I am generally against dystopian ideas like "predictive policing", and I used to like Doctorow very much. I vividly remember being inspired by Little Brother. Also, of course he is right in saying that crime data is not completely objective.

But TBH, the article itself is also guilty of confirmation bias. It is full of "everybody knows" [1] kind of arguments and exaggerations. "Driving while brown" isn't really a "de facto crime" in any reasonable sense.

The analysis of leak by The Markup he endorses doesn't look very objective either. The fact that the Predpol's algorithm suggests that the police should patrol the more dangerous neighbourhoods isn't really a good argument for showing that it is racially biased. At the very least, they should try to control for other factors somehow (e.g. by comparing poor neighbourhoods with different racial compositions).

There is indeed "a pool of mathematicians who can produce convincing confirmation bias as a service", as Doctorow says, but their prejudices are not mostly in the category and direction he suggests. [2]

[1]: https://archive.ph/VYsZ3#selection-335.61-335.76

[2]: https://en.wikipedia.org/wiki/Political_views_of_American_ac...