Abstract

The revelation principle suggest that a theoretician searching for an optimal mechanism need only consider ones with truth telling as an equilibrium.  This optimal mechanism may not be practical, and corresponding non-revelation mechanisms are generally complex and highly dependent on details.   In this talk, I will give a theory for the design of simple non-revelation mechanisms (e.g., with "first-price” payments).  These mechanisms (a) will have unique and easy to find Bayes-Nash equilibria, (b) are amenable to inference, (c) are easy to optimize over, and (d) approximate the optimal mechanism in general single-dimensional environments.  These result come, respectively, from Chawla and Hartline (2013); Chawla, Hartline, and Nekipelov (2014, 2015); and Hartline and Taggart (2015) which will be surveyed.

Video Recording