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.