Every now and then, I write an LW comment on some topic and feel that the contents of my comment pretty much settles the issue decisively. Instead, the comment seems to get ignored entirely - it either gets very few votes or none, nobody responds to it, and the discussion generally continues as if it had never been posted.
Similarly, every now and then I see somebody else make a post or comment that they clearly feel is decisive, but which doesn't seem very interesting to me. Either it seems to be saying something obvious, or I don't get its connection to the topic at hand in the first place.
This seems like it would be about inferential distance: either the writer doesn't know the things that make the reader experience the comment as uninteresting, or the reader doesn't know the things that make the writer experience the comment as interesting. So there's inferential silence - a sufficiently long inferential distance that a claim doesn't provoke even objections, just uncomprehending or indifferent silence.
But "explain your reasoning in more detail" doesn't seem like it would help with the issue. For one, we often don't know beforehand when people don't share our assumptions. Also, some of the comments or posts that seem to encounter this kind of a fate are already relatively long. For example, Wei Dai wondered why MIRI-affiliated people don't often respond to his posts that raise criticisms, and I essentially replied that I found the content of his post relatively obvious so didn't have much to say.
Perhaps people could more often explicitly comment if they notice that something that a poster seems to consider a big thing doesn't seem very interesting or meaningful to them, and briefly explain why? Even a sentence or two might be helpful for the original poster.
Because the karma is numeric, those buttons should all also have numeric values +1 (want more of this) or -1 (want less of this). It would be probably good to preserve the old-style "upvote" and "downvote" for people who just want to vote and not analyze why exactly they felt this way (because having to analyze everything would be a trivial obstacle to voting).
So perhaps the good mechanism would be to first show the choice of downvote and upvote, and after the vote is made, display relevant subcategories and let voter optionally choose one of them which fits most.
Upvote -- nice, interesting, informative...
Downvote -- trivial, wrong, off-topic, flamebait, trolling...
Then the total score of the comment should also display a most frequent flag or two. Such as "5 points, interesting", or "-10, trolling". Maybe display the flag only if overwhelming majority of voters agree on it; otherwise just numerical value without flag.