[derefr]

It's a model that predicts their behavior well, is what I meant. It "makes sense of" their decisions. It doesn't justify those decisions; of course they're awful decisions.

I'm doing kremlinology here: attempting to model an opaque organizational structure so that you can change your behavior to avoid its inferred structural inadequacies. Because any organization in need of kremlinology ain't gonna be able to change in response to even a top-down order to do so.

Google can't stop being catastrophically Balkanized, any more than IBM can stop being catastrophically monolithic, or Facebook can stop being catastrophically lackadaisical toward code-base quality. It's in their DNA: in the kind of people they have hired, who now choose who they hire, and will continue to be biased however they're biased even if a top-down order tries to push against those biases.