If I have read their argument correctly the author proposes that the interview "bar" was lower only relatively speaking. Allow me to elaborate: in a somewhat clumsily stated thesis they point out the disproportionate care granted to the favoured class is evidenced by an allegedly reduced false negative rate. If that's true then, in a way, Google haven't reduced the technical "bar" per se except by making the test more reflective of their actual expectations for hires. This is, I believe, the heart of the author's argument in this case: these improvements in the interview process apply only to favoured groups. If true, these exclusive improvements yield a test which, though just as technically thorough, is paradoxically still easier to pass given a favoured candidate. If my understanding is correct, there really has to be a better way to phrase the author's argument such that it can't be misinterpreted in the most common way it appears to be being taken now.