Simon Willison wrote up six cool things you can build with OpenID and reiterated some of his very good ideas about a simple trust network.

A site-to-site trust network should use degrees of trust rather than simply asserting membership in a "white" or "black" list. Perhaps a set of common violations of trust - from spamming to griefing - and common traits of a trustworthy user - from contributing to helping new members - could be agreed-upon. Individual web sites would then publish data about any violations made by OpenID users.

It would be up to violation aggregators to determine if each user's behavior represents an isolated incident or an ongoing pattern. Sites would also be accumulating a reputation of their own for trustworthy reporting.

Eventually, though, I believe that all of this IDing and trusting is going to evolve into a reputation-based currency :)