In psychiatry danger means that a psychiatric disorder can cause risks and complications and that people can get into trouble because of their impulsive decisions or actions. Such a danger can be, for instance, that someone resigns from his job without due consideration and without having arranged alternative income. So it does not have to be a danger in the legal sense of the word. The concept of danger in psychiatry is more vague and less narrowly demarcated. Judges approach it differently. They want to know exactly which danger is posed, how serious it is, how likely, how imminent, how acute the hazard is, and how likely it is to recur. Below I will examine more closely this legal view of danger as the problems that arise in predicting it are extraordinarily large.

A first problem, pointed out by Stone who was quoting Livermore, is statistical. The lower the incidence of a certain event in a certain population the less accurately can the cases in which it will occur be predicted. Livermore offers the following figures: when the method of prediction has a 95% accuracy rate, and, say, among 100,000 people 100 can be found to be dangerous, then 95 of these 100 people can be correctly identified and 5 will be incorrectly identified as dangerous. Of the remaining 99,900 people, however, another 5%, that is 4,995, will be wrongly identified as dangerous. This means that in order to lock up 95 dangerous people 4,995 people who are not dangerous will have to be locked up as well. In reality the situation is much more serious as dangerousness is not predictable with reasonable reliability and the error margins are probably much higher. Stone for instance also refers to research by Kozel et al regarding 31 perpetrators of sex crimes who were released against the advice of psychiatrists. Twelve of them (38%) repeated the offense. That means that the other nineteen (62%) were unjustly considered dangerous. The same research by Kozel et al reports that of 304 people released from an institution 26 (8.6%) repeated the offense. So prediction has a statistical value but no value in correctly identifying individuals. The number of “non-dangerous” people who repeat the offense is double the number of “dangerous” people who repeat it. Cocozza and Steadman who proposed some refinements in predicting dangerousness nonetheless conclude that statistically the best strategy is to assume that nobody is dangerous. Such a prediction, although not correct, is closer to correct than any other prediction, regardless on what it is based.

A second problem is that psychiatrists who have concerned themselves a great deal with prediction, particularly regarding criminals, usually posit that dangerousness can be predicted only intuitively. The only meaningful factor seems to be that the behavior has occurred several times in the past. The more often people have done something, the more likely they are to do it again. That makes predictions regarding first offenders highly speculative.

A third problem is that dangerous behavior does not come “falling out of the sky,” but as every other behavior, is determined by an unpredictable series of circumstances. These relate both to people’s personalities as their existential situations and all other sorts of factors. The examination necessary to predict danger could be one of those factors. The social context of people’s lives is so totally changed by involuntary commitment that subsequent behavior bears almost no relation to the problems and predictions which preceded the commitment. It could well be that people who would never have become violent “outside” respond with violence to involuntary hospitalization, or conversely, that people who during their incarceration are not violent would have been so if not incarcerated. Yesavage et al found that in a group of people committed involuntarily due to the threat of danger there was no more violent behavior than in a comparable group who were involuntarily committed for other reasons. This could mean that a. there is in fact no difference in the two groups in respect of the chances of presenting dangerous behavior; b. the contextual factors are primarily determinate; c. the procedure that is followed is of decisive influence; or d. treatment was so effective that the feared behavior did not present itself. Rofman et al conducted similar research, this time with a control group of nearly all voluntary patients. They found that during the first 10 days there was significantly more aggressive behavior among involuntary patients committed for dangerousness. Here too it is unclear what exactly is being measured. Circumstances in society and on the mental hospital ward are extremely different. Besides, violence could have been induced by the involuntary commitment itself.

A fourth problem is that it is fairly impossible to ascertain whether the prediction is valid. When people are committed on grounds of a prediction their behavior changes so drastically that it is doubtful that subsequently displayed behavior has any value as feedback for the predictor. However, when people who are not committed subsequently display the feared behavior, the presumption is that the evaluator was wrong. This implies that an evaluator cannot be found wrong when involuntarily committing someone. If the committed person later displays aggressive behavior that is (possibly wrongly) viewed as a confirmation that the evaluation was correct. If the person does not display such behavior the change of environment or efficacy of treatment is credited. However when that person is not committed and later displays dangerous behavior such is counted as a failure on the part of the evaluator. In particular when the events take on a dramatic form and become front page news, the evaluator is faced with a most difficult confrontation with his own apparent failure. This state of affairs naturaly evokes a constant urge to “play it safe,” to not take risks, and therefore involuntarily commit more people than necessary. High reliability can never be expected from predictive procedures for which there is no feedback.

A fifth problem is that psychiatrists have developed their concepts for the purpose of intervening with treatment. When a psychiatrist is asked to predict someone’s dangerousness he cannot do so without observing the person’s psychological condition and the presence of psychiatric problems. Rubin considers the notion that certain psychiatric disorders are associated with danger incorrect. According to him psychiatric diagnoses have no predictive value in respect of dangerous behavior. Psychiatrists are preoccupied with treatment. No doubt they allow their conclusions regarding dangerousness to be influenced by their opinion on the desirability of treating the patient.

A sixth problem related to this is that psychiatric examinations and the “clinical eye” are inefficient ways of approaching the prediction of danger. Psychiatric examination neither was developed for that purpose nor is it suited to it. In a follow-up examination of 17 people who were considered insane while committing major crimes Rubin found that repetition of offenses was mainly connected to social factors. He calls the notion that psychiatric evaluation of individuals can reliably predict danger a myth and points out that it is wrong to consider impulses and actions interchangeable. He considers it a mistake to assume that certain psychiatric disorders are in themselves dangerous. Here a problem discussed in 1.4.1 returns. The classification of psychiatric disorders, intended for indicating treatment, is unsuitable for serving as a prediction of danger.

The fate of the so-called Baxstrom patients is illustrative of these problems. In 1966 The United States Supreme Court ruled that 650 people incarcerated in “maximum security” clinics for the criminally insane were to be transferred to “ordinary” psychiatric hospitals. All 650 had remained in detention after expiration of their sentence as they were considered too dangerous to be released. After four years only 20% of these people were reported to have displayed aggressive behavior, whether inside a psychiatric hospital or outside of it. This implies that 80% no longer displayed dangerous behavior.

Additional conclusions can be derived from the above and from research referred to by Stone and Robitscher.

The matter of how high the risk is that someone will in the future display dangerous behavior is so complex that a reasonably accurate evaluation is fairly impossible. Inasmuch as it is possible to research the likelihood of dangerous behavior psychiatrists have been found to be no better at predicting it than others. In fact, neither psychiatrists nor other professionals can do so reliably. For each correct prediction there are always several incorrect ones. In short, future dangerous behavior is not predictable. Even regarding repeat criminal offenders prediction is inaccurate. An even remotely accurate prediction is impossible regarding psychiatric patients who have never violated any law nor proven to be dangerous.

In “normal” criminal justice cases great care and accuracy is taken to determine whether people have actually performed the acts of which they are accused. Psychiatric patients are routinely locked up because of an off chance that they might in the future become dangerous. The discrepancy between the aspired levels of certainty for these two types of detention is so bewilderingly great as to evoke the impression that from a legal viewpoint having a psychiatric disorder renders a person fair game. Ellis and Robitscher among others have pointed out the dire social consequences of involuntary commitment to people so committed. The grounds on which such decisions are made, inasmuch as can be investigated, are strictly inadequate for making such an invasive decision.

Possibly the presumed dangerousness ascribed to psychiatric patients by many authors contributes to that. Snowdon reveals that dangerousness is much more common among non-psychotics than among psychotics. He voices the fear that the criterion of dangerousness makes involuntary commitment for psychotics impossible. Melick et al note that several reviews of frequency of arrests of ex-psychiatric patients before 1965 led to the conclusion that arrests among this population were less frequent than in the general population. Reviews after 1965 reveal a gradual increase in arrests. This difference is explained by assuming that criminal behavior was gradually becoming more “medicalized” causing more people with criminal behaviors to wind up in psychiatry. This would mean not that psychiatric patients are more dangerous than other people but the opposite, that in the last decades dangerous people are ever more being considered psychiatric patients. I might add that I consider the conclusion by Melick et al reversible. A diametrically opposite interpretation is possible as well. It could be that the small number of arrests before 1965 was due to ex-psychiatric patients being recommitted instead of arrested. The rise in arrests could then be explained by the “criminalization” of psychiatric disorders after 1965. This could be related to the large-scale closure and reduction of the Mental Hospitals in the United States after 1965, and the inadequacy of alternative facilities. Furthermore, experience in the United States shows that when involuntary commitment is made more difficult psychiatric disorders are proportionately “criminalized.” People are then no longer eligible for commitment so when arrested for minor infringements they wind up in jail. Be that as it may, it is clear that both people with psychiatric disorders are to be found in jails and people with criminal behavior are to be found in psychiatric hospitals.

The Dutch law is aimed at making involuntary commitment a legal matter. Accordingly psychiatrists are expected to estimate future dangerousness. Psychiatrists, however, cannot predict that, unless threatening behavior is already concretely, directly, and immediately manifest, such as when someone is angrily swinging an ax and shouting that he will murder his wife. Yet even then a concrete prediction of what will happen remains difficult.

De Winter’s view that every psychotic patient can be considered dangerous to himself, others, and the public order seems to me not only factually wrong but also dangerous in the sense that the legislator may labor under the illusion of having made an efficient law when in fact that is not the case. De Winter’s proposal would be an example of incorrectly enforcing a law. Asking about future dangerousness is not only pointless, as there can be no answer, but also poses an important ethical dilemma for psychiatrists. That dilemma is whether judging possible risks to third parties can and may be counted as one of their duties and whether such judgments may be used against their patients. Citizens’ safety and maintaining public order in the community belongs in the realm of the police and the courts. Is it justified to expect psychiatrists to take this task upon themselves? It seems to me a 180º turn in their actual obligation: helping as well as possible people who ask for help because of an illness. Even when psychiatrists function not as therapists but solely as evaluators it remains to me questionable whether it is justifiable that they are asked not only about psychiatric diagnoses but also, more or less based on those diagnoses, whether they regard the patient as dangerous. Not only are psychiatrists incapable of such judgments, other than statistically, but they are compelled to become the adversaries of the people being judged and “accomplices” of the judicial system. If psychiatrists must be accomplices let them be the patients’ accomplices also in their evaluating role. It is up to judges to make pronouncements about danger and its seriousness. Peszke pointed out the altogether unmedical nature of the job imposed on psychiatrists. Stone and later Robitscher supported him in this view.

Likewise Cohen Stuart pointed out psychiatrists’ conflict of interest evoked by the Dutch commitment laws. The problem is not only that psychiatrists are expected to perform the impossible task of determining danger. The law also expects the treating psychiatrists to signify the point at which the danger has abated to the point that patients may be given their freedom even though their psychiatric disorders remain. Cohen Stuart is right in pointing out that both roles, that of therapist and that of evaluator of danger, cannot be fulfilled by one and the same person. In the one role the psychiatrist is the patient’s adversary, while in the other, his ally. After all, psychiatrists and patients are expected to set up the treatment plan together. In Chapter VI, 5, I mentioned that the separation of treatment and regulation in the Netherlands is a valuable tradition. The combination of therapist and evaluator of danger is even less admissible. Psychiatrists are thus compelled to make decisions from the judge’s point of view, determining whether freeing the patient serves the interests of society. When psychiatrists accept this dual role, which from a medical-ethical viewpoint is utterly inadmissible, they become officers of social control, which not only corrupts their role as therapists but also will confront them with constant failure in both of these mutually exclusive roles.

The law has attempted to avoid the problem of unpredictability regarding danger by posing that the danger must already be manifest in the person’s actions. The question is whether the solution is not worse than the problem.

In the first place is the insoluble problem that it is unknown what must be regarded as the manifestation of danger. Is it an argument? Making threats? A slap? Or must actual harm be done? How is behavior deriving from psychiatric disorders to be distinguished from that which is not? After all, “No psychiatric disorder exists which autonomously and predictably leads to direct danger.”

Secondly, if the respective manifestation poses immediate danger, then as a rule a crime will have been committed. Threatening violence or putting others in danger is a criminal act. Criminal law does not wait for the harm to have actually transpired. That would mean that criminal behavior is “psychiatrized.” Stone pointed out the shifting roles of the massive institutions of psychiatry, justice, and welfare. “What has happened in the last two decades is that in the name of reform, the professionals within each of these social institutions have taken on the roles, functions, and goals of each other.” Luckey and Berman wrote about a change in the law regarding danger in Nebraska stipulating that certain people who used to be tried according to criminal law would now be involuntarily committed. This would mean that certain forms of criminal behavior would cause a person to be involuntarily committed while severely disabling psychiatric disorders, in particular many psychoses in which there is not a clear threat of danger, would not meet the criteria for involuntary commitment.

Thirdly, it seems to me justifiable to fear that soon we will no longer be able to distinguish between a criminal act of posing danger and the manifestation of a psychiatric disorder. This distinction depends on the decision which interpretation of events is more valid. The choice will sometimes be in one direction and sometimes in the other. As what happens to people once they are channeled into psychiatry seems to be very much determined by legal procedures and rulings anyway, psychiatry will be “criminalized.” This risk is confirmed by Cocozza and Steadman’s findings that identical behaviors led to some ex-psychiatric patients being recommitted while others were arrested. Neither civil rights nor psychiatric clarity will be served when people displaying manifestations of danger are shifted arbitrarily whether in the direction of the criminal justice system or in the direction of psychiatry.

In summary, it is extraordinarily difficult to predict danger. Such predictions cannot or almost not be extrapolated from actual behavior. No criteria are known by which dangerous behavior can be predicted other than that such behavior has already been repeatedly displayed. Predicting danger with reasonable reliability is impossible.