Hacker News Reader

Comment by tptacek | Hacker News Reader

No, it's clearly not. By an overwhelming margin, most people driving "the wrong car" are not criminals; "wrong car" is not a real predictor for criminality.

View on HN

Imagine a jar with two colors of gumballs: red and green. Now imagine you knew that 1 out of 1000 green gumballs were poisoned and 1 out of 10000 red gumballs were poisoned. Which gumball would you choose to chew if you had to choose? Why?

My guess is that you would choose the red gumball. Even though the vast majority of green gumballs are perfectly fine, the color of the green gumball is still a predictor for poison.

Here’s the kicker though: the more times you test a gumball for poison, the more poisonous that colour gumball becomes.

Low socioeconomic status is a predictor for criminality, as well as a predictor for car type owned. Thus car type is correlated with criminality.

High socioeconomic status is a predictor for white collar criminality, as well as a predictor for car type owned. Thus car type is correlated with criminality.

sure, but how often does a traffic stop lead to the discovery of evidence that justifies a charge for a white collar crime? the person would have to be driving around with incriminating (likely very technical) documents and do something that gives the officer probably cause to search the vehicle. the papers won't even look bad unless the officer already knows what to look for.

on the other hand, traffic stop -> suspected intoxication -> field sobriety test / drug dogs seems like a very common routine.

And this is how negative feedback loops happen.

Is it?

First part of your sentence, about most people, is obviously correct, but has no relation to the second part of your sentence about it not being a predictor for criminality.

Reply on news.ycombinator.com