[Paper Review] Complexity of Mechanism Design

Vincent Conitzer, Tüomas Sandholm|arXiv (Cornell University)|May 28, 2002

Auction Theory and Applications18 references249 citations

TL;DR

This paper proposes automated mechanism design for preference aggregation in multiagent systems without side payments, showing that deterministic mechanisms are NP-complete to design for both dominant-strategy and Bayes-Nash implementation. However, allowing randomized mechanisms reduces the complexity to polynomial time, enabling better social welfare through strategic uncertainty without sacrificing truthfulness.

ABSTRACT

The aggregation of conflicting preferences is a central problem in multiagent systems. The key difficulty is that the agents may report their preferences insincerely. Mechanism design is the art of designing the rules of the game so that the agents are motivated to report their preferences truthfully and a (socially) desirable outcome is chosen. We propose an approach where a mechanism is automatically created for the preference aggregation setting at hand. This has several advantages, but the downside is that the mechanism design optimization problem needs to be solved anew each time. Focusing on settings where side payments are not possible, we show that the mechanism design problem is NP-complete for deterministic mechanisms. This holds both for dominant-strategy implementation and for Bayes-Nash implementation. We then show that if we allow randomized mechanisms, the mechanism design problem becomes tractable. In other words, the coordinator can tackle the computational complexity introduced by its uncertainty about the agents preferences BY making the agents face additional uncertainty.This comes at no loss, AND IN SOME cases at a gain, IN the(social) objective.

Motivation & Objective

• To address the challenge of designing nonmanipulable mechanisms in multiagent systems where agents may misreport preferences.
• To overcome limitations of classical mechanisms that require unrealistic assumptions like quasilinearity and side payments.
• To study the computational complexity of mechanism design when mechanisms are automatically generated for specific settings rather than applied generally.
• To investigate whether randomization can reduce computational complexity while improving social welfare.
• To demonstrate that randomized mechanisms can outperform deterministic ones in both efficiency and computational tractability.

Proposed method

• Formulates mechanism design as an optimization problem with inputs including agent types, prior distributions, and social objectives.
• Defines deterministic mechanism design as selecting a fixed outcome based on reported preferences, constrained by nonmanipulability.
• Introduces randomized mechanism design where outcomes are chosen stochastically, using probability distributions over outcomes.
• Uses linear programming to solve the randomized mechanism design problem efficiently under both dominant-strategy and Bayes-Nash implementation.
• Applies a transformation to convert any mechanism into a nonmanipulable one with identical performance, justifying the focus on nonmanipulable mechanisms.
• Employs a proof by contradiction to show that randomization strictly increases expected social welfare in some cases, implying P ≠ NP if it did not.

Experimental results

Research questions

• RQ1 Is the problem of designing nonmanipulable deterministic mechanisms computationally hard when side payments are not allowed?
• RQ2 Can randomized mechanisms make the mechanism design problem tractable where deterministic ones are NP-complete?
• RQ3 Does randomization in mechanisms lead to higher expected social welfare compared to deterministic mechanisms?
• RQ4 Can the computational complexity of mechanism design be reduced by introducing uncertainty into the mechanism itself?
• RQ5 Is there a fundamental gap between the performance of deterministic and randomized mechanisms in nonmanipulable settings?

Key findings

• The deterministic mechanism design problem is NP-complete for both dominant-strategy and Bayes-Nash implementation when side payments are not allowed.
• Randomized mechanism design is solvable in polynomial time via linear programming for both dominant-strategy and Bayes-Nash implementation.
• Randomized mechanisms can achieve strictly higher expected social welfare than any deterministic mechanism in the same setting.
• In a 3-outcome, 2-agent example, a randomized mechanism achieves 5.5 expected social welfare, while the best deterministic mechanism achieves only 5.
• The existence of such a gap implies that P ≠ NP, assuming that randomization cannot increase social welfare in the absence of computational complexity.
• The paper demonstrates that introducing uncertainty into the mechanism can reduce computational complexity without sacrificing social welfare, and sometimes improves it.