Fast questions. Unreasonably specific consequences.

Pick what is most true, not what looks flattering. The algorithm already assumes you are trying to look normal.