Combinatorial Prediction Markets
Sep. 2nd, 2008 09:48 pm![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
guslThe video of Robin Hanson's UAI lecture only has 20 views. I'm hoping this blog post will correct this. I found this lecture quite a good overview of his ideas on bet-based epistemology. Domingos got very enthusiastic about this, and they seemed to chat for many hours afterwards.
Robin is interested in algorithms for keeping a consistent belief state (i.e. immune to Dutch book), which is a hard computational problem when you have lots of variables. If you had a powerful computer, you could set up such Dutch books for profit, essentially selling consistency to the computationally-deprived market subsidizer (i.e. the party interested in the information).
From the other side, I'm also hoping that scientists begin to automatically elicit knowledge this way. It sounds like a crazy idea that just might work.
Robin is interested in algorithms for keeping a consistent belief state (i.e. immune to Dutch book), which is a hard computational problem when you have lots of variables. If you had a powerful computer, you could set up such Dutch books for profit, essentially selling consistency to the computationally-deprived market subsidizer (i.e. the party interested in the information).
From the other side, I'm also hoping that scientists begin to automatically elicit knowledge this way. It sounds like a crazy idea that just might work.