Disclaimer: I have not read any literature on the economics of prediction markets, and I know nothing about the mechanics of Polymarket/Kalshi.

I would imagine that in theory, everyone thinks they have the best information at the time, something like:

House: "Odds that X happens? We'll put $1 on both sides to get it started. 50/50."

Someone comes along: "Oh dang, I'm definitely more than 50% confident that X is happening. Let me put $1 in." Now it's 67:33.

Someone else comes along: "Oh I'm more than 67% confident X is happening, let me put $1 in." Now it's 75:25.

And of course, you get people going: "I'm more than 25% confident that X is _not_ happening, let me put $1 in!" And now it's 60:40.

The murky part, I would imagine, comes when the odds and the payout actually act as something that influences the outcome, but in perfect theory-land, if everything goes as planned, this should move the odds to the most informationally-accurate measurement, which should, in theory, benefit observers by making this measurement public.