to "Update" List
Click to Notify Us
In Court Today
Bills in the
Letters / e-mails
How To Access
No SO Zones
Risk Assessment Overview:|
Sec. 1 Declaring:
Sec. 2 Predicting:
Sec. 3 Process & Correctness:
Sec. 4 Tools in Use:
Sec. 5 Tools Used by States:
*** Risk Assessment Systems: Process & Correctness ***
Risk assessments, as they relate to sex offenders, are supposed to "Predict" the likelihood of a re-offense in the future! Something like trying to predict whether a person will be alive tomorrow or not; we all know thats impossible with any certainty.
However, we could say, given a normal healthy person, it is more likely they will be alive tomorrow, than it is likely they will not. Yet, if a person had cancer, heart disease or other diseases, we could say, that person has a higher risk of not being alive tomorrow, than a person who did not have any of those diseases. Such is the principle behind risk assessments.
Problems with All Risk Assessment systems:
So, the process of risk assessment takes "known factors" that can cause a "outcome," and compare to a suspect person, to see if a person possesses those "known factors" so that you can say, probably the "outcome" will be true; or in the case of previously convicted sex offenders the "outcome" being, that they will commit another crime.
Currently risk assessments have established "ranges" being "High-Medium-Low (or some other grouping)" probability of re-offending. Ranges are established based upon the number of "known factors," for instance, a high range (high probability) might be 10 or more "known factors," while a medium range might be 5-10 "known factors," and the low range being 0-4 "known factors."
Accordingly, if a person has 8 of these "known factors" present in his history, then s/he would be considered a Medium Probability of re-offense. What is so interesting is, the Racing Sheets one gets at horse racing tracks are more likely to produce a winner than risk assessment systems will foretell when a sex offender will re-offend!
1) Systems are inherently biased (skewed). If you were to use facts from any person on earth, man, woman, child, and answer the risk assessment questions, it would calculate out that, that person is a Low Risk of re-offending, at very least. The problem is, all risk assessment systems, start their lowest range at "1" (meaning anyone can commit a crime, which is true); inherently biased.
Problems with "Known Factors" used by systems:
That means every risk assessment assesses a person, one range higher than they should be. The public is immediately mislead, and those persons assessed, are further damaged in their public reputation.
To illustrate a point: If you were to use the facts of a DEAD person, it would still say, that person has a low risk of reoffending. The problem here, the assessment doesn't ask if the person is alive.
2) Systems also ignore the age of the offender, as to his "physical" ability or likelihood of committing another offense. We all know that when a person gets old, or is in a wheelchair, confined to a nursing home, or any other such circumstance, it is very unlikely they will be out committing crimes.
Systems ignore "age-as-a-factor" that is the age of the offender. There have been studies which prove this, but all risk assessment systems ignore them. Again, the public is automatically misled as to older sex offenders.
3) Systems also fail to consider "age between offenses." Risk assessments are rarely used on persons who have only one offense, such would be purely speculative, but are used when offenders have more than one offense.
When systems are applied to multiple offenses, we have not seen one that considers the time period between offenses. The problem that this causes is, offenses committed maybe 20-30 years ago (possibly urinating in the park, or youthful encounters, etc.) coupled with the current offense, makes the person seem as if they are a Jack-the-Ripper type of person. Age between offenses is sometimes very relevant and is ignored.
"Known Factors" simply do not exist, what researchers have done is, reviewed groups of offenders who have re-offended, and looked for similar factors present in each of those offenders. The theory being, if a certain group of offenders, that did commit a second crime, possessed factor xx dd ww hh dd ff etc., then that became what they call a "Known Factor" to look for in another person, and presume if present then that other person will re-offend.
Above points out a few of the problems within each system, now we will get into other areas.
Given I will be taking significant heat for that very general statement, I'll back it up by saying, given there are MANY DIFFERENT Risk Assessment Tools (systems) being used today, why don't ALL OF THEM use the same "Known Factors?" Simply because researchers DISAGREE as to "Known Factors."
Now, in defense of the research that has been done, when a researcher develops a risk assessment system, they then test it by applying it to another group of offenders to see if it will pick out those that did re-offend in that group. This is called cross-validating the risk assessment system. Studies have been done on current systems, later you will see a few of them, which show about, half the time it is right, and half the time it is wrong; 50-50! No better than chance.
*** Risk Assessment Systems: A Layman's Understanding ***
Risk Assesment Systems:
A Layman's Understanding
It is important to understand how Risk Assessment Systems determine the likelihood of an offender committing another crime. It is equally important to understand the differences in the various systems.
(See Charts comparing systems: Chart-1 Chart-2 Note: Some of this information is very technical.)
Generally speaking, based upon historical factors related to offender recidivism (factors selected from prior research projects and new theories)(factors can be static [unchangeable], dynamic [changeable], or some combination of the two), each factor is assigned a weight (e.g. -1 to +3 [OR] -1 or +3). Then researchers or coders (trained personnel or the offender) assign the weights to the individual factors (by reviewing offender's history), and the weights are totaled.
Interesting is, the Predicted Recidivism Factor can be different from risk system A, to risk system B. This is caused by each system using different "historical/dynamic factors" used to score an offender. Every system has its own theory of factors relevant to recidivism. Today, researchers are striving to arrive at what historical and/or dynamic factors are relevant to determining recidivism. Hence the differences in the systems and the resulting score assigned to the offender.
Then the offender's cumulative total determines whether the offender is considered a low risk, medium risk, or high risk offender. For Instance: -# to +3 = Low Risk, 4-7 = Medium Risk, 8 and Over = High Risk (Each system having its own ranges). This cumulative number should be called the offender's "Predicted Recidivism Factor."
The Predicted Recidivism Factor is critical to the offender because it determines how s/he will be perceived by others, and handled by the agency administering the risk system. It must be remembered that it is a prediction only, based upon theoretical unproven historical/dynamic factors, each system having its own theories. Currently the predictive validity of these systems is less than 50/50; an offender may, or may not, re-offend?
(See Risk Assessment)
The Dangerousness Assumption Error:
While the predicted recidivism factor represents no more than the possibility of re-offense (and not necessarily a sex offense), that recidivism factor has been mentally equated to how violent the offender is considered. This is where the myth begins, society has interpreted this recidivism factor to mean how violent the offender is, when in fact it does not mean violence at all.
This quantum leap stems from media reports (frequently inaccurate and missing selected facts), and public speakers (including legislative declarations: ex: Michigan Legislative Declaration about sex offender registrants.) are reporting and generalizing about all sex offenders, citing non-typical high profile crimes as their basis. Then individual offenders lose their identity, and it is assumed each offender takes on the perceived identity given the class of offenders. All bad, equally dangerous and violent, the myth is perpetuated!
However, there are many other factors to consider before equating, recidivism factor as a person's dangerousness (violence) level. e.g. specific crime type (rape being more violent than touching type crimes) and circumstances of criminal incident/s, etc..
Risk Assessment Systems: Types--
Basically there are two types of systems, actuarial and clinical. See "RISK ASSESSMENT IN CANADA" for a very clear explanation of the differences in the systems. The Minnesota Sex Offender Screening Tool is a combination of, actuarial (historical static variables) and dynamic (institutional dynamic variables).
Clinical risk assessment is, by definition, an exercise in human judgment. The susceptibility of human judgment to error has been the subject of considerable empirical scrutiny (Garb, 1998). Actuarial Risk Assessment removes the possibility for human biases which enter into Clinical Judgment.
Risk Assessment Systems: Use--
Risk assessment systems are used for various purposes, to determine whether to parole an inmate, to determine the level of supervision a parolee or probationer will have, and to determine whether to civilly commit an inmate.
When states do use "tiered registries," the Predictive Recidivism Factor determines the offenders level of community notification. Most states ignore assessments entirely, and declare all registered offenders dangerous, falsely labeling many registrants, this raises serious due process questions.
At least one risk assessment system (the Minnesota Sex Offender Screening Tool) was first designed to be used only for determining whether to civilly commit an inmate, and now is being used by many states for other purposes. e.g. Tiered Registries, etc.. (See Community Notification Policy) One can only wonder if these other uses produces valid results, given usage against its design, or whether it further stigmatizes registrants. See The Fargo Police Department (South Dakota) will continue to use the Minnesota Sex Offender Screening Tool--Revised (MnSOST-R) to screen and score sex offenders who register as Fargo residents. NOTE: They cite its effectiveness in use within the Department of Corrections, however, that is where it is designed to be used, but not in society because a few of the questions pertain to the prison environment.
The Minnesota Sex Offender Screening Tool contains, 16 factors (questions) in total, the last 4 pertain to when the offender was incarcerated. Accordingly, when used on an offender who was incarcerated elsewhere, or never incarcerated (i.e. probation), the ultimate rating can be incorrect (from -3 to +8 range for these 4 questions). Further, when coding for "number of sex offences" the manual (available below) makes no allowances for cases where offenders "plea to lessor charges," the manual instructs to code, that circumstance, as 2 convictions. Further, a few of the factors have overlapping theories which results in, twice charging points for the same event (ex: factors 2 & 7).
It is a well known fact that, on average, a person's rate of recidivism decreases with age (Report Summary 2 pgs). (or Full Report, 29 pgs)
The MnSOST-R makes no allowances for the length of time between, the sexual offense and when a form is coded, the offender is considered the same risk from day one to his/her death. The DSM-IV, used by clinicians even requires consideration of the age of the offense with respect to today, when a diagnosis is being made, and uses seven modifying specifiers (mild, moderate, severe, in partial remission, in full remission, prior history). Yet, all risk assessment systems ignore this glaring fact!
In the Jones case (see below), overprediction in error seems to be the norm. Offenders suffer all sorts of burdensome punishments from this treatment, both in prison and in society.
Risk Assessment: Methods--
Here I distinguish System from Method to highlight an alarming trend, to use multiple risk assessment systems to determine a predictive recidivism factor. That alone shows the inaccuracies of individual systems. Many of the risk assessment systems have been individually "cross-validated," and more recently these are suspect, but when two or more are used together, the results prove what? Remembering that each individually is no better than mere chance. Clearly the offender is prejudiced and misclassified, and society is mislead.
When researchers develop these systems they may look at other systems, but they do not completely incorporate another system into the one they are developing. While they may use some of the same factors, never the complete system. Yet, agencies feel qualified and justified in using multiple systems to arrive at a predictive recidivism factor. Have we reached a mere guessing game?
The first problem is, each system uses "historical factors," the offender is then charged twice for these factors; this can also occur with "dynamic factors." Then comes the "factors" that are not duplicated, why were these NOT INCLUDED in the other system? Mere oversight, or was there proof that they "proved nothing" as to recidivism? Finally, it is critical to note that, there are many (multiples) different risk assessment systems that are being used today, researchers agree there is no perfect or even near perfect system. Sex offenders have become unwilling participants in experimental testing, which by the way, the US Supreme Court has already held is unconstitutional.
Yet, as these imperfect and defective systems generate misclassifications, the ex-offenders are tagged with misinformation in their state held records, and no way to either contest it or to remove it. It won't be long before "prior misclassifications" will be a "historical factor" further prejudicing and again misclassifying the person. "The most dangerous untruths are truths slightly distorted."
Risk Assessments: Performed by--
For the purposes of this section (Performed by--) we are excluding assessments not related to placement in society. Once a jurisdiction selects a risk assessment system (i.e., MnSOST-R, VRAG, Static-99, RRASOR, etc.), then personnel are needed to implement and perform the assessments. These systems require some level of offender background investigation. What skills must the investigator possess? Can investigators inject personal biases or prejudices into an assessment?
What laws cover the records generated during an assessment, and are the records considered private not subject to an FOIA request? Is there a system in-place whereby the offender may contest his/her classification? Can the offender be represented by counsel, and does the jurisdiction have to pay for counsel if the offender is indigent? These are some of the questions that come into play, and under current state of law, some have no answer.
In jurisdictions with "tiered registries" an investigative assessment team usually includes, local law enforcement, maybe child protective services or a similar agency, the local prosecutor's office, or the state has appointed a board consisting of persons representing these agencies. Finally and lease desirable are, jurisdictions that select a local police person to perform assessments. A closer look at these personnel will reveal they are all from the side of the law opposite the offender, and biases and prejudices are likely to be inflicted into the assessment.
If for a moment we were to look at the process afforded in civil commitment proceedings, either as the result of a diagnosed mental illness or at the end of a sentence (sex offenders), these folks are entitled to a "process" which includes the ability to contest the commitment. The theory being, the entitlement flows from their deprivation of societal liberty. Societal Liberty goes far beyond physical restraint, in the case of risk assessments that scarlet label offenders in society their entire life is now composed of subtle denials, as are the lives of those close to them. (See Societal Issues) Registered offenders are now targets for vigilantism (subtle and blatant), and many are becoming homeless; in some jurisdictions forced to report weekly to tell authorities where they stayed the last 7 nights. As homeless persons the registration laws are rendered useless, homeless offenders have no regular residence.
An assessment, what is it? It is a process, performed by a person or group of persons, whereby they arrive at (after reviewing the specific offender's history) a predictive recidivism factor. However, there are some jurisdictions (ex: Michigan) where the process is performed by the Legislature, whereby they legislatively declare (by reviewing old predictive recidivism studies generated by studies [which used small selected offender populations to prove a theory]) this offender's predictive recidivism factor! Crucial is, offenders assessment based upon his/her own history ---OR--- offender assessment based upon the history of offenders of studies!
Due process mandates, a person be held accountable for his own actions, not the actions of others he has no connection to! Individual offenders' interests are not represented at all. Fairness is supposed to prevail, but what is in-place if the offender feels he is dealt with unfairly? Only a select few states have a defined process by which the offender may contest an assessment.
The risk assessment process (by declaration or by a system that prevents appellate process) is severely flawed, citizens are ostracized from the community, in every sense of the word, and it flows directly from legislative acts. A lesser form of a Bill of Attainder is, a Bill of Pains and Penalties, which are legislative acts that inflict punishment without a trial (a process), and have been unconstitutional for many many years. See United States -v- Brown.
A STANFORD UNIVERSITY psychiatrist thinks that evaluating the danger (Risk Assessment) posed by sex offenders should not be left solely in the hands of the police. Nebraska, local police. See Ques: Who completes an assessment.
States with no risk assessment systems:
Most states do not use any Risk Assessment System for sex offenders in society, instead the states legislatures merely declared ALL REGISTERED SEX OFFENDERS ARE DANGEROUS! I.e., Michigan. No system whatsoever, no hearings, no actuarial system, no clinical assessments, legislative declaration, this used to be called a "Bill of Pains & Penalties" in violation of the US Constitution!
*** Risk Assessment Systems: How Often Are They Correct? ***
In Jones -vs- United States, 463 US 354 (1983)(and a few other cases) the court examined the "prediction of dangerousness from any set of facts," and said:
"It is worth examining what is known about the possibility of predicting dangerousness from any set of facts. Although a substantial body of research suggests that a consistent pattern of violent behavior may, from a purely statistical standpoint, indicate a certain likelihood of further violence in the future, 8 mere statistical validity is far from perfect for purposes of predicting which individuals will be dangerous. Commentators and researchers have long acknowledged that even the best attempts to identify dangerous individuals on the basis of specified facts have been inaccurate roughly two-thirds of the time, almost always on the side of overprediction." NOTE: American Psychiatric Association, 1982 confirmed this report!
Jones went on to say, that the longer an offender does not repeat the same act, the likelihood that s/he will repeat the offense is greatly diminished. In current systems the offender's age is a factor ignored! Studies have proven that -age reduces recidivism- (R.K.Hanson 2001). This is not factored into the systems at all. "The most dangerous untruths are truths slightly distorted."
Under the Jones analogy, if an actuarial system declared 100 offenders dangerous, the reality is, only 33 of them would possibly be dangerous. Under a static (unchangeable) actuarial risk factor system, the offender's dangerousness classification would never change because it uses static risk factors (factors being fixed in history).
Cross-Validation/Multiple Systems: More recently several of the actuarial Risk Assessment Systems have come under criticism because they have been found to work with one type of offender and not another type. (Article 8-Dec-2002 Scotland) Also because of no standard method of cross-validation, or cross-validation simple has not been done at all. (9-Nov-2002 Scotland: System of assessing risk of sex offenders is flawed) Further, when states uses multiple risk assessment systems, using the second or third one to modify the finding of the first system, all cross-validation of the individual systems is rendered void, because there has not been any cross-validations of the use of multiple systems. It becomes a guessing game.
Finally, whatever burdens or consequences result from mislabeling (low-medium-high) of offenders, it is clear that 66% are having their lives, and their families lives, affected, but it is impossible to tell which 66%?
However, R.Karl Hanson, Ph,D. recently (1999) compared the predictive accuracy of three sex offender risk assessment measures: the "RRASOR (Hanson, 1997), Thornton's SACJ-MIn (Grubin, 1998), and a new scale, Static-99 in this study on the Static-99 system (FULL REPORT: a 29 pg PDF file) or you can read his summary report (2 pg PDF file).
Then in 2000, recognizing the "Static Nature(unchangeable historical facts)" of the above systems, and that there were no scales to determine CHANGE in a sex offender's risk level, he developed SONAR (Sex Offender Needs Assessment Rating) system to fill this recognized gap. See Full Report: 25 pg PDF file or you can read his Summary Report 2 pg PDF file.
Courts have been discussing the merits and failures of both, actuarial and clinical types of systems and admissibility of each in court: Arizona Court of Appeals; Illinois Court of Appeals. The fact of the matter remains that, even with improvements, risk assessment will continue to be no more than educated guesswork.
See paper on Myths About Violent Sexual Predators and Pesky Legislation