I think prediction markets are limited by the fact that most major societal issues are complex and long term. Prediction markets are bad at both. It's very difficult to create markets that are about complicated issues and they are hard to fairly resolve, and long term markets get wacky odds due to the time cost of money (even fake money) and longshot bias.

It is often more important to understand the causes of outcomes than the outcome itself. For instance, we don't want to predict "will there be another Jan 6 type insurrection in the US" - we want to predict what would cause it.

Finally, rationality doesn't fix problems. Power fixes problems. And so long as power is undemocratically concentrated (due to economic inequality and other forms of inequality) in the hands of a few, and our society is moving towards less democracy - even if we do identify the causes of major problems - we won't be able to solve most of them.

Patrick McKenzie recently wrote that there's an exchange rate between competence and power, and any system that demonstrates sufficient competence will inherently become powerful, and thus subject to all the forces in the world that have opinions about what they want you to do with that power.

I worry that someone similar might happen with the "play money" thing. Manifold only gets away with what it does because it's play money that doesn't matter, but to the extent that people actually care about mana, that becomes proportionately less true.

We already had a case where someone spent months trolling a different site just because they cared about their mana so much, and all the insider trading and the whale vs minnows fiasco and so on, and that's just with weird internet points that at best convey a bit of status among a group of nerds in the valley and can't actually be used for anything. The more people care about mana, the more they'll care about correct resolutions and fraud and so on, and the more that Manifold will be inherently regulated.

My second favorite part is the happiness James has to describe NOT dealing with "crypto" or "real money" several times throughout the talk. Every time he says "We're not doing crypto", you can tell he's extremely glad to not be doing that and dealing with regulations, etc.

My favorite part was the discussion about attracting people who are really interested in a specific niche and possibly don't care as much about the overall site outside their niche. Subject matter experts (or superfans) sticking to their area of expertise and owning that section of the site via question creation and dashboard management is exciting because it gives the rest of the site's users an insight into that area.

Using market values at fixed time intervals to calculate Brier scores seems weird to me, because you have a huge number of extremely correlated data points, and the number of data points (and hence weighting in the final result) is also correlated with the outcome.

Many markets take the form of "well X happen before Y date", and so YES will resolve immediately, while NO will only resolve at close and hence there will be more "observations" in the data set. Hanania got nonsense results when trying to calculate the "calibration" of the Salem contest partly for this reason.