Following up on last week’s research review, here are some new articles from the ever-controversial practice niche of sexually violent predator cases:
Facts? Who cares about the facts?!
Once a jury is empaneled to decide whether someone with a prior sex offense conviction is so dangerous to the public that he should be civilly detained, the verdict is a foregone conclusion. Dangerousness is presumed based on the prior conviction, rather than having to be proven.
Researchers Nicholas Scurich and Daniel Krauss confirmed this by giving jury-eligible citizens varying degrees of information in a Sexually Violent Predator (SVP) case and asking them to vote. Some mock jurors were told only that the person had a prior conviction for a sex offense. Others were also given information that the person had a mental abnormality that made him likely to engage in future acts of sexual aggression.
It mattered not a whit. The mock jurors voted to civilly commit at the same rate, whether or not they had heard evidence of current dangerousness.
“The mere fact that a respondent had been referred for an SVP proceeding was sufficient for a majority of participants to authorize commitment,’ the researchers found. “These findings raise concerns about whether the constitutionally required due process occurs in SVP commitment proceedings.”
No surprise, really. In this practice niche more than others, fear and hype often overshadow reason. Sex offenders are not the most appealing human beings, and no one wants to shoulder the responsibility of voting to release someone who could go out and rape or molest again.
The study is:
The presumption of dangerousness in sexually violent predator commitment proceedings, Nicholas Scurich and Daniel A. Krauss, Law, Probability and Risk. A copy may be requested from the first author (HERE).
Sexual disorder diagnoses not reliable
Meanwhile, even when jurors do hear evidence of mental abnormality, it is not especially accurate.
Examining the diagnoses given to 375 sex offenders referred for civil commitment in New Jersey, researchers found “questionable” diagnostic reliability to be a widespread problem across the range of clinicians.
Pedophilia was the only diagnosis in which two evaluators were likely to agree at a level above chance. The rates of agreement were far worse for other disorders that are typically rendered in SVP cases, including “Paraphilia Not Otherwise Specified,” Sexual Sadism, Antisocial Personality Disorder and Exhibitionism. In fact, among the six cases in which Exhibitionism was diagnosed, there was not a single case in which both clinicians agreed.
The study, by Anthony Perillo of John Jay College and colleagues, adds to a burgeoning body of literature (some of which I’ve previously reported on) suggesting that psychiatric diagnoses in SVP evaluations are often dubious and not to be trusted.
The article is:
Examining the scope of questionable diagnostic reliability in Sexually Violent Predator(SVP) evaluations, Anthony D. Perillo, Ashley H. Spada, Cynthia Calkins and Elizabeth L. Jeglic, International Journal of Law and Psychiatry. A copy may be requested from the first author (HERE).
Race bias in actuarial risk prediction
Okay, so the diagnoses aren’t reliable. But we’ve still got another tool of science up our sleeves -- actuarial risk assessment.
Not so fast.
As I’ve previously reported, the predictive accuracy of actuarial risk assessment tools is pretty wimpy. And now, researchers from Sam Houston State University are finding that the most widely used actuarial tool, the Static-99, doesn’t work at all with Latino offenders.
The findings are based on research with a large sample of about 2,000 sex offenders, almost 600 of whom were Latino.
“Findings have implications for fairness in testing and highlight the need for continuedresearch regarding the potentially moderating role of offender race/ethnicity in risk research,” note researchers Jorge Varela and colleagues.
The study is:
Do the Static-99 and Static-99R Perform Similarly for White, Black, and Latino Sexual Offenders? Jorge G. Varela , Marcus T. Boccaccini, Daniel C. Murrie, Jennifer D. Caperton and Ernie Gonzalez Jr. International Journal of Forensic Mental Health. To request a copy from the first author, click HERE.
How to lie with statistics: “The Area Under the Curve”
Listen to any defender of actuarial risk prediction for a few minutes, and you will likely hear "Receiver Operating Characteristics” and “The Area Under the Curve” touted as indicators of statistical accuracy.
But in a new study in the Journal of Threat Assessment, two European scholars argue that these arguments are “fundamentally misleading.” Using the Risk Matrix 2000 instrument -- widely deployed in the United Kingdom -- as an exemplar, they found that a prediction of reoffense for an offender who scored in the “Very High Risk” range will be wrong an astounding 93 percent of the time.
“The numbers necessary to detain in order to prevent one instance of recidivism are large,” write David Cooke and Christine Michie. “On further reflection, from a statistical rather than a psychological perspective, should we be surprised? It has long been recognized that low-frequency events are hard to predict.”
The authors argue that the weak performance of actuarials is being systematically camouflaged by “statistical rituals” that are confusing and non-transparent, raising fundamental questions of fairness in legal decision-making.
The article is:
The Generalizability of the Risk Matrix 2000: On Model Shrinkage and the Misinterpretation of the Area Under the Curve. David Cooke and Christine Michie. Journal of Threat Assessment and Management. To request a copy from the first author, click HERE.
Not everyone agrees with Cooke and Michie’s analysis. One detractor is Douglas Mossman, of the Department of Psychiatry at the University of Cincinnati College of Medicine. Using a fictional scenario, he attempts to illustrate how "group data have an obvious application to individual decisions.” His paper goes on to argue that “misinterpretations of mathematical concepts and misunderstanding of the aims of risk assessment have led to mistakes about the applicability of group data to individual instances.”
The paper is:
From Group Data to Useful Probabilities: The Relevance of Actuarial Risk Assessment in Individual Instances. (Unpublished.) Douglas Mossman. Paper available online (HERE).
Who is minding the store?
If nothing else, the above research snippets demonstrate the high level of controversy and complexity in the implementation of Sexually Violent Predator laws. If psychologists -- who must master psychometrics and statistics in order to earn our PhD’s -- have a hard time with these concepts, imagine how difficult it is for attorneys. With people’s lives at stake, do they have the knowledge base necessary to avoid being hoodwinked, and to educate jurors and judges?
In a new paper, prolific legal scholars Heather Cucolo and Michael L. Perlin of the New York School of Law argue that more stringent standards for representation are necessary for effective assistance of counsel in SVP cases.
They propose that counsel should be required to “demonstrate a familiarity with the psychometric tests regularly employed at such hearings, and with relevant expert witnesses who could assist in the representation of the client.” Furthermore, they argue for a pool of court-appointed experts who could be appointed at no cost, similar to those provided in insanity cases.
“There is no question that the population in question is the most despised group of individuals in the nation. Society’s general revulsion towards this population is shared by judges, jurors and lawyers. Although the bar pays lip service to the bromide that counsel is available for all, no matter how unpopular the cause, the reality is that there are few volunteers for the job of representing these individuals, and that the public's enmity has a chilling effect on the vigorous of representation in this area.”
The paper is: