I don't want to waste valuable time figuring out exactly how sure I am that a particular title is bad, or how polite I should be, or whether or not I should take advantage of this high-bandwidth feedback mechanism to start some kind of conversation.
Learn the design lesson of Twitter: More degrees of freedom doesn't necessarily help.
I think this is a GREAT idea if it uses pre-selected phrases to convey specific / useful info to the author.
Given those two prerequisites, I think it's a great idea.
I'd hate for people to abuse the anonymous nature of it sometimes and be rude-- we haven't had any "private" messaging so far on HN, so this would open a little of that world for moderators to think about.
Yes, it would definitely be a good idea, if private comments attached to flags are implemented, to say explicitly that moderators/curators are allowed to view all of them.
A few other thoughts:
Would it be helpful have 1 word positive and negative flags like: offtopic, uncivil, helpful, insightful, inspiring, trolling, etc...? That way, positive as well as negative behavior could be reinforced.
Would it be helpful to have a delay in how soon the flag shows up. So much online discussion get's aggravated people get angry at getting downvoted, flagged, etc... and then they quickly flame back. If there was a delay in when the flags showed up, would it decrease people's tendency to post "why did I just get downmodded" an hour later.
I have enough spam and distraction in my life.
I don't know how to explain it, but I'd almost rather the flag options be worded as "Sorry, this comment seems: uncivil" or "Sorry, this comment seems: off-topic" to lighten it a bit. If they were a friend in real life, I'd take them aside and say "hey man, that comment was great, but you might want to soften the tone. It kind of came across as rude."
Downvotes end up being seen as judgments of your ideas, and I would love to be able to provide better feedback. This would be great if this could somehow give the user a nudge that they wouldn't feel defensive about.
It may seem like a minor point, but I feel a little guidance would go a long way in helping users get the sort of social cues they already get in a real-life conversation.
Edit: Instead of calling it out as flagging, I'd be okay with a downvote system that simply/optionally let you specify a reason to provide (with 3-4 common nicely-worded options). I don't care about upvote reasons at all.