Everyday Peeple

Forget Parcheesi. Bayesian games are proper fun for the whole family. In a Bayesian game, one (or both) players doesn't know what type of person the other is. This is important because play will vary depending on the nature of your opponent. If you face the decision about whether or not to be the tough guy in a barroom confrontation, it's in your best interest to know whether or not the other dude can mop the floor with you. By reverse induction, it's therefore wise in such situations to signal to others that you're tough, or crazy or dangerous. This is true whether or not you're actually a badass. Acting tough in a barroom brawl game is therefore a pooling equilibrium. Contrast this with, say, the automobile market. A '31 Duesenberg Phaeton gets a motorist from Point A to Point B roughly as effectively as an '86 Yugo (provided you can find one still in working condition), yet for the price of one Doozy ($1.4M at auction), you can buy close to 1500 Zastava Korals (again, provided that many exist in the wild. A Dusenberg isn't 1500 times as efficient at transporting humans, so the price reflects something else about the owner. This is a separating equilibrium.

For most existing Bayesian-type games, we have pretty well-established heuristics about what to expect and how to play. We know when to be sincere (the marriage game) and when to be strategically dishonest (the dating game). But that's for well-established, properly-understood situations.

Meet Peeple, aka Yelp, but for humans. Yes, yes: the idea was stolen from either Dinosaur Comics or SMBC, depending on whether you ask Ryan or Zach (I do so hope they duke it out in the comments [they won't]). So far, most of the press has been pretty unkind. It's easy enough to imagine anonymous vandals ruining moderate-to-high profile people's reputations with relative ease and impunity. If this happens, the sensible response from folks with chronically bad ratings is to poison the well, so to speak. That is, if a separating equilibrium arises, where actually awful people tend to accumulate negative reviews, then those awful people will have an incentive to sock-puppet up and leave piles of negative reviews on the profiles of their rivals, thereby creating a pooling equilibrium where everyone looks awful. Sort of like Encyclopedia Dramatica, but with less fun and imagination.

Still, I think it'd be a nice experiment to see what sort of Bayesian game arises. Recall that some Bayesian games have no pure dominant equilibrium strategy. If I were a betting man, I'd say that there would be some chaos (pooling eq) among second-tier culture warriors, but most other folks would enjoy more-or-less accurate ratings. If I were a betting man.

Edit: the above assumes that the proposed venture isn't merely a silly hoax (pr ~.6)