Background
I like the idea of Manifold Markets. Prediction markets should be a great way of accumulating information. It's not so clear that you need real money for incentives because online forums often have people participate for fun.
However, outside of a few forecasting questions that others have made, I don't generally find myself reaching for Manifold when I want to know something. I basically always use other methods such as Google, ChatGPT, Wikipedia, or asking people; even when it comes to questions where prediction markets in theory should shine, such as when we don't have a definitive answer yet but expect to have one in the future.
I think there are multiple reasons for this:
Lack of activity: Predictions are often obviously incorrect, and when they aren't, they often lack documentation or additional info that can be used to understand them.
Difficulty in creating markets: I need to come up with a way to explain the market, create objective resolution criteria, and eventually resolve the market in a way that I think is reasonable.
Lack of informativeness: Most markets are yes/no markets. That's just 1 bit. You can get a few extra bits by having it be multiple choice, but it is still pretty limited. There were experiments with free text options but I mostly didn't dare make such markets (was kind of paranoid about resolution - maybe that would change if I researched more) and even if I did, they seem to be hidden/removed for now (did something happen?).
Resolution criterion
If, during 2027, I start having a class of questions that I regularly use Manifold to learn answers to, this market will resolve YES. As an anchor, I will focus on the above three problems being solved (conjunctively, i.e. being easily able to ask an open-ended question that gets good answers), but I am open to solutions that don't involve these problems being solved (e.g. maybe it is just a skill issue).
Like if there are lots of times during 2027 where I am wondering about some question and I go "I know! Manifold Markets will give me the answer!" and I genuinely believe it, then the question resolves YES. Otherwise it resolves NO. Unless we have some annoying ambiguity that I don't want to count, in which case I might resolve PROB or N/A.
1,000
1.00I keep thinking that something like "Manifold News"/"Manifold Journal" might help, where basically people are incentivized for dumping high-bandwidth information such as personal observations, studies, arguments, etc..
In theory classical prediction markets should achieve this just by buying based on the information, and then publishing the information. But in practice liquidity seems to be too low so it doesn't really work.
However it is hard to think of good ways of rewarding people for dumping news/studies into a stream. Because you would somehow have to quantify 1) its importance, 2) its truth. One of the main options I can see for this would be coupling it to a trading bot somehow. Like you dump some news into the stream, and based on linguistic similarity it suggests some trades which you buy into. And maybe then the news get "auctioned off", by allowing people to bid on getting the payouts from the trades. But it is sort of convoluted and questionable.
@tailcalled I think what is needed first and foremost is just volume of participants. With high enough amount of participants and exposure, what you described as "high-bandwidth information" will also come, in comments. At the end of the day, nothing compells a human more then proving someone else wrong, just look at reddit. The big question is how to enable that the actually insightful comments get pushed to the top and are most visible? I think something along the lines of "this contained useful information" button on a comment, like the "like" button, could come in handy.
Anyways, have you tied out the new multiple choice? the mechanism is actually good now.
@Sinclair You mean because one can also bet NO? ๐ Possibly there's some subtler change I am not aware of.