Does this kind of "pay to vote" poll effectively solve the problems of other polling strategies on Manifold?
13
16
100
resolved Oct 5
Resolved
N/A

Lots of people create polls on Manifold. They're usually one of two types, both of which have problems:

  • A Keynesian beauty contest, where the option(s) with the highest market share wins. This means the incentive is to bet on the option you think will win, not the one you want to win, making the outcome unlikely to accurately reflect people's preferences.

  • A poll in the comments, where people type out their vote and the creator counts them up at the end. This is a proper poll, but there's no benefit to running it on the Manifold platform; you may as well post a link to a Google form or Twitter poll. This type of poll doesn't allow people to express how much they care about a certain option, nor does it reward better predictors. It's also a hassle for the market creator to have to manually count up all the responses, and sometimes people leave ambiguous comments.

I think there's a better way to run polls that solves all of these issues and is in keeping with the ethos of prediction markets: Pay-to-vote polls.

The way they work is simple: People vote by betting in the market, but the market resolves to an option that is not part of the poll. Anyone who voted is guaranteed to lose their investment. This means that the incentive is to only vote if you actually care about the result, and to set the amount of your vote in proportion to how much you care.

There's a flaw in this system though, which is that there's still a small incentive to predict the most popular outcome; you can bet on it early and then sell your shares later once it becomes more popular. A solution is to N/A the market rather than resolving it. This also means people get their mana back and there's much less cost to participating in the market. There is still some opportunity cost of having the mana tied up for the duration of the poll, which I expect will still be enough to stop people from voting if they don't care about the outcome.

(If Manifold ends up deciding to implement native polls, they could get the best of both worlds by implementing a new type of poll market that's the same as regular markets except it's impossible to sell shares after they've been purchased and the mana spent on the poll just gets burned once it closes.)

This market is an example of the N/A version of this system. If you think this is a good system and people should adopt it for their own polls, buy YES. If you don't like it and think people shouldn't use it, buy NO. This market will resolve to N/A after it closes.

(It's not a great example, because the outcome of this poll doesn't actually determine anything, so there's little incentive to vote in it. Consider it more of a proof of concept.)

Get Ṁ200 play money
Sort by:

Ended at 42%, which seems reasonable given the thoughts posted in the comments. I'll probably try this style at least once again for something where the poll results actually matter and see how that goes.

may as well post a link to a Google form or Twitter poll.

A great idea! Let's try it.

  • A poll in the comments, where people type out their vote and the creator counts them up at the end. This is a proper poll, but there's no benefit to running it on the Manifold platform; you may as well post a link to a Google form or Twitter poll

I disagree, I think that polling works well on Manifold (bare the second part of your point, which is more of a technical issue). Having a market allows to common-knowledge what we think the consensus is going to be, while keeping it separate from a Keynesian beauty contest.

I'm in favor of more ways to spend mana on this platform than just tips though. I think paying to vote is a good idea if you can only cast one vote

  • A poll in the comments, where people type out their vote and the creator counts them up at the end. This is a proper poll, but there's no benefit to running it on the Manifold platform; you may as well post a link to a Google form or Twitter poll



@JoyVoid Oops, got tricked by the editor, didn't know that ctrl+enter would just send immediatly

I think we should just implement polls because people keep trying to use markets for this purpose.

Proposal: quadratic poll. First vote is free, 2nd vote costs 1 mana, 3rd vote costs 2 mana ...

@Sinclair If votes are discrete things and you have to click for each vote, I think that's too cheap. I'd do at least 0-10-20-...
But if you just enter a number in a textbox then it doesn't matter because all parabolas are the same.

@Sinclair Should probably be capped though? I don't like that users with way more mana are able to override the consensus (especially if Manifold transitions to a real money market)

predicted NO

@JoyVoid As we move to a world where human labor is worthless it will be increasingly important to let the rich redistribute their wealth by winning meaningless prizes on social media.

@MartinRandall It's not really redistribution if the mana just goes to Manifold. But what if the size of the daily bonus was proportional to the mana invested into polls? 🤔

predicted NO

@Yev Oh, I thought it would go to the person who created the poll. It could also go to other voters. Not sure what's best.

I like the idea that mana bonuses are a dividend based on how much Manifold is earning from adverts or sponsorships or votes or whatever.

@MartinRandall I think going to creator makes the most sense?
Like I expect people to use quadratic for actual votes over public goods like "what should I draw next?"

For info/opinion polls, I don't think voters should have to pay to vote at all - they're the ones doing work after all.

bought Ṁ41 of NO

Why yes, I am willing to lock up a bit less than m200 for a week to say that this is not a great idea.

But at least it's not resolve to MKT.

bought Ṁ300 of NO

One problem with using a market to poll is that by default, liquidity is terrible. Allow me to demonstrate :)

sold Ṁ204 of NO

Of course, people can set limit orders and stuff, but I think simpler polls would likely do better as a default way to ask questions that don't change over time or don't get a huge amount of interest. Not sure, I think there's still a lot of interesting questions about polls vs markets.

sold Ṁ53 of YES

This seems like the same problem faced by regular markets, and I expect whatever solution is arrived on for those will likely work for polls too.

@IsaacKing In a regular market people have a financial incentive to correct the price.

@Yev People who care about the outcome of a poll have the same incentive.

@IsaacKing Not really. In a prediction market everyone has an incentive to move the price towards the true probability. Here some people have an incentive to move it up and others have an incentive to move it down. But no one has the incentive to move it towards the "true answer".

Featuring because this is an interesting and well-written proposal!

I do tend to think that instead of paying-to-vote, most polls probably should be pay-you-to-vote -- by voting, you're contributing information and deserve a mana perk. One of my fun side project ideas is to have a section on "Get M$" where instead of paying money, you can do microtasks on Manifold like voting in polls or commenting in markets in exchange for M$; inspired by Amazon Turk and the sketchy websites where teenage Austin would fill out online surveys to get Nexon Cash (Maplestory's premium currency).

@Austin That'll just incent people to vote randomly and leave low-effort comments. Scratch that. It'll incent bots to vote randomly and leave low-effort comments.

@Austin And we already have a way to reward good comments: tips.

@MartinRandall suggested something similar on my market - the suggestion was to use quadratic voting. https://manifold.markets/jack/what-are-the-best-ways-to-operate-a#answer-7

I have a better solution

Tip this if you think my poll is better than Isaac's

Tip this if you think my poll is worse than Isaac's

All tips will go to charity.