Kevin Drum assesses a research result:
A new paper with access to Uber’s massive database of driver records concludes that female drivers earn 7 percent less than male drivers. Why? Mostly because women drive more slowly than men.
There are a couple of other factors as well that are tied to experience, and that’s interesting enough by itself. But the authors call their result “surprising,” and I think that’s the wrong conclusion. The proper conclusion is that in a job that pays via algorithm and has no special rewards for working long hours, the gender gap is only 7 percent. That’s what you get when there’s no opportunity for discrimination.
Without knowing anything about Uber’s “algorithm,” I think Kevin is committing an indiscretion when he assumes an algorithm is gender-blind. The press (NewScientist, I’m sure, although I don’t happen to have any links handy and I’m feeling lazy today) has reported several instances of machine learning algorithms returning results which are not gender-blind. This happens because the data on which these algorithms are trained reflect society’s biases, and the machine learning algorithm is incapable of compensating for such biases.
I’d also like to say that it’s possible the authors of the study were surprised because, well, it makes for more publicity if you’re surprised.