If you had perfectly rational people, all uniform in their rationality, one might want to distribute that trait along the bell curve, such that most were "mostly" rational, with a few highly irrational individuals and a few highly rational ones. It's like the ant colony... send agents out along all the possible paths simultaneously, see who comes back with cake.

Even in that model though, they all average out to "mostly rational". People do too.

What about this argument requires that people be individually rational rather than on-average rational?