Ask YC: Voting Karma
Most of my Karma comes from comments because I honestly take a pretty passive role here, taking in and commenting instead of actively adding nodes; however, I have an interesting habit. Whenever I write a comment, before I submit, I double check to see what value I'm really adding. If it's not pretty significant — if my comment is pretty much just agreement with something someone said with little additional information — I drop it and just upmod. It's a small effort to try to bring up the signal-to-noise ratio recognizing that even well meaning comments can be be noisy.
The part I want to call attention to is that, to me, it seems like you're still adding value when you upmod. It's far less than a full comment or submission, but you're calling attention to signal instead of noise.
My suggestion is that Karma could be incremented based on some fraction of the number of upmods that occur on an article you upmodded after you did so. This rewards the people who read articles (especially those on the new page) and vote so as to bring attention to something interesting.
Sound interesting?
I'd still have trouble seeing a good way to combine commenting, submitting, and voting ability scores into one number though. I might leave them as separate scores.
The current system is just as vulnerable to robots that submit everything from good rss feeds.
Basically, award original motion and value increasing actions by Karma. (Also, as some other comments mention, there should be some risk — cost — involved in taking those actions)
Yes, I'm working with a similar scheme on my site. This rewards people for predicting what will be popular and provides a big incentive to check out the new content.
The basic system is something like a stock market (that's how I initially visualized it). It costs zero to vote on something but if it goes negative then you lose points, positive then you gain. Like "icky" says here, this approach would have some vulnerabilities here since content is rarely voted down (which is good.)
Ultimately, the closer the karma number correlates with how useful the user is being, the more benefit the community will receive through some properties of game mechanics. Right now, that correlation isn't very high.
Without some sort of constraint the only thing keeping people from upmodding continuously and rampantly is... RSI!
The core idea is that upmods and downmods need to embody some sense of value. Without a buy-in cost, the only contributing source of value is 1-vote-per-story exclusivity. Without a buy-in cost, nothing is stopping me (edit: or someone else, hah. I don't mean to implicate myself for the sake of rhetoric) from getting a bot together to upmod every single story essentially nullifying the benefit of my vote and still maximizing the return.
The 3rd one concerns the karma system : makes a quota of free comments each day then the additionnal ones cost some karma to post. Sometimes your silence has a greater value than a new comment.
The system could group users based on their voting patterns and account for these grouping when computing a point score of the submission for specific user. For example, if A, B and C are voting up and down same articles, then if A votes something up, this should mark it up in B's and C's view in some fashion.
It is a form of a dynamic interest profiling, which should increase per-user quality of the top articles. This way karma can be an indication of how compatible you are with a general population .. though it wouldn't be karma strictly speaking.
Thoughts ?
I think an exponential decay on how much karma you get based on upmods per second would be sufficient to avoid people rampantly upmodding. It takes time to read an article and its comments, and it seems like measuring that would be an effective way of preventing rampant abuse.
Of course, people could just write scripts then to upmod all the time once they figured out the optimal delay. Then there would be an entirely different form of karma inflation to fight. Like spam, it seems to be a never ending battle.
In any case... is it that difficult to reverse-engineering the points system? For those who care...
Or perhaps a limit on upmods/downmods to a certain amount per submission viewed?