Rambling Preamble
There's a lot of conversation right now about the MSM competition. I also feel like I've noticed some behavioral changes amongst stackers. That leads me to reflect on what this contest is showing us about the SN rewards structure.
Many of us have written about how the rewards structure here is largely responsible for cultivating this very unique internet community. I've written many times about how people are really short-changing themselves by not taking advantage of the reward incentives: i.e. you can come out ahead by zapping content you like liberally.
There's plenty of room for improvement, of course, and one of the things I most value about being here is that the development team talks with us openly about our ideas for improvements. (I was very interested in experimental econ in graduate school, but don't get to flex my mechanism design muscles in my current job.)
Actual Point
That's a lot of preamble to get to my point. Million Sat Madness seems to be causing stackers to examine the particulars of the reward system more carefully, and that might be a problem.
The reward system Stacker News uses is a form of Keynesian Beauty Contest: our rewards aren't really based on zapping and making the best content. Our rewards are based on zapping and creating the content that other users zap the most (slight oversimplification).
I often say "Outcomes follow incentives" and Keynesian Beauty Contests are known to have a flawed incentive structure (assuming the goal is to elicit honest responses about what people like the best).
When people were being super miserly with their daily zaps, this wasn't an issue. However, I feel like I've noticed a shift in zapping habits that conforms with the expectations of a Keynesian Beauty Contest. If the goal is to zap posts that you think will get lots of zaps, then there are certain people who are a safer bet to zap, like @kepford or @Natalia, and you might not zap other posts even though you liked them. If you're trying to zap the "best" comments, then you might try to zap good responses to @k00b or @elvismercury, since they're really generous comment zappers, but ignore equally good comments left on other people's posts.
Please note that I'm not saying any of those creators shouldn't be getting zapped a ton. I'd still love to see people zapping them more. I drew them from the set of stackers whose work I most value and who I've noticed other people also highly value.
Part of why I think this might be showing up for MSM, but not for the daily rewards, is that the larger sample size is reducing uncertainty in the outcome greatly. On any given day, I might have one of the best performing posts or comments. However, over the course of a month, there's a near certainty that I won't. That means there's a disincentive to zap my content. Or, stated differently, the marginal benefit of zapping my content is less than the marginal benefit of zapping other content of a similar quality.
Possible Solution
There's probably a super clever and elegant idea in mechanism design that solves this incentive problem, but I'm not aware of it.
I really like the idea of rewarding creators more for having the top content over a longer span of time, so I don't actually think the best answer is to go back to daily rewards. I also think the monthly rewards are more fair to international stackers who get slightly screwed by time-zone issues.
What I do think would largely head off this incentive issue is making the incentives more difficult to game. My suggestion is to split the reward pot into rewards over four different timespans:
- 25% rewards paid out daily
- 25% rewards paid out weekly
- 25% rewards paid out monthly
- 25% rewards paid out yearly
That way there's an incentive to zap content you think is good, but not best-of-the-month good, because it still might get you some daily and weekly rewards. There's also more incentive to zap great posts even more, since at the end of the year those rewards might be very large.
I also think some form of randomization could be used to make the daily rewards are a bit more fair across time zones.
I'm curious to hear what you all think.