Personally, I would love to see Yonatan's explanations be made more official. I don't know if there are any plans for that (I write code for a living, I'm not i

Personally, I would love to see Yonatan's explanations be made more official. I don't know if there are any plans for that (I write code for a living, I'm not in PR :-) ) Yonatan has a great thread going on in his personal account where he is responding to questions from users. He is the primary architect of both G+ and the new policy. I would recommend reading through it and posting your questions there (https://plus.google.com/u/0/103389452828130864950/posts/YJbzDptWGQt https://plus.google.com/u/0/103389452828130864950/posts/YJbz...) I know that the team would appreciate real-world examples of names that are edge cases, or that would likely be false-positives in the current system. As far as how they determine what it means to be "name-shaped", I'll just copy one of Yonatan's responses. I really recommend reading through the thread. "OK, and now to answer the really damned tough question of "just how the heck do you define name-shapedness?" And I can't go in to too many of the technical details, but the zeroth-order answer is "using humans, boosted by computers." Humans look at names and flag things which look like they aren't names; anything which is flagged gets reviewed by lots more humans, cross-checked, vetted by people with cultural familiarity, etc. This process tries to err on the side of treating things as OK. (!) We also build up computer models to basically cache this and handle the "easy" cases automatically, since obviously having humans do everything doesn't scale. The goal is that most things which are marked as "not a name" are genuinely cases of something being meant as either a nickname or an organization; whenever the appeals process is triggered, and even more so whenever something passes an appeal, that's a sign that the first-stage check failed and we need to improve our rules. So then we can look at the pattern of appeals, see if there are classes of names which we are systematically getting wrong, and learn from this to improve the process and reduce the chance of someone being sent through it incorrectly."