> SFU study confirms statistics can’t precisely predict who will reoffend

Document Tools

Contacts:

Stephen Hart, Psychology: hart@sfu.ca; 604.291.5485 (out of the country until July 11 but available by e-mail)
Kevin Douglas, Psychology: douglask@sfu.ca; 604.268.7050
Ron Roesch, Psychology: roesch@sfu.ca; 604.291.3370
Julie Ovenell-Carter, PAMR: joc@sfu.ca; 604.291.4323

Related web site: www.sfu.ca/mhlpi/

A new study by SFU forensic psychologist Stephen Hart confirms that risk assessment models long favoured by social agencies to predict the likelihood that a convicted criminal will reoffend are “fraught with problems and give a false sense of security.”

The study, recently published in the British Journal of Psychiatry, examines the precision of statistical risk assessment models and evaluates the margins of errors of group versus individual predictions of violence. The study concludes these traditional models “cannot be used to estimate an individual’s risk for future violence with any reasonable degree of certainty and should be used with great caution or not at all.”

The tests work by assigning a score to people based on factors such as their age, their school history, and the nature of their past crimes. The resulting scores are used to assign them to a group in which a known proportion have gone on to commit a crime after release from prison. Their individual risk of recidivism is thought to be similar to the risk posed by the group as a whole.

However, observes Hart: “It’s a mistake for agencies to assume that predictive success at the group level will translate to predictive success at the individual level. You can have perfect knowledge at the group level and still never know what will happen in any given case.

“Reliance on these statistical models encourages a passive managerial approach,” he says. “Agencies like to be able to point to a number and say, ‘See: this person is likely to reoffend and there’s nothing we can do about it.’ It blinds them to a more proactive approach.”

To that end, Hart and his colleagues in SFU’s Mental Health Law and Policy Institute have developed protocols for individualized assessments that are widely used around the world. These assessments help agencies manage the risks posed by individual offenders by providing appropriate community supports and services, without the need for making predictions of probability.

Hart likens the approach to dealing with the threat of earthquakes. “We can’t predict individual earthquakes with any degree of accuracy, but we can identify those areas that are most at risk and then take rational planning steps to minimize the possibility of harm.”