I've been trying to get places like Manifold or Metaculus to add latent variables to their prediction systems. As a brief summary, latent variables would allow you to make bets and look up predictions on abstract concepts such as "How well will Ukraine do in their defense against Russia?" by linking together many concrete outcomes/indicators.

Yesterday, Nathan Young suggested that I should design the interface for it. I think that was probably a good idea, because I got the impression that I wasn't really explaining the concept properly, and also because I have a bunch of knowledge of latent variable models that might be useful to get implemented.

Here's a link to the mockup I have so far. I apologize for it being rough around the edges, and for the poor code quality; it's just something I quickly threw together.

Now for the designer request: The interface in my mockup is probably still confusing. I hope it makes it clearer what sorts of predictions I want to introduce the possibility for, especially if one plays around with the interface a bit. But ideally the meaning of all the options and graphs should be intuitive even for new users.

So if there are any designers in the audience here, I would be interested in hiring a designer to spend some time with me designing an interface that would be easier to understand.

Also if there are any programmers who are familiar with the Manifold codebase (including programmers working at Manifold), I would be interested in paying part of the development costs.

(I have a max budget of $10k for this.)

Beyond this, I would be interested in whether there is any discussion or ideas for latent variable prediction markets, including but not limited to ideas for ways to make it more intuitive to people.