Quoting Eliezer:
One of my ever-pending posts to write is on what sort of simple interface might prevent online arguments from retracing the same points over and over. I suspect it will not be graphical with boxes, because that makes poor use of screen real estate. I suspect it will not have lots of fancy argument types and patterns, because no one really uses that stuff. I think it does need to have a karma system, because otherwise there's no way to find the good stuff.
The basic rationale is filtering; the current system is satisficing, given some reasonable assumptions as to how online communities work, and has the "feature" of being inherited from a successful web site and thus being a "proven solution" rather than something speculative designed from scratch.
The major issue with LW's current karma system is information cascades; that has been acknowledged almost from the start of LW. Yet solutions intended to correct this have not been widely adopted.
People who go back and downvote every post or comment a Less Wrong user has ever made, please, stop doing that. It's a clever way to pull information cascades in your direction but it is clearly an abuse of the content filtering system. It's also highly dishonorable. If you truly must use such tactics then downvoting a few of your enemy's top level posts is much less evil; your enemy loses the karma and takes the hint without your severely biasing the public perception of Less Wrong's discourse.
(I just lost over 200 karma in a few minutes and that'll probably continue for awhile. This happens to me every few weeks. Edit: I mean it's been happening every few weeks for a few months for a total of only three or four. Between 400 and 700 karma lost total I think? I don't mean to overstate the problem.)