The unreasonable ineffectiveness of mechanism design

Eugene Wigner famously wrote of the unreasonable effectiveness of mathematics in describing the physical world. It seems crazy that a few simple equations suffice to predict the motions of everything from atoms to cannonballs, from planets to galaxies. That we can extrapolate from a few measurements and a few simple principles and build a novel kind of bridge or skyscraper that won’t fall down. The physical world is unfathomably large and complex and chaotic, yet it seems to turn into putty in the hands of a few monkeys armed with an equation or three.

How different is the social world! Mechanism design, our best attempt to derive simple principles for the social and economic world, gives us simple methods for how to allocate scarce resources in a socially optimal way (the Vickrey-Clarke-Groves mechanism, a generalization of second-price auctions); for how to solve all tragedies of the commons (dominant assurance contracts); for how to aggregate information optimally despite incentives to lie (prediction markets). It’s hard to name a social problem that couldn’t be solved in principle by one of these three mechanisms alone, let alone the thousands more that have been invented in the last fifty years. Yet - our social problems persist, and all these beautiful mechanisms are almost nowhere used in practice. One of the central questions the Institute for Mechanism Design much investigate is: why? What explains this unreasonable gap between the enormous promise of mechanisms and the tiny scope of their real-world application?

The answer can’t just be “no-one has tried”. Noone has better incentives or resources to develop an efficient market in attention than Google; yet even Google has long abandoned the simplest and most robust incentive-compatible mechanism we know of, second-price auctions, and replaced it with first-price auctions which we know to be less efficient - because advertisers prefer them, and even Google’s near-monopoly market power was not enough to force them to switch.

We know the mathematics of the mechanism designers are solid; the problem must rest in their assumptions. The standard assumption to dispute is that of “homo economicus”, but I think that’s a red herring - people are actually quite good at discerning their best interests when real resources are at stake. There’s some other way in which mechanism designers are mis-modelling people or the competitive ecology of extant mechanisms into which new mechanisms much be deployed. Identifying and correcting these missing assumptions is the best way to unleash the unreasonable power of mathematics in service of the social good.