Bayesian incentive compatibility
Honesty is optimal in expectation, given that everyone else is reporting honestly and given shared beliefs about the distribution of private information.
Often misread asUsing it as though it delivered the same guarantee as dominant strategies, and so not noticing that the honesty rests on beliefs about the distribution and on the others reporting honestly too.
How to see it
Honest on average, if everyone else is too
The weaker and more usable standard: honesty is your best move in expectation, given that the others are also reporting honestly and that everybody shares the same view of how private information is distributed. It can fail for particular draws and still hold overall.
The checkChange the distribution, or let one participant lie, and the incentive to be honest can disappear. That fragility is the price paid for the extra reach.
Try itAsk what happens to a bidder's optimal report if the distribution of rivals' values turns out to be different from the one assumed. Under the dominant strategy standard, nothing. Under this one the whole calculation changes, and that is the difference between them.
Know these first
Easily confused with
Sources
- Myerson, R. B. (1981). Mathematics of Operations Research 6(1), 58-73.
- Myerson, R. (1991). Harvard University Press.
- Krishna, V. (2002). Academic Press.