By Andy Ho
SERIAL sex offender Bala Kuppusamy, 48, was recently sentenced to 42 years. First convicted in 1987 and sentenced to 11 years, he was released in 1992 only to reoffend 45 days later. Convicted again in 1993 and sentenced to 23 years, he was again released early. Within 41 days of his release in March last year, he sexually violated three women again.
Relying mainly on clinical judgments to determine who is more likely to relapse, prison authorities tailor preventive treatment accordingly. Moderate risk offenders undergo four months while high risk ones undergo six to eight months of treatment. However, low risk convicts are exempted from treatment. Yet 6.9per cent of the low risk ones re-offend within two years.
Making treatment decisions based on clinical judgments for which there are no confirmatory lab tests may be improved if augmented by a more accurate and structured way of assessing the recidivism risk.
Ideally, clinicians must be able to articulate how they arrive at their conclusions and enumerate the factors upon which they base their recidivism predictions for any particular case. But how they turn such information around in their heads is a process that can hardly be ferreted out by outside observers.
In fact, cognitive biases are known to plague clinical judgments. One is the tendency to collect data that you hope will support your preferred hypothesis and not collect data which you think might contradict it. This is called the confirmatory bias, which may also cause you to interpret the data you collect in ways that support your pet hypothesis rather than an alternative and plausible explanation.
But such things are hard to pin down, so it all comes down to an expert's credibility. In the past 15 years, however, several actuarial instruments have been developed to objectively assess the risk of future sex offences. Locally, both clinical assessment and actuarial instruments are used to assess their risk level when sex offenders enter prison.
These actuarial instruments are based on a method used by insurance companies to measure risk. First, the characteristics of recidivist sex offenders are identified. These risk factors are then linked to real outcomes (actual rates of recidivism), so a quantitative measure of future risk for a person with a particular cluster of characteristics can be defined. These traits may include deviant sexual arousal as measured by penile engorgement, a history of substance abuse, placement in residential care in childhood, and so on.
Such actuarial instruments have been tested for accuracy in many empirical studies. These studies, in turn, have been reviewed by, among others, Canadian crime prevention authorities in 1997, 2002, 2004, and 2007. Strikingly, all these reviews came to the same conclusion.
In 2002, reviewers found that 'one of the most consistent findings is that evidence-based, actuarial measures are more accurate in the prediction of...recidivism than professional, clinical judgment'.
In 2007, reviewers opined: 'Given the weight of evidence supporting them, we believe actuarial risk tools should be a major consideration in the evaluation of recidivism risk potential.' Similar analyses of these empirical studies that academics have carried out agree.
While a clinician may weigh causal factors wrongly and/or inconsistently, actuarial instruments are based on empirically observed relationships between predictors (measurable offender characteristics) and real outcomes (recidivism rates) over a follow-up period. Moreover, different experts can use the same instrument to verify its reliability, so you don't have to take assessor ability on faith alone.
But do such instruments identify the actual risk a particular sex offender poses? In Preventing Sexual Violence (American Psychological Association, 2005), author John La Fond, who is a law professor at the University of Missouri-Kansas, argues that it is unfair to keep a person in jail just because he 'looked like' others in a group with the same score.
Yet, apart from wild guesses, all predictions in the real world do refer to the behaviour of a relevant group. Supposing a desperately overweight smoker whose father died at 55 from a massive heart attack sees a doctor who says to him: 'Group data says you are at high risk of dying from a heart attack. But, as an individual, you needn't worry about it.' Any court would find that doctor negligent.
In many things, then, we predict our individual risk from the appropriate group risk. Even in sexual recidivism, clinician predictions are also based on a convict's similarities to other known recidivists. The clinician who has seen 500 cases (and presumably compares the present case with them) is considered to be more credible than one who has seen only five cases. But the clinician cannot describe those 500 cases that guide him in the case under consideration. By contrast, actuarial instruments rely on explicitly identified reference groups and unambiguously enumerated rules about how conclusions are reached. Hopefully, the use of actuarial instruments to augment clinical assessments of recidivism risks will, over time, improve decision making as to which convicts require treatment and for how long.