The Washington Post was beside itself about the use of computerized “predictive policing,” where computer algorithms are used to assess the risk of recidivism and to set jail sentences:

.

The WP feels that racial bias creeps in because the initial data fed into the computers makes blacks and Latino defendants look riskier than Whites. There are racial biases in arrest and conviction rates because – the liberal faith has it – the existence of “racial biases,” so a Black or Latino defendant will be more likely to have a prior conviction than a White.

This is a perfect example of circular reasoning. There must be a racial bias because, well, there are a disproportional number of Blacks and Latinos convicted, and because they lack free will and agency, their plight must be socially constructed. The proof: in 2012 a White attorney tried to get himself arrested for carrying graffiti stencils and spray paint in Brooklyn, but did not, while 3,598 people (race not stated) got arrested in 2013 in New York for the same crime. Now there’s hard social science for you! For an alternative see:

.

It is left as a homework exercise for this Methods in Statistical Analysis 101 class to discuss the errors in the WP article. A brick of chocolate will be awarded to the student who detects the most errors. All papers must be handed to Professor LaFond by 9 am tomorrow morning before the start of Professor LaFond’s lecture on Bayes’ Theorem and shank fighting.