I'm not quite sure how to answer JJ Marr and Nathan, but if you watched the Berkman panel I posted about earlier [1], the conclusion of the WMF harassment survey is that the effect of harassment on women in Wikipedia, is that they leave.

And not to beat around the bush, for those who are not going to read the piece, the "certain topic" is sex, sex, and sex:

"...behaviors that were seen as being sexual or having a sexual edge to them were far more likely to be creepy than more innocuous ones. Women, especially, noted that behaviors like unwanted sexual advances, constantly turning the conversation towards sex, requests for photos, dates and invading their personal space were signs that a person was creepy."

So this goes back to defining harassment.  How do you tell the difference between someone who genuinely does not want to appear creepy, as in the hotel example, and someone who is deliberately skirting the boundaries, in order to harass people while flying under the radar.

There is a long history of defining harassment and "hostile work environment" in employment situations.  For in-person interactions, there is a whole set of non-verbal signals that tell you when to back slowly away, the "odd smile" for example.  But obviously in online communications, you are not going to be able to see how someone smiles.  Harassment on the internet is something new, the old HR harassment definitions can't just be copy-pasted.

And how far can you go in telling someone they have to adjust to something that creeps them out?  On enwiki, we have seen women advised to "keep a low profile" if they don't want to be photoshopped onto porn.  So the "encyclopedia that anyone can edit" is now "the encyclopedia that anyone can edit as long as they don't mind potential employers finding non-consensual pornographic images of them on the internet".   Paradoxically, the WMF has gone the opposite direction from arbcom, particularly in their recent safe space event policy, although the means of enforcement are not very evident.