|Sponsored by the HealthcareTrainingInstitute.org providing Quality Education since 1979|
Documents such as the DE/DJ Guide themselves may be responsible for their own misapplication. After briefly stating the previously quoted caution, the DE/DJ Guide goes on to list the various purported early warning signs with little to no differentiation among them. Specifically, the DE/DJ Guide (and others like it) fail to discriminate among the indicators in terms of two critical dimensions. First, there is no differentiation among the indicators regarding the widely disparate base rates in the general adolescent population. Second, there is little reference to differential validity of the various indicators; in other words, there is no evaluation of the extent to which each purported warning sign has been shown to be correlated with subsequent violent behavior. The net result of promulgating these indicators as effective ways to identify potentially violent youth, but failing to discriminate among the signs along these dimensions, is an unexamined but theoretically predictable elevation of decision errors. The following sections deal with the effects of differential base rate and differential validity, respectively.
Differential Validity: Warning Signs Versus Low-Level Violence: The validity of the early warning signs in the DE/DJ Guide in relation to extreme violence is patently unknown. The extent to which each individual warning sign listed in the DE/DJ Guide is associated with extreme violent actions is not clear. Stated differently, it is unknown what proportion of all people who show the indicator will later perpetrate extreme violence; nor is it known what proportion of all adolescents who have exhibited extreme violence have previously exhibited the indicators. The latter proportion (of perpetrators showing the signs) is likely far higher than the former (proportion of positive indicatees later perpetrating); this difference logically accounts for what behaviors or traits have been chosen as potential early warning signs. However, this retrospective logic requires prospective validation with acceptable positive and negative predictive power.
Differential Utility Of True Positives Versus True Negatives: In most prediction tasks, minimizing prediction error (thereby maximizing accuracy) is die only goal. Correct classification is implicitly assumed to have inherent utility. But as Finn (1982) noted, this assumption is often suspect; moreover, the utility of true positive and true negative predictions must inform the extent to which we tolerate the different types of prediction errors.
True Positives: Take the extreme example of a diagnostic prediction. A test is used to diagnose persons as having or not having a potentially fatal disease. There is a treatment for the disease, but it is effective in only a small percentage of cases and has side effects (occasionally lethal) for the majority of cases. Thus, the utility of a true positive is quite low. In blatant terms, the mortality rate with a correct diagnosis may be only slightly lower than it would have been if no diagnosis had been attempted. Just as importantly, some "innocents" (persons who never actually had the disease; false positives) are seriously harmed by the diagnosis and subsequent intervention. In the case of predicting violence in adolescents, the utility of a true positive should be seriously questioned. In fact, many of the publicized school shootings reviewed previously had perpetrators who had previously been identified as requiring behavioral intervention. For example, both perpetrators from the recent Littleton shooting were in the juvenile justice system (reportedly receiving "good" prognoses from their program officers; "Anatomy of a massacre," May 3, 1999). Likewise, as previously mentioned, the Springfield shooter had been arrested for stealing a gun the day before his shooting spree. Short of preemptive, long-term confinement (which, given the error rates discussed previously, would be unconscionable), there are no known interventions for potentially violent persons that yield anything close to high efficacy. In fact, even when adolescents are "known" to be violent (rather than merely suspected as "potentially violent") and assertive efforts are applied toward correction, treatment, and supervision, the success rate estimates are as low as 5% (Greenwood, 1994; Lab & Whitehead, 1988). Much like the diagnostic example described previously, the violence perpetration rate is likely only slightly lower in the context of attempted early identification than it would have been otherwise; thus, there is some minimal benefit to potential victims. However, many innocents are harmed via the prediction errors and subsequent stigmatization and ostracization. The societal impact of these competing benefits constitutes the value judgment being explicated.
True Negatives: Adolescence is a time of turmoil and disequilibrium — one that has been lengthening since the onset of the industrial revolution. The adolescent's world is in a rapid and constant state of flux. Adults often communicate radically mixed messages regarding the legitimacy of adolescents' status as emerging adults, leaving teenagers with the sense that they have the responsibilities of adulthood (judgment, learning, earning, little play time) but the limited rights of a child. Adolescents need consistent and positive attention from important adults (parents, teachers, coaches, etc.), even when they might seem to be eschewing that attention in favor of peer interaction; this is the means by which adolescents are welcomed and socialized into the world of adulthood. In the current context of searching and scanning the student body (as well as the students' bodies) for signs of potential violence, there is a real danger on the part of well-meaning adults of failing to attend to the crucial needs of the vast majority of adolescents (including but not limited to those who will never become violent). There is questionable utility (for the individual and for society) in categorizing an adolescent as "not a risk" and thereafter focusing attention elsewhere.
Prediction Versus Prejudice: Authoritarianism In The Guise Of Safety
Treatment of Self-Identified Troubled Youth: Efforts are often directed toward identifying and treating youths who do not view themselves as in need of intervention. However, many adolescents do acknowledge troubled emotional and social experiences and present themselves to school counselors, teachers, physicians, and other adult authority figures for help. Professionals who work with adolescents must be ready and willing either to provide requested help or make appropriate referrals to other professionals. Morley and Rossman (1996) described an integrative school- and community-based program intended to orient educators and other adults toward assessing and providing for the needs of children and adolescents who present in distress. Training teachers, school administrators, general practice physicians, police officers, juvenile corrections officers, and other noncounselor professionals in how to address the needs of self-identified troubled adolescents would be a far more productive expenditure of resources than would training them to identify subtle (and likely invalid) violence warning signs in otherwise functional youth.
Treatment of Already Identified Troubled Youth: It has become nearly tautological to assert that the best predictor of future behavior is past behavior (cf. Rich, 1992, in reference to violent crime). This is particularly the case when the physical and social circumstances surrounding the individual are largely unchanged across time (Silver, Mulvey, & Monahan, 1999). When adolescents commit identifiable acts of violence (even at a low level) and enter into the juvenile justice system or their schools disciplinary-intervention programs (often termed alternative education programs), they are identified as troubled and in need of help. With these adolescents, it is unnecessary to search for subtle warning signs that they might be at risk; they are clearly at risk of future violent behavior based on their past perpetration. There is a growing literature on the prediction of violence among those who have previously been violent (e.g., Gardner, Lidz, Mulvey, & Shaw, 1996; Monahan, 1996); but the applicability of these technologies cannot be seen to generalize to persons who have never been violent. Beyond risk assessment, there is a need to rehabilitate violent youths and prevent future violence. How effective are juvenile justice systems and specialized educational programs at preventing these clearly at risk adolescents from perpetrating mass violence? Although there are some existing estimates of general recidivism (e.g., Greenwood, 1994; Lab & Whitehead, 1988; Tate, Reppucci, & Mulvey, 1995), the base rate problem precludes precise assessment with respect to extreme violence. However, there are at least two glaring examples that call into question the extent to which such programs can prevent mass violence: the two young men who killed students, teachers, and themselves in Littleton, Colorado in 1999. These two boys had been placed in a juvenile justice diversion program for various theft charges. One of the boys had been reported to law enforcement officials on several occasions for threatening the life of a peer. As mentioned previously, they appeared to be complying with the disciplinary activities required of them vis à vis the diversion program. Obviously, simply knowing that the two youngsters were at risk did not lead to effective prevention of their eventual violent acts. Pointing this out is not meant as an indictment of the Littleton juvenile justice programs or any other person or system attempting to treat already identified troubled youth. This example is provided to illustrate the need to focus on improving our treatment technologies for those we know are at risk, rather than filling our intervention programs with false positives identified via subtle, nonviolent, purported indicators of risk.
Treatment of Adolescenthood: As discussed in the True Negatives section, adolescence — even in the best of cases — is a developmental stage fraught with anxiety and tension. Although the physical maturation and information availability of most midadolescents (15-17 years old) approximates that of adults, there are biophysiological factors (e.g., forebrain myelination; cf. Kalat, 1998, p. 436) and social and experience-related factors (e.g., relationship and occupational attempts) that prevent adult-level judgment for another 8-10 years. The net subjective result for the adolescent is a world of extremes. Any kind of violent act is imaginable (via today ' s media technology) and accessible (given the availability of weaponry and other dangerous items and information), but adults demand a very narrow range of expression of the adolescent's own feelings of anger. Any kind of physical experience is imaginable and accessible (given the availability of illicit drugs; virtual reality; body piercing; and fast cars, games, and amusement park rides), but adults expect experimentation with most of these experiences to be laid aside in favor of "Just Say 'No'" mantras. Any kind of sexual experience is not only imaginable but publicly fantasized about (via glamor magazines, popular music, film and television, and styles of dress, not to mention available pornography), but adults make "abstinence" the most important concept in what passes for sex education. Society must determine what experiences and behaviors to make available and acceptable to citizens of all developmental levels. However, adolescence merits reevaluation in terms of the dichotomies currently existing between availability and acceptability. As much as we might benefit from determining how best to treat troubled adolescents, we are in need of finding better ways to treat adolescence.
- Sewell, Kenneth W.; Mendelsohn, Michaela; Profiling Potentially Violent Youth: Statistical and Conceptual Problems; Children's Services: Social Policy, Research & Practice, 2000, Vol. 3 Issue 3, p147-169
Reflection Exercise #12
Others who bought this School Shootings Course