Wojtek Sokolowski wrote:
> [WS:] I think it is only a problem when you limit yourself to
> paper-and-pencil testing. If, otoh, you start observing the suspect
> (identified based on those test) and see in what kind of behavior he
engages
> in everyday life - is he being set off by trivial events, is he
aggressive,
> does he torture animals, is he trying to acquire weapons ? then the
> predictive power of your profiling will increase rather dramatically. .
>
> Wojtek
>
> _
No, that doesn't help. The vast majority of people who torture animals
do not go on killing sprees. Same methodological problem.
[WS:] Have you heard of multiple logistic regression? It predicts the probability of an event from multiple predictors. If you get that probability high enough that may warrant various protective measures. Will it infringe on the "civil rights" of a few deranged individuals? Sure. Does it matter? It depends what that infringement is - but it is possible to contain potentially dangerous individuals in a humane way. More importantly - it is highly desirable, as it sacrifices comfort of a few deranged individuals for the lives of many, which is priceless - which is the essence of Pascal's wager.
Wojtek