pg, you could present cleaned data in a Netflix Prize-style challenge. Let the hackers see the patterns in the data (whether bad upvotes are coming from new users, from old users without a lot of karma, etc) and make the prize be XX minutes of your attention (or money).
It seems like a lot of the comments on this thread are asking for more information -- or at the very least working from very different personal experiences.