This question occurred to me when looking at one of the answers on this question: Showing a homeless person that you care about their well-being.
The asker had a complaint about some specific predictions the answerer had made about the likely outcome of following their advice. The asker felt uncomfortable about some perceived implications of those predictions, and accordingly made a comment requesting that those predictions be removed.
I made a comment suggesting that the perceived implications were not obvious in the text being discussed, and added that it seemed odd to ask answerers to remove information they felt was relevant to the advice they gave for (what appeared to me to be) a thin reason. That comment was deleted.
The summary of the comment was that the assertion which prompted the complaint (that the predictions suggested all homeless people must be grateful about receiving basic human contact, and are bad people if they don't) seemed to go way beyond what the answer stated. The answerer simply listed two predictions about what they expected to happen, and there was nothing I saw in the surrounding text suggesting that anyone's moral status depended on them responding in that way.
I'm not complaining about the deletion of the comment, but as it's apparently not a good place to discuss the matter in vivo on that answer, meta seems like it might be a better place.
What are the standards we want to enforce for deciding which elements of answers answerers should or must remove, and when are requests or demands for those alterations ones we want to enforce?
(sidenote: my nontrivial experience with homeless populations suggests to me that those particular predictions are very unlikely to happen, but this meta question isn't about the accuracy of the predictions so much as it is about information the answerer felt was meaningful to include in the answer)
Update: After discussion with the commenter in the above-linked thread, I think that this specific situation was mostly a case of tone being hard to express and understand well over the internet.
I do think that this is a good thing for the stack to consider, but the question and answer linked above are not a very good example of it.