Can someone versed in ML explain how generating an "algorithmically written crime report" from other features that the network will already be incorporating adds to the predictive performance of the network?
> came under fire from activists and civil libertarians who argued the firm provided a sort of “tech-washing” for racially biased, ineffective policing methods.<p>Well, yes. The function of labelling crime as "gang related" is primarily to impose higher sentences on nonwhite criminals.<p>The most effective approach is pre-emptive intervention in the style of the Violence Reduction Unit: <a href="http://www.actiononviolence.org.uk/about-us" rel="nofollow">http://www.actiononviolence.org.uk/about-us</a> (invented in Boston, extremely effective in Glasgow)