Article by Judge Daniel Conviser - Re: Registration

"It's Past Time to Reform NY's Sex Offender Risk Assessment System," Part I, by Daniel Conviser [Manhattan Judge], New York Law Journal, 1/26/21

"There is little evidence...that the RAI [Risk Assessment Instrument] has any predictive validity. The instrument was created in January 1996...by the five state employees who comprised the Board [of Examiners of Sex Offenders], and even at the time was not based on actuarial data correlated to reoffense risk. It mixes risk predictions with 'harm' value judgments and other policy considerations to create numerical rankings which have no direct relationship with either the modern science of sex offender risk assessment or the implicit value judgments about the relative harms caused by various kinds of sex crimes made by New York's Penal Law. Judges, moreover, given the absence of expert risk evidence at the vast majority of SORA hearings, lack the appropriate tools to make valid risk decisions."

"[T]he risk determinations of the RAI are arbitrary. I am not the only person or organization who thinks so. Many of the issues discussed here were also outlined in great detail in a report reissued earlier ths year by three Committees of the New York Bar Association...These concerns are also routinely raised during SORA hearings and are well understood by professionals in the field. Yet, despite the well-established flaws of this sytem, nothing about it has changed in 25 years."

"For 12 years, I have...listened to hundreds of hours of expert testimony by psychologists and psychiatrists about sex offender risk assessment and have learned how valid sex offender risk assessments are made.

"The outlines of how such assessments are created are simple. An expert psychologist or psychiatrist...reviews every piece of information about the offender which can be found and conducts an extended interview with him, if possible. The expert may administer tests and interview collateral contacts, like an offender's relatives or sex offender treatment providers. When determining risk, evaluators also score or review 'actuarial risk assessment instruments.' These are conceptually no different than the actuarial tables used by insurance compaies to underwrite insurance policies.

"Decades of research on tens of thousands of convicted sex offenders throughout the world have determined the factors which are correlated to offenders who reoffend and those who don't."

"The RAI looks like an actuarial risk assessment instrument because it is a scale with numbers. But it isn't. It wasn't designed to correlate to risk data. Rather, when it was created in 1996, the five state employees who developed it (members of the Board in consultation with some outside 'experts') picked things which they felt were especially egregious or had some relationship to risk, some of which were required under the SORA statute, and combined them into a numerical scale with 15 categories they made up."

"Some of the factors on the scale are about harm, some are about risk, some are apparently about both, some are based on other policy goals, and the rationale for some are simply unclear. Thus, under the first risk factor, an offender gets 10 points if forcible compulsion is used in a second offense, 15 points if physical injury is caused, 30 points if the offender is armed with a 'dangerous instrument.' These are obviously all significant aggravating facts about a sex offense.

"But the specific manner in which the RAI defines and grades these characteristics has no relationship to the science of risk assessment.

"An offender with a prior nonviolent misdemeanor conviction for a drug crime gets 5 extra points. But, if it is a felony, 15 points are added. If an offender touches a victim's sexual or intimate parts over clothing, 5 points are scored. But if the contact is under the clothing 10 points are scored. None of these distinctions are supported by any research on sex offender recidivism."

"The precise numerical scores under the RAI have no correlation to risk data. Each factor is scored with 5, 10, 15, 20, 25 or 30 points."

"Nor do the scoring ranges have any appropriate rationale. Scores from 0 to 70 are low risk; from 75 to 105 are moderate risk and from 110 to 300 are high risk. Thus the 'low' risk category comprises 23% of the scoring range; the 'moderate' risk category 10% and the 'high risk' category 63%. No one has ever attempted to explain why. Roughly 25% of SORA offenders are classified as 'high risk' even though other measures indicate the actual percentage of high risk offenders in New York is far lower. See City Bar Report, p. 7."

[Given that there are three risk levels, one might have expected that the lowest one-third of the score range would be 'low'-risk, one-third would be 'medium,' and one-third would be 'high.' But that's not the case. While the 'low' share of 23% comes close, the 'high' proportion is a whopping 63%, roughly double what would have been expected. (And yet, only about 25% of offenders actually get assigned the highest risk level, which is believed to still be far more high-risk offenders than actually exist.) All of this is yet more evidence that this so-called 'Risk-Assessment Instrument' has little to no construct validity.]

"It would be easy to determine how predictive the RAI actually is in predicting sex offender recidivism...All of the data is readily available. Each sex offender's risk level could be put into a computerized program. Then, data about which offenders were arrested or convicted of additional sex offenses over time could be compared. Matching the data would tell us how predictive SORA risk level determinations have been. No such analysis has ever been published by the State.

"In 2011, Dr. Jeffrey C. Sandler of the State University of New York at Albany conducted a statistical study of sex offender outcomes under SORA but his work is available only through a slide show. The City Bar Report summarized his analysis as finding 'that an individual's score on the RAI marginally corresponds to their risk of re-offending, but that a better-designed risk assessment instrument would much more accurately predict an individual's risk.' The report also said that Dr. Sandler's analysis 'concluded that the RAI does not significantly predict the harm caused by an individual's reoffending...' City Bar Report, pp. 7-8."

"The RAI does not consider most of the significant actuarial factors which do correlate to risk; and for the issues it touches on, considers these in a way which does not reflect the data."

--- Part II, NYLJ 2/10/21

"[I]n the absence of reliable evidence courts are ill-suited to make risk predictions. SORA determinations are rarely based on expert evidence or testimony by qualified psychologists or psychiatrists who have assessed an offender's risk. Judges are rather asked to make judgments about risk by looking at basic facts about an offender and his offenses, reviewing a couple of pages of information from the Board and a proposed RAI score and hearing arguments from lawyers who are not experts in sex offender risk issues. Assessments about risk, however, are not intuitive or moral decisions. They are not based on whether the hair on the back of your neck stands up when you review an offender's crimes. They are empirical assessments; predictions about how likely it is that future behaviors will occur. They are predictions which should be informed by the decades of studies and science which have been published about the factors related to reoffending...But under our system, they rarely are."

"Consider age at an offender's release, perhaps the most robust of all actuarial risk factors...[R]e-offense risk begins to decline for healthy 40-year-old men.

"Case law, however, has consistently upheld the rejection of downward departure motions based on age for offenders in their 50s, 60s and 70s...I am not aware that any trial court decision to deny a downward departure based on an offender's age has ever been reversed on appeal."

"Risk and harm, the two factors purportedly measured by the RAI, are completely different considerations. The first is an empirical prediction. The second is a moral judgment. Yet...the RAI mixes and matches the two considerations in ways impossible to decipher..."