Live data from Hacker News

Ask HN: Do social sites with karma need a decay function?

news.ycombinator.com

21–30 of 33 posts

Re: Ask HN: Do social sites with karma need a decay function?

#21

I think a good idea would be to treat karma the same way that tennis rankings are counted: Only the past 52 weeks are counted. So if you gain 1000 karma in a year, then don't participate, your karma will be 0. That will encourage continual good behavior.

You assume that karma encourages good behavior. If that were the case, then I think this would be a good idea.

I couldn't agree more. Karma encourages karma getting behavior. Sometimes that overlaps with good.

Re: Ask HN: Do social sites with karma need a decay function?

#22

I think a good idea would be to treat karma the same way that tennis rankings are counted: Only the past 52 weeks are counted. So if you gain 1000 karma in a year, then don't participate, your karma will be 0. That will encourage continual good behavior.

That encourages lock-in. What if my wife had a baby and I went offline for a year because I didn't have time. Would that make my opinions suddenly invalid?

Tennis rankings are based on physical form. Physical form is something much more likely to change year to year than intellectual form.

Someone might be out of date with the latest topics, consensus from the community, etc but it doesn't imply their thinking apparatus or world view has dramatically altered.

Re: Ask HN: Do social sites with karma need a decay function?

#23

I think a good idea would be to treat karma the same way that tennis rankings are counted: Only the past 52 weeks are counted. So if you gain 1000 karma in a year, then don't participate, your karma will be 0. That will encourage continual good behavior.

Karma ensures popular behaviour, not "good".

If that's what you're gunning for, it's a great idea. What's popular is only dependent on your users and their tastes, and this can change over time.

Re: Ask HN: Do social sites with karma need a decay function?

#24
To be honest, I think the up/down notion of karma is the problem.

I'd prefer to see words being attached to comments instead of numbers. "Funny", "Insightful", "Accurate", and their opposites could be the method of rating, with "leader boards" showing the funniest, most insightful etc.

The notion of me as a user being "worth" #### points would vanish, and it would be replaced with a general "run4yourlives: slightly funny, not often accurate" descriptor based on the ratings of my peers.

Slashdot tried this, but they kept linking it back to numbers, which I think is where it failed. The idea should be to remove the notion of a numerical worth altogether.

Re: Ask HN: Do social sites with karma need a decay function?

#25
While you can probably improve a karma system with decay or a more complicated algorithm, there's something to be said for having a system simple that your users understand. I'd make a wild guess that karma wouldn't be as interesting to users if they can't figure out how their actions directly affect their score.

Re: Ask HN: Do social sites with karma need a decay function?

#26
The answer depends on what problem you think the karma-delay function is intended to solve and whether or not there is evidence that this problem is real.

I think the only place a decay function makes sense is in providing historic views of the leaderboard.

Re: Ask HN: Do social sites with karma need a decay function?

#27
One thing I've thought about in the past is should an upvote/downvote result in a percentage gain/loss instead of absolute.

The thing is, if you've only got 3 karma, and someone downvotes you, that's a big deal. But if you've got 3000 karma and someone downvotes you, it's not. Maybe a vote should result in a 1% gain/loss.

On the other hand, it probably doesn't matter that much :)

Re: Ask HN: Do social sites with karma need a decay function?

#28
post #21

Earlier quoted context omitted.

You assume that karma encourages good behavior. If that were the case, then I think this would be a good idea.

I couldn't agree more. Karma encourages karma getting behavior. Sometimes that overlaps with good.

Agreed. Karma might not be the best system for encouraging good behavior... But does anyone have any other ideas?

It's also slightly more difficult because what constitutes good behavior isn't always black and white and isn't even always agreed upon.

Re: Ask HN: Do social sites with karma need a decay function?

#29

To be honest, I think the up/down notion of karma is the problem. I'd prefer to see words being attached to comments instead of numbers. "Funny", "Insightful", "Accurate", and their opposites could be the method of rating, with "leader boards" showing the funniest, most insightful etc. The notion of me as a user being "worth" #### points would vanish, and it would be replaced with a general "run4yourlives: slightly f…

I like this idea a lot.

Re: Ask HN: Do social sites with karma need a decay function?

#30
post #10

Earlier quoted context omitted.

> Why should people work so hard to gain something that is crumbling into dust all around us? Like money which will decay with inflation, or perhaps the regard of people who will be dust in a few decades? > If karma is constantly going down anyway it will confuse the issue If that's a problem, the site can provide the 1st derivative (ignoring decay). A bit like music charts ("at number 10, up 5, it's..."). Some sites…

Then again, I don't see what the utility of decaying karma is personally. It smacks of "what have you done for me lately". I think it's OK to respect people for past glories. Incidentally, this is one of the reasons why I don't participate on Perlmonks. There are a lot of people there that have "been around forever" and answer questions as though it's 1999. I really enjoy reading the "don't use Moose, it's bloat" thr…

Interesting. Perhaps that's perl's main problem. Ruby (or python) snatched up that generation of hackers since they could make more of a splash. Someone commented on reddit a few weeks back "perl used to have a really cool culture"

OK - I can now see a benefit of favouring recent contributors. It helps stop ossificiation.

Post reply on HN