Economists' analysis of mechanism design under incomplete information relies on strong assumptions about economic agents' beliefs and common knowledge of those beliefs. This talk will survey an approach to mechanism design that aims to relax those assumptions and thus derives more robust mechanisms. The relationship to prior-free optimal mechanism design in the computer science literature will be discussed.