I'm becoming more and more convinced that Google employees don't give out reasons for their decisions because they can't! They trust too much in their machine learning and there's no way for employees to know why this event triggered over that one.
If your ml model is a magic black box and there's zero visibility into the system, even internally, as to the signals that resulted in a decision, you've hired a monkey that can run a tensorflow tutorial, not actually someone that understands machine learning.
Or (more likely) because they are not allowed to by policy.
Or because the machine won't tell them why because privacy. (Although this applies less in the ad side of the business than consumer facing bits)