Hacker News Reader

The naive implementation would create an incentive to go to the "new" page and indiscriminately upvote everything.

Sounds like a karma singularity.

The solution is to have a karma cost for voting on a submission/comment that isn't upvoted enough (for example doesn't make the 40th percentile) or to finally add downmod for submissions and make you pay for supporting something the community rejects.

I see this as being a pretty good way of doing this, too. Uncertain stories, stories with little Karma, are riskier business and are therefore more expensive to support, but if they're good stories the payoff can be lucrative when large parts of the community agree.

Just make the points awarded the average number of points per 100 items gained through tel's system. Better?

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.

I've considered doing something like that.

that's like amie.st, where you get cash for recommending cheap songs.

Amie.st sort of does this. So does thesixtyone.

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)

I thought I heard somewhere that your vote gets more "powerful" if you were one of the early voters on a story that garnishes a significant number of votes. Was I mistaken?

I think the old bellwether award on Reddit somehow fed into the recommendation system. I'm pretty sure that stuff isn't turned on here yet though.

> 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.

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.

Any site like this is designed so that upvoted content is seen by a much larger number of people than down voted content (which is, after all, pretty much the point). So you're going to get a lot more upvotes than downvotes, making rampant upvoting still strategically sound. You could weight negative positions more than positive ones, or you could also have a buy in cost. Buy in cost does provide a disincentive to voting which probably isn't desirable.

Disincentives to voting give more abstract value to an upmod. Right now the only thing that gives an upmod value is that you can only upmod a story once.

Without some sort of constraint the only thing keeping people from upmodding continuously and rampantly is... RSI!

We're still talking about a single upvote per article. Its just that if you upvote every new article you'd end up ahead of the game whilst adding nothing of value.

Bingo, I was being hyperbolic there.

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.

We shouldn't write base on karma, we should write because we have something valuable to say.

But karma tells you whether what you wrote is valuable here.

I posted 3 suggestions yesterday that went unoticed : http://news.ycombinator.com/item?id=133618

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.

Here's an idea. Please bear with me for a moment, I'll explain how it's related to the karma.

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 agree. I personally don't really want to listen to people that care immensely about their karma scores.

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.

I'm new here, but it seems this site is less about karma-whoring that digg and reddit.

In any case... is it that difficult to reverse-engineering the points system? For those who care...

It is much better than digg or reddit, in my limited experience. Karma mongering still happens though, and it will probably increase with time as more users are added unless karma mongering is carefully considered in the design.

I think the vote-weighting system is a better solution to the same problem, but PG hasn't turned it on (nor completely implemented it, as of arc2).

I have a theory that drive-by modders who only read short posts cause one-liners to out-karma thoughtful posts. Without seeing the data, it's hard to know, but I'm wondering if a penalty for that voting pattern would be worthwhile. The odds of learning something in a 2 word post that justifies my time here seem lower, comparatively.

Sounds like an interesting idea, but what is to stop someone from just upmodding every comment in a thread?

That could be solved by using a suggestion from another thread. Basically you have to spend some karma to give karma out. And once you upmod one thing, it will cost more to upmod something else soon after that. This system might have to be seperate from the karma, but you could pay karma to get points in the other system. It would definitely need to be refined.

This is an interesting idea. It's a variation of intelligent betting. However it seems that mods with mainstream interests will end up with huge karmas, while those interested in more quirky subjects will not get any.

A limit to how many upmods/downmods a day could help minimize something like that...

Or perhaps a limit on upmods/downmods to a certain amount per submission viewed?

You could let people bet karma on stories on the new page...

yeah that's definitely gameable but i think there's some merit in your thoughts -- it certainly follows a better outline of what a meme is all about.

Ask YC: Voting Karma