Mechanism Design

Ordinary game theory hands you the rules and asks a question: given this game, how will self-interested players behave? You find the best responses, locate the Nash equilibrium, and predict where everyone gets stuck. But the most valuable question in all of strategy runs the machine backwards. Suppose you don’t like where the game settles — the auction that overpays, the committee that lies about its budget, the two heirs squabbling over one house. You don’t have to accept the equilibrium. You are allowed to change the rules. And if you change them cleverly, you can make the outcome you want be the very outcome that self-interested players walk themselves into, with no one to police them and no sermon required. That inverted craft is mechanism design, and it is the closest thing decision-making has to an engineering discipline.

Mechanism design is game theory in reverse: rather than taking a game and predicting its equilibrium, you fix the equilibrium you want and then design a game that produces it. The economists Leonid Hurwicz, Eric Maskin, and Roger Myerson shared the 2007 Nobel Prize for turning this into a rigorous science, but you have already met its logic in disguise — every time a parent settles a fight over the last slice of cake by saying “one of you cuts, the other chooses”, they are doing mechanism design. The rule is built so that the cutter’s own greed forces a fair split. The genius of the field is that it never asks people to be better than they are. It asks the rules to be smarter.

This is an advanced strategy course, and it assumes you have already met Nash equilibrium (best responses, dominant strategies, why stable outcomes are so often the bad ones) and incentives (that people respond to the rewards and penalties they actually face). From there it builds the design mindset floor to ceiling. We start with the great flip — the difference between analysing a fixed game and choosing the game, and why that single move is worth a Nobel Prize. We define the field’s central prize, incentive compatibility: a mechanism is incentive-compatible when honest, cooperative behaviour is itself a best response, so nobody can gain by gaming it. We work the two canonical machines until they are obvious in your hands — “I cut, you choose”, where self-interest manufactures fairness, and the second-price (Vickrey) auction, where bidding your true value is a dominant strategy (and we’ll see exactly why the first-price auction tempts you to lie instead, and why eBay’s proxy bidding is a second-price auction in a trench coat). We’ll state the field’s deep organising idea, the revelation principle — if any mechanism can reach a good outcome, then a simple honest one can reach it too — and then open the toolbox: a Pigouvian tax is a mechanism, and so are deposit-refunds, cap-and-trade, and the matching markets that place kidneys with patients and children with schools without any money changing hands at all.

And because this is a course about tools, not magic, we will be ruthlessly honest about where the craft bites back. You cannot design away every bad incentive: the impossibility results (Myerson–Satterthwaite, Gibbard–Satterthwaite, Arrow) prove that you often cannot have efficiency, honesty, and a balanced budget all at once — something must give. Mechanisms that look airtight get gamed in the wild (Goodhart’s law, unravelling markets), people refuse to participate when the deal is bad for them (individual rationality), and every design quietly assumes players actually reach the equilibrium you intended. By the end you will be able to look at a broken situation and stop asking “how do I get these people to behave?” — the wrong question — and start asking the designer’s question instead: what rule makes the behaviour I want each player’s own best response? That is the difference between wishing a system were better and building one that is.