Your points are very well taken - there is most certainly a trade off in
computation and pay off. But let's ask ourselves: how do real world people
make this trade-off? Having just done some consulting work, I know that
organizations routinely make such computations. Before hiring a consultant
or devoting resources to solving a problem, managers frequently ask: will
the cost of finding the solution get me a pay-off? The managers have the
collective intelligence of the organization (other workers, books,
reports, corporate archives, etc) that will help them answer this
question.
But in other cases, framing the issue as a computational cost issue is
misleading because it assumes that actors will engage in an even harder
computation. Most beginning chess players seem to learn in a fashion
similar to a nueral net. For example, they aim to acheive a state
("win") by doing certain actions ("control the center"). If they fail to
control the center, they loose a bunch and then shape up. In a broad
sense, they are utility maximizers, but in a narrow sense they are not
doing anything resembling the actors depicted in, say, Krep's
Microeconomics text. They are in a feed back loop with their environment,
not engaged in an utility maximization computation.
Fabio
