My point is that the fact that the probabilistic inference is valid does not imply that you should e.g. use examples that assume the user is that way, which was the reason you were making that point to begin with. I can safely assume that an unknown user is male. Doesn't mean I should use male-experience specific examples for elucidation.

(Picture me saying this in dramatic tones, standing on a podium wearing robes and frequently howling "Fools!")

For a perfect Bayesian, it works. For humans, not so much. Just having a category exist makes us develop silly beliefs around it. If they're categories of people, we start loving our category and hating others - the ingroup/outgroup dichotomy. We treat ourselves as default and other, er, others, increasing the status differential. If a power structure already exists on top on that, forget it. It's really not innocent.