Study Finds Crime Predicting Algorithm Is No Better Than Poll of Random Internet Users

Many people have claimed that algorithms and other advanced technology can help law enforcement determine who is or isn't more likely to commit a crime. But it turns out that those algorithms are often no matter than some random person on the internet.

A study by researchers at Dartmouth compared the algorithm COMPAS, which helps determine how likely a former convict is to reoffend and is used by multiple states, with random poll users on Amazon. The study found that the random internet users, who had no criminal justice experience, were just as accurate as the algorithm.

The researchers gave the poll users information about a defendant and asked whether they believed the person would commit a crime again. The poll users were correct about 67 percent of the time. The COMPAS algorithm was correct only 65 percent of the time.

The company behind COMPAS said the study actually confirmed how effective their algorithm is, noting that the standard for risk-assessment tools in criminal justice is around 0.70.

It turns out that when you actually look at all the data, there are only two things that matter in determining whether a person will commit a crime again: age and prior convictions. The older a person is, the less likely it is that they will reoffend. And the younger the person is, the more likely.

Do we really need a computer to tell us that?

(h/t Gizmodo)


Nowadays, would your parents still be upset if they caught you consuming cannabis? Parents these days have much more progressive opinions on cannabis, and perhaps if they caught their kids consuming, they wouldn't necessarily punish them. While some parents still want their children to wait until the legal age to consume (if they choose to do so, at all), others don't believe it would be the end of the world if they "caught" their kids smoking pot earlier than that.

Can we see some ID please?

You must be 19 years of age or older to enter.