Wednesday, October 5, 2011

Combating the pull to overpredict violence

Like the moon's effect on tides, the pull to overpredict violence exerts a powerful influence, even on seasoned forensic evaluators who know its strength.

When directly informed that an event has a low base rate of occurrence -- for example, that a homicide offender has only a 1 in 100 likelihood of being arrested for another homicide -- both laypeople and professionals will markedly overpredict violence.

In an article in the Journal of the American Academy of Psychiatry and Law, eminent forensic psychologist Stanley Brodsky and postdoctoral fellow Sarah L. Miller analyze why this is so.

For one thing, the risk of underpredicting violence has more potential to negatively impact the evaluator. Bad publicity, public outrage, even civil litigation. Not to mention the harm committed by a high-risk individual who reoffends. 

Far safer to "err on the side of public safety," goes clinical lore. A claim of dangerousness is well nigh impossible to disprove. And especially in the context of civil commitment of sex offenders, the issue is not framed as punishment but, rather, as "an acceptable restriction of individual rights in the interest of public safety and rehabilitation." It's not as if these guys are sympathetic characters, with a constituency of supporters looking out for their rights.

Certain psychological mechanisms also contribute to bias in the direction of overpredicting risk. These include confirmation bias, or seeking information to support a preconceived conclusion, and illusory correlation, in which the evaluator assumes two things are related just because they co-occurred.

The purpose of Brodsky and Miller's well-argued review is to make evaluators more aware of the natural overprediction tendency, and to provide a checklist that evaluators can use to assess and correct their potential biases.

It's a great idea, although I am a bit skeptical that such a simple approach will make much of an impact in the adversarial arena.


The full article is available for free download HERE.

No comments:

Post a Comment

Note: Only a member of this blog may post a comment.

 
Real Time Web Analytics