Curation “gamed” through unvoting?
I was thinking about the feature of unvoting and its game theory extensions.
1. No unvoting
Supposing that no unvoting exists, then someone who cast a vote on an article which was provably fake or copied, cannot take it back. So in this case, there is plenty of incentive to cheat the platform through fake content. There is also fear of voting / reservations to upvote legit content without doing due diligence (too much effort, curator might skip voting altogether).
2. Unvoting implemented
If unvoting does exist, then there are two main cases.
a) After unvoting, the voter gets back their voting power
b) After unvoting, the voter doesn’t get back their voting power
In case (b), the voter/curator would be penalized if he acknowledged his mistake. Thus it would be in his interest to preserve an upvote, even in fraudulent content – otherwise he would lose money.
Case (a) is where it gets interesting. Case (a) is someone unvoting and getting their voting power back.
Supposing curation reward split takes place between 15m and 30m after a post is created, then we have the following situation:
- Curator votes in the 15 min mark
- Curator sees the votes accumulated till the 30 minute mark or later
- If enough rewards have been accumulated he retains the vote, otherwise he unvotes so that he can vote something else with better rewards
The process can probably be automated through a script – which makes it more problematic.
3. Fixing the issue
I tried to think how one can have solution 2a (unvoting allowed / user gets back his voting power) but without the gaming.
Disallowing unvoting isn’t good – it’s a regression. Allowing unvoting but not giving the voting power back is also problematic due to incentives because then why would someone unvote. So, what to do?
I ended up with the following duct-tape solution: Limit the number of unvotes one has during 24hrs. Perhaps one unvote per 24hrs is OK. Or it could be slightly variable for witnesses/large curators who cast a lot of votes, allowing them to make a few more errors.
If anyone has any better idea, or if my rationale is flawed somewhere, please contribute to the discussion.
Thank you...
There is no useful gaming here. Removing a vote does not benefit you in any significant way, other than: a) making the recommendation feature work better for you, and b) slightly increasing the payout on other posts (generally insignificant).
More specifically, voting is in effect spending the perishable resource known as vote power (which recharges from empty to full in five days). Removing the vote does not give you back the voting power you originally used, so you end up worse off than had you left the vote there (assuming the vote would have earned a reward >0).
Obviously this corresponds to your case (b), which as you say is the less interesting, and perhaps means that people won't remove erroneous votes. Even there, the incentive is to be a bit more careful and not make too many errors (which could mislead others about the state of voting). This is not necessarily a bad thing. The designers of the platform identified the same issues you did with (a) so they did not use it.
Great idea! I hope you will get more upvotes!