>And precisely how was this done?

Can't find the article atm, but it was basically pre-crime from Minority Report (without the pre-cogs, obviously). They looked at large datasets and built a predictive model, correlating things like race and prior criminal history to infer who was more likely to re-offend. At scale, this works. Ethical issues abound, however.

> At scale, this works.

We're going to need a definition of "works." The false positive rate seems to be notable since those stories readily percolate into media whenever these schemes are implemented and the damage done from those is absolutely massive.

The idea that criminals are likely to re-offend is not new. What to do about this has always been the challenge. Simply over policing this segment is not any type of solution. Unless, of course, you are invested in the "private prison" industry.