I often upvote both a comment and its rebuttal, because in many debates I have no dog in the race and I often find that people on both sides have valuable or otherwise interesting insights. Even if I don't agree with them completely, I sometimes find that I'm able to view an issue from another perspective that I hadn't considered before.
Not to mention, many debates really just don't have a clear answer, and to pit commentors against each other like that is to enable one of them to be declared "winner", which I think would be misleading in many cases.
In a discussion being had in good faith, a thread will frequently go into a pseudo-debate mode to "find the truth" of a statement someone offered. Kind of like we're doing here. We all add anecdotes and counterfactuals and so forth, and see where the inductive process takes us. None of these comments would/should be tagged as rebuttals.
But there's a very specific situation where, I think, this feature is an important addition: when the original comment (or the post-link heading the discussion!) just doesn't know what it's talking about, and the reply outlines why this is so.
The concept is less like "this post disagrees with its parent, and so up-voting (agreeing) with it should mean down-voting (disagreeing) with its parent", and more like "this post is a petition to flag/retract its parent, and up-voting it is signing that petition."
This is why I had point 4 in the above, which might otherwise seem an unrelated feature: negatively-scored posts are not "disagreed with", but rather "flagged dead." But negatively-scored posts must still stay visible, in order to give requisite context to their positively-scored rebuttal-comments. And, in fact, a thread containing a negatively-scored top-comment might even still be the top-sorted thread, if its replies are considered valuable enough.
(One might still want to add some visual effect to remind those reading the post that the community considers it "retracted." Perhaps adding a strikethrough that disappears on hover, or a background of faint red Xes. You don't want to make the post illegible—it's still necessary context for its subthread, unlike the current HN/Reddit system where the post "fades" to nothing, and then the whole subthread gets considered a lost cause and collapsed.)
I wonder if the problem is the binary nature of so many things. 5 star/1 star reviews are the most common, we have up and down votes, but no real context for that. I wonder if the problem that needs to be solved isn't context, rather than directional.
This is me spitballing, so feel free to IGNORE everything below.
I was listening to the Freakonomics podcast on proper voting (e.g. elections) the other day - this one: http://freakonomics.com/podcast/idea-must-die-election-editi... - and there was an idea called quadratic voting. http://ericposner.com/quadratic-voting/
Anyway, the idea is that we all get X votes, and we can vote more on some things, e.g. if we REALLY want marijuana legal, I can vote 4 votes. The twist is that the cost of extra votes is the square of the number of votes, 2 costs 4, 3 costs 9 (we all get maths works).
The theory is this lets people vote on what they care about, much less than simply voting equally.
I was wondering there isn't some way to create a system where you earn votes, that you can spend in some similar way, e.g. you "earn" upvotes from others, and can spend them elsewhere. If everyone got a single vote, but you earned extra votes you could spend in interesting ways at increasing cost, e.g. you could heavily downvote an idea at an exponential rate, it might make people less a victim of common denominator views, give fringe views more airtime, and make people consider a vote more deeply. Not just yes/no but how much do I hate/love this idea? Enough to blow 100 points on 10 down/up votes?
Anyway, just some random braincrack (https://www.youtube.com/watch?v=0sHCQWjTrJ8) I needed to get out.
1. both of the 'unit-weight' up- and down-votes were only done through implicit/passive actions (basically a piece of Javascript running in a browser extension, that tracks whether you read the whole article/comment—i.e. scrolled through it at a reasonable reading speed—before scrolling away/closing/going back/beginning to reply)
2. there are above-'unit-weight' up- and down-vote mechanisms that are explicit, but require something humans consider very-slightly onerous: solving a CAPTCHA, paying a tenth of a cent of pre-paid credit, etc.
In my conception, the small votes would be named "OK" and "Meh", and the large votes "Love" and "Hate", with about a 10x difference between their power. The site, of course, would be called "Mehddit." ;)
I am wary of this, especially when it comes to the web. I can see this system being "controlled" by early and/or very active users.
1. there are replies and up-voting, but no down-voting;
2. you can classify a comment reply, at time of posting, as a "rebuttal";
3. up-voting a rebuttal comment will also be considered as down-voting its parent;
4. the aggregate score of a subthread is calculated as a Euclidean distance, with all the positively-scored comments as dimensions. Thus, if you've got one great comment made in response to a bad comment, the thread will stay un-collapsed; if you've got two equally-great comments, the thread will rise, etc. (Likewise, if you've got good discussion happening in response to a really bad post, the discussion's value will still propel the post to the hot page.)