Forensic Use of Actuarial Risk Assessment with Sex Offenders: Accuracy, Admissibility and Accountability

Abstract

The decade of the nineties ushered in an unprecedented number of state and federal laws intended to manage sexual offenders. Among the most controversial are the so-called sexually violent person (hereinafter SVP) laws - schemes that use civil commitment to supplement criminal sentences in order to incapacitate the most dangerous sex offenders. Risk assessment - the prediction of sexual recidivism - is essential to this legislative agenda. As a result, the demand for specialized risk assessments has been rapidly growing, and has produced a cottage industry of forensic psychologists and vigorous development of actuarial and other structured approaches to supplement the traditional clinical assessment of this risk.

These modern legislative initiatives for the management of sexual offenders have generated heated controversy, and their reliance on the assessment of risk has, in important ways, been central to the controversy. An initial wave of litigation in the mid-1990's questioned the constitutional foundations of this new legislative agenda. By calling attention to the inherently problematic task of assessing future risk, these challenges questioned the constitutionality of locking up people for future crimes that they might - or might not - commit. The courts were unimpressed, and confirmed the basic constitutionality of risk-based deprivations of liberty, despite well-known shortcomings in the prediction of dangerousness.

A second wave of litigation is now focused on a more detailed articulation of the standards for assessing risk in these legal settings. Courts are asking in more detail what the legal standards for dangerousness mean, and what sort of evidence is legally available, and ought to be required, to prove those standards. In large measure, the legal challenges in these more recent cases have addressed the use of actuarially-derived (as distinct from clinical) risk assessments, and it is this subject that we examine in this article. Clinical judgments of dangerousness - judgments that ultimately rest on the ipse dixit of a mental health professional - are a routine part of the judicial landscape. In contrast, actuarial risk assessment (hereinafter ARA) - which employs empirically derived mechanical rules for combining information to produce a quantitative estimate of risk - is novel in the legal arena and seems to be setting off a variety of alarm bells. Critics of ARA have focused their objections on the admissibility of ARA-derived expert testimony. Pointing to a variety of shortcomings, they argue that the relatively new ARA techniques do not merit admissibility under prevailing Frye or Daubert standards. Such challenges have met with mixed success.

In this paper we explore the forensic use of actuarial risk assessment testimony, particularly in the context of SVP laws. Our thesis is straightforward: actuarial methods have proven equal or superior to clinical judgments. Given the legislative mandate to assess risk, as well as the routine, widespread use of clinical assessments of risk in the judicial system, it is logically incoherent to exclude evidence that presumptively improves upon the reliability and accuracy of these judgments.

But the issues raised are larger than those of admissibility. Like any tool of science, ARA can work for good or ill. Used well, ARA can ameliorate some of the concerns about justice, efficacy, and public policy that swirl around SVP laws and the other recent legislative initiatives. In fact, ARA is a state-of-the-art technique, and courts should insist that it be employed as a major instrument of risk assessment. Used poorly, however, ARA can exacerbate concerns about justice and utility. Further, the use of ARA might have unanticipated and undesired consequences for broader areas of public policy. Improved ability to identify persons at high risk for violence may make expanded preventive detention laws politically impossible to resist. New laws, in turn, may demand better risk assessment, which may beget more aggressive and expansive prevention laws, and so on. In view of these negatives, we urge caution and mindfulness in using ARA.

Our argument for the admissibility of ARA must be understood in the context of two important preliminary points. First, as we will more fully discuss later, the development of ARA, like all good science, is evolutionary. The sophistication of ARA evolves over time as more is learned about the task of assessing sex offender risk and the functioning of particular ARA scales. ARA scales differ in their reliability and accuracy. Hence, they should not be considered equivalent and thus interchangeable. Like all products of science, they are works in various stages of development.

Second, it is imperative that ARA be used properly. This second issue raises several legitimate concerns; among the most critical is the proper interpretation of ARA information. To the extent that courts seek to measure the long-term, presumptively stable risk posed by individuals, ARA provides the most accurate information. But courts ought to be concerned as well with how risk can be managed and modified in the short- and medium-terms, through intervention such as treatment and community supervision. This domain, generally referred to as dynamic risk assessment, represents the most recent entrée to the scientific literature and will likely be the focus of attention among scientists for the foreseeable future. Given its focus on long-term risk, however, ARA is of less direct relevance, at least given the current state of the art. Therefore, on these important questions of risk management and modification, courts may, for the time being, need to rely more heavily on carefully done clinical assessments, though it is likely that dynamic ARA will eventually complement these assessments as well.

In urging the use of ARA, we do so against the backdrop of existing SVP laws and serious concerns raised about their constitutionality and wisdom. Although our paper is premised on findings that ARA is superior to clinical assessment methods, we do not take the position that either is sufficient to justify the massive and long-term deprivation of liberty inherent in SVP laws. Our point is simply that if courts deprive people of liberty based on assessed risk, then ARA should be part of the assessment. Courts should use ARA in part because it will improve risk assessment. But more importantly, from our perspective, ARA brings a transparency that will allow for a clearer understanding of the true nature of risk assessment, including its significant limits and potential for misuse.

This paper proceeds as follows. We first set the context by briefly describing SVP laws, highlighting two salient features: their lack of clear standards for confinement, and their extraordinary cost. Second, we discuss notions of risk assessment and predictions of dangerousness, and outline the key legal concepts that control their use. Third, we introduce the distinctions between clinical and actuarial methods of risk assessment, and summarize the empirical basis for the claim that ARA is generally superior to clinical assessment of risk. Fourth, we discuss the treatment of ARA under prevailing standards for admissibility, the Frye and Daubert tests, arguing that SVP courts should admit ARA testimony. Finally, we propose a set of guidelines for courts to use to minimize the potential for misuse and prejudice, and maximize the beneficial consequences, in connection with ARA testimony. In closing, we argue that performing risk assessments without ARA is unethical for mental health professionals and improper for courts. But we warn that increasingly accurate methods of risk assessment may encourage the expansion of civil-commitment-style violent person laws, an approach to prevention that, in our view, is questionable both morally and practically.