowden, for example, argued that risk, unlike dangerousness, does "not contain pejorative connotations" and would therefore invite psychiatrists to a "more objective and robust analysis." 108 Most legislators and legal scholars, however, stuck with the term "dangerousness." According to Sophie Holmes and Keith Soothill, the divergence between these disciplines has made it more difficult to understand and reach consensus on the meaning of dangerousness. 109 In the 2000s, the interest in dangerousness increased again in legal theory. Acts of terrorism and the subsequent adoption of anti-terrorism legislation in many states caused what could be called the second renaissance of theorizing dangerousness. These anti-terrorism laws introduced preventive measures that The Elusive Concept of Dangerousness target presumed terrorists and known terrorist offenders that might reoffend. 110 Since the 2000s, three concepts of dangerousness have been proposed in the legal theoretical literature by-in a chronological order-Antony Duff, Christopher Slobogin, and Andrew Ashworth and Lucia Zedner. Unlike in the previously discussed periods, there is no clear common ground to be found in these concepts. Therefore, we discuss these concepts chronologically, starting with Duff's.

Duff
In Section IV, we discussed the concept of dangerousness of Floud and Young, who argued that dangerousness lies in a "pathological attribute of character," namely a disposition to cause willful harm to others. Duff builds on this idea of Floud and Young. Duff discussed his concept as part of his proposal for "special selective detention" (SSD), an order designed to incapacitate dangerous offenders. He described SSD as follows: "SSD is special in that it involves longer periods of imprisonment than those to which the offenders would normally be liable … ; it is selective in that it is focused selectively on offenders … who are thought to be particularly likely to commit further serious crimes if released at the end of a normal prison term." 111 Offenders who would be the target of SSD are typically seen as dangerous, according to Duff. 112 Duff, however, held a different view than most authors on what makes an offender dangerous. He argued that dangerousness can be found in the possession of certain character traits that cause criminal behavior, although he did not specify which traits. 113 According to Duff, criminal behavior is the manifestation of the dangerous offender's faulty character. As Duff put it, "the connection between the character traits which make a person criminally dangerous, and the criminal conduct which would manifest those character traits, is logical, not contingent. To have those character traits is to be disposed to behave in those criminal ways in situations of the appropriate kind; to say that the criminal conduct manifests those character traits is to say that that conduct constitutes the public actualization of those traits." 114 This definition of dangerousness raises the question whether every person with these character traits is dangerous. This is not the case. Duff stated that the mere possession of these character traits is insufficient to label someone as dangerous: he or she must also be a persistent offender of crimes of serious violence against the person. Being a persistent offender, however, is not a separate requirement. Duff argued that it is not the repeated commission of crime that renders an offender dangerous: it is his or her malicious character that these crimes expose. 115 Therefore, to define an offender as dangerous, the crimes that this offender committed should evidence a moral deficiency in his or her character: "Nor is [defining an offender as dangerous] justified merely by the fact that he has been convicted of several crimes of violence over the years-that he is a repeat offender, whose crimes could be seen merely as a series of discrete aberrations in an otherwise value-sensitive life. He must be a persistent offender-one whose criminal career can only be interpreted as manifesting an utter and continuing disregard or contempt

Max de Vries and Johannes Bijlsma
for the values that he flouts and for those whom he attacks." 116

Slobogin
Slobogin discussed his concept of dangerousness in relation to preventive measures that involve "longterm, pure preventive detention" of dangerous offenders. 117 He argued that for the imposition of such measures the dangerousness of the offender should be determined by two criteria: a psychological criterion and a prediction criterion. Slobogin described these criteria as follows: "The psychological criterion describes the psychological traits that distinguish those dangerous people who may be committed from those who may not be. The prediction criterion describes the level of risk that must be shown before preventive detention may take place." 118 Regarding the psychological criterion, Slobogin argued that "the core trait that normatively distinguishes the dangerous person who may be preventively detained from the dangerous person who may not be is imperviousness to criminal punishment, or what I shall call undeterrability." 119 According to Slobogin, undeterrability can arise out of two psychological tendencies: either "unawareness that one is engaging in criminal conduct," or "extreme recklessness with respect to the prospect of serious loss of liberty or death resulting from the criminal conduct." 120 The mental state of the offender distinguishes the unaware from the reckless. The subcategory of the unaware comprises mentally disordered offenders. For this reason, the criterion for this subcategory is similar to the cognitive prong of the insanity defense. 121 The subcategory of the reckless consists of legally sane offenders who know that they are committing a crime but commit it anyway, "while aware of a very substantial risk [of being] caught and subjected to a serious deprivation of liberty." To narrow down the scope of this second subcategory, Slobogin added that "the anticipated/ignored loss of freedom be substantial. That caveat ensures that the person is truly undeterrable, as opposed to some who could be deterred with significant enough disincentives." 122 When an offender meets the psychological criterion, he or she can in principle be subjected to preventive detention. However, whether this offender can actually be preventively detained, depends on whether the prediction criterion is met. Like Duff, Slobogin argued that offenders with a dangerous character are only truly dangerous if they translate their beliefs or desires into action. But contrary to Duff, Slobogin believed this requires prediction of the offender's future behavior. 123 Slobogin phrased the prediction criterion as "the degree of dangerousness necessary to justify preventive detention." 124 Slobogin did not specify which degree of dangerousness is required, as the prediction criterion depends on the preventive measure that is considered (and he did not consider a specific measure). Instead, he proposed two principles, the proportionality principle and the consistency principle, which should guide what the prediction criterion of a specific measure should look like. Firstly, the proportionality principle "states that the degree of danger required for preventive detention should be roughly proportionate to the degree The Elusive Concept of Dangerousness of liberty deprivation the state seeks." 125 Slobogin argued that this in line with legal criteria for dangerousness, which require "a lesser showing of dangerousness as one moves down the hierarchy of interventions." 126 Secondly, the consistency principle "states that the degree of dangerousness required for preventive detention should be similar to the degree of dangerousness sufficient to authorize like liberty deprivations associated with other manifestations of the state's police power." 127 This means, for example, that the degree of dangerousness that is required for preventive detention should be roughly equivalent to the degree that a conviction of an inchoate offence requires. 128

Ashworth and Zedner
The foundation of the concept of dangerousness of Ashworth and Zedner is the presumption of harmlessness. This presumption entails, according to Ashworth and Zedner, that "in principle, every citizen has a right to be presumed harmless, and this presumption of harmlessness can be rebutted only in exceptional circumstances." 129 Ashworth and Zedner argued that an offender does not forfeit this presumption on conviction for a crime, as "a past harmful intention in respect of a single circumstance or victim cannot simply be extended into the future." 130 Therefore, the imposition of a preventive measure on a known offender still requires that this presumption is rebutted.
According to Ashworth and Zedner, the presumption of harmlessness can be rebutted if the offender committed a serious violent offence and is classified as dangerous. 131 Ashworth and Zedner argued that the classification of an offender as dangerous requires a "positive risk assessment." 132 The approach of Ashworth and Zedner is similar to the actuarial concepts of dangerousness that we discussed in Section V. According to Ashworth and Zedner, risk assessment "requires, first and most obviously, an assessment of the gravity of the harm in prospect. Second, it requires separate assessment of the degree of probability that it will actually occur." The degree of gravity and the degree of probability represent "dual axes," i.e., two interval scales. A risk assessment is "positive," i.e., the offender can be classified as dangerous, if both the degree of gravity and the degree of probability are above certain predetermined levels. 133 Since Ashworth and Zedner did not consider a specific preventive measure, they did not indicate how high these levels of gravity and probability should precisely be. However, they formulated two principles that should guide which levels are at least required for the imposition of measures that involve preventive detention. The "grievous risk principle," the first principle, requires "that the prospective harm is adjudged to be so serious as to justify the deprivation of liberty suffered by the individual detained." This principle should be combined with a "high probability requirement," the second principle, which entails "that the predicted occurrence is of such imminence and high likelihood as to justify detention." 134 The dual axes of gravity and probability are interrelated. According to Ashworth and

Max de Vries and Johannes Bijlsma
Zedner, it is "appropriate to weight the scale so that a higher level of probability is required for less serious harms, a lower level of probability for more serious harms, and a lower level still for potentially catastrophic harms." 135 Consequently, for the verdict that the offender is dangerous, the state must prove that "the person presents a significant risk of serious harm to others and the required level of risk should vary according to the seriousness of the predicted harm." 136

VII. Discussion
In the current debate on the preventive state, little attention is paid to the concept of dangerousness in criminal law. As we have shown in the previous sections, considerable work that is nowadays somewhat overlooked in the discussion about the preventive state has already been done on this concept in the past. Our account of the state of the art of legal scholarship on the concept of dangerousness also shows, however, that this concept is not yet fully developed. Rather than to put forward our own theory of dangerousness, we believe it at this point to be more fruitful to canvass five "aspects" of the concept of dangerousness. These aspects concern major differences between the concepts of dangerousness that were discussed in the preceding sections, as well as theoretical shortcomings in these concepts. Therefore, these aspects merit further consideration by legal scholarship. By parsing these aspects from our discussion of concepts of dangerousness in criminal legal theory, we aim to guide further research in this field. The first aspect is the nature of the harm. Preventive measures are put in place to prevent future harm, so the most natural aspect of dangerousness to start with is what kind of harm should be prevented by preventive measures.
One of the most striking findings of the discussion of concepts of dangerousness in the preceding sections is that the harms that ought to be prevented by criminal law are very much a product of their time. In this respect, Karen Harrison rightly observes that our concepts of dangerousness tell us at least as much about ourselves and the age that we live in as about the offenders that are considered to be dangerous, as these concepts show what types of offending we fear the most and what we place the most value on. 137 This is a relevant observation for the current debate, since it means we cannot assume that concepts of dangerousness put forward in the past reflect a society's present preoccupations with danger.
While the nature of the harms that should be managed through preventive measures is certainly influenced by the particular concerns of an historical age, there is agreement among most scholars that danger of (intentional) violent crime may warrant preventive measures. Also, with the exception of the Advisory Council for the Penal System, scholars agree that danger of property crime is not a basis for preventive measures. However, these concepts are ambiguous about how severe the violent crime should be. An Scholars who do not consider a particular preventive measure but discuss preventive measures in general tend to be unspecific about the nature of the harm that could justify preventive measures. Floud and Young, for example, focus on intentionally committed violent crime, but in the end, they seem to be indifferent towards the nature of the harm since they argue that "fear converts risk into danger." As long as comparable dangers are treated equally, 138 they take no a priori stance on the nature of the harm that could justify preventive measures; a point that Slobogin also makes with his consistency principle.
The second aspect is the probability of harm. Scholars focus on the question what probability of harm is required to find an offender dangerous. The reason for this appears to be that scholars argue that a danger only merits consideration when the probability of harm is above a certain level. If it is insufficiently certain that an offender will commit any harm, then, however serious this harm may be, it cannot justify preventive measures. From the discussion of the concepts in the preceding sections, it appears, however, that scholars struggle with the question of how probable the occurrence of harm should be. Only some attempt to answer this question. In any case, the required probability is never quantified on an interval scale from 0 to 1 but is always qualitatively described in a broad range of terms that are meant to express the level of probability that is required. Some terms used seem to suggest that a rather low probability would suffice, such as "likely to commit serious harm" (Advisory Council on the Penal System), while other expressions seem to point at a rather high probability, such as "substantial risk" (Bottoms and Brownsword). Some scholars allow for varying degrees of probability. Floud and Young, for example, put the fear of the public forward as decider whether the probability presents merely a risk, which may or may not be taken, or a danger that ought to be avoided. Slobogin as well as Ashworth and Zedner argue that the required probability should depend on the severity of the preventive measure under consideration: the less intrusive a certain preventive measure is, the lower the probability and seriousness of the harm may be, and the other way around.
The third aspect regards the question to what extent a finding of dangerousness is a factual matter and/or entails a value judgement. In this respect, scholars hold opposing views. On the one hand, some scholars, particularly those writing in the 1980s-the age of the emergence of actuarial risk assessment-tend to stress that judgements of dangerousness are to a large degree factual statements. The premise of these concepts is that dangerousness can and should be measured empirically through actuarial risk assessment. Morris and Miller as well as Walker capture dangerousness in a formula, in which dangerousness is the product of the severity and the probability of the danger. This formula suggests that the degree of dangerousness can be expressed quantitatively. Still, while arguing that assessing dangerousness

Max de Vries and Johannes Bijlsma
is an empirical matter, Morris and Miller admit that whether or not a certain degree of danger justifies preventive measures is a social and a political task. With this last step, they grant that at the end of the line a value judgement, or maybe more accurate: a policy decision, has to be made.
On the other hand, some authors stress that the concept of dangerousness in itself contains value judgements, while they differ on the extent that they allow for value judgements. The Butler Committee, for example, does not believe that the probability of the occurrence of harm can be captured in a single probability. Instead, it requires a holistic assessment of dangerousness, apparently leaving room for value judgements. Floud and Young argue that while risks are mattersof-fact, it is the fear of people that converts risks into dangers. 139 Whether or not a certain harm or a certain offender is to be feared is at least in part a value judgement, varying per person, society, andas was already discussed abovehistorical age. While Floud and Young insist that only rational fears may give rise to preventive measures, thus demanding that fears must have at least some consistency, 140 fear is still a phenomenon highly influenced by value judgements of various kinds. 141 More recent authors explicitly incorporate value judgements in their concepts of dangerousness. Slobogin's proportionality principle, as well as Ashworth and Zedner's grievous risk principle, require an evaluation of the degree of expected harm in light of the degree of liberty deprivation.
The fourth aspect is whether dangerousness refers to a personal characteristic of an offender, or to the outcome of an assessment of abstract risk factors. In this regard, there is an apparent dichotomy between actuarial and non-actuarial concepts of dangerousness. With the advent of actuarial risk assessment, it has been proposed that the individual offender has "dissolved" in preventive criminal law. While previously only specific, often mentally disordered, offenders were considered to be dangerous, actuarial risk assessment is based on a "collation of a range of abstract factors deemed liable to produce risk in general," according to Castel. 142 The former approach can be recognized in the older concepts of dangerousness, such as the Butler Committee's definition of dangerousness as the propensity of an offender to cause harm. More recently, Duff and Slobogin echo pre-actuarial concepts of dangerousness when they look for the basis of dangerousness in the personality of the offender. Therefore, as a concept, the more traditional point of view on dangerousness has not been completely overshadowed by the emergence of actuarial approaches to dangerousness.
Actuarial concepts of dangerousness do not explicitly refer to the personality of the offender. Actuarial risk assessment instruments, however, often include risk factors associated with the offender's personality, like impulsivity and mental disorder. 143 Moreover, underlying actuarial concepts of dangerousness might be an assumption that a high score on risk factors is proof of a dangerous personality. Morris and Miller, for example, while employing an actuarial concept of dangerousness, relate dangerousness to the The Elusive Concept of Dangerousness behavior of a "dangerous criminal." Conversely, Duff's concept of dangerousness as a personality trait, relies in part on the criminal career of the offender-an important actuarial predictor of reoffending. 144 Thus, the dichotomy between actuarial and non-actuarial concepts of dangerousness is perhaps not as strong as it looks.
The fifth and final aspect is the temporality of dangerousness. Ingrained in any concept of dangerousness is the possibility of future harm, as preventive measures are imposed to prevent this harm from materializing in the future. Therefore, dangerousness is always in a particular sense related to the future. However, the concepts of dangerousness that were discussed in the preceding sections show that there are different views on what role the past, the present, and the future should play. Actuarial concepts of dangerousness seem to be primarily concerned with the future, as it is the possible future harm that takes central stage in the risk assessment of the offender. Yet, the assessment of the risk of future harm is in the here and now. When actuarial risk assessment is used, proof of dangerousness may lie in the past (e.g., the criminal history of the offender) and in the present (e.g., the presence of a mental disorder).
As was discussed, some concepts regard dangerousness in part (Floud and Young; Slobogin) or in whole (Duff) as a characteristic of an offender. These concepts place dangerousness in the first place in the present, as the danger is present in the personality of the offender in quite a literal sense. Duff, in particular, states that "a judgement of dangerousness is a statement of a present condition, not the prediction of a particular result." 145 The proof of dangerousness in such concepts-in Duff's concept, for example, the persistence in offending-may lie in the past. Interestingly, Bottoms and Brownsword also use temporality in a normative sense: the danger is only "vivid"-a value judgement-when the future harm is expected to occur with a particular frequency within a certain timeframe and with a particular immediacy.

VIII. Conclusion
The concept of dangerousness in criminal law has been studied by several scholars in the past, but these past concepts do not yet present a fully developed theory of dangerousness. The five aspects of dangerousness that were discussed above show that dangerousness is a multifaceted concept to which vastly different meanings are ascribed. These aspects are therefore natural starting points for further research into dangerousness. The importance of closing this gap in the legal doctrine is obvious, as the legal establishment of dangerousness permits intrusive preventive measures, such as preventive detention for an indeterminate period of time. Accordingly, the authority of the state to exercise these powers hinges to an important extent on the concept of dangerousness that applies. Therefore, the debate on preventive justice is incomplete when discussion about the elusive concept of dangerousness is being evaded.