The author(s) may just be learning what the reactions are to the comments. Here for instance, the comment elicited downvotes. This may be used to teach a model to avoid the style of the comment