Skip to content

Preference Revelation

Why It Matters

Almost every coordination problem involves information you don't have. Who actually values this resource most? Which project proposal has the best expected return? What's the real cost of delivering this feature? If you ask directly, people lie strategically. Effective mechanisms must elicit truth despite this - either by making truth dominant strategy or by making lies detectable and costly.

The Key Move

When you need information from strategic agents, never just ask. Instead, design the mechanism so the response triggers consequences that make lying costly. Ask: "If this person lies, how do they benefit?" Then structure outcomes so the lie hurts them or the truth is revealed. Alternatively, cross-check reports against observable outcomes or other parties' reports to catch inconsistencies.

Classic Application

Vickrey-Clarke-Groves (VCG) mechanisms. When allocating goods, ask each person their valuation. Allocate efficiently based on reports. But charge each person not their reported value - instead, charge them the "harm" they impose on others (the difference in total value with and without them). This makes truthful reporting optimal because your payment doesn't depend on your own report directly, only on how you affect others. Lying about valuations can't help you.

Surprising Application

Sprint planning in software teams. Developers estimate task complexity. If promotions reward "high output," developers underestimate complexity to appear more productive, then miss deadlines. Better mechanism: historical velocity tracking. Estimates are cross-checked against actual completion time. Systematic underestimation becomes visible and damages credibility. The mechanism makes lies detectable, incentivizing more honest estimation.

Failure Modes

Over-reliance on revelation mechanisms in settings where direct monitoring is cheaper (sometimes just check the facts rather than building elaborate incentive structures). Ignoring computational complexity (optimal revelation mechanisms may be too complex for participants to reason about). Assuming common knowledge of the mechanism (if people don't understand how their reports affect outcomes, the incentive properties fail). Creating mechanisms where small lies are undetectable but large ones aren't (leading to small-scale systematic lying).

Go Deeper

Green and Laffont (1979), "Incentives in Public Decision Making." Clarke (1971), "Multipart Pricing of Public Goods." Groves (1973), "Incentives in Teams." Holmström (1982), "Moral Hazard in Teams" - shows some settings make truthful revelation impossible.