[Paper Review] Stable Nash Equilibria in the Gale-Shapley Matching Game

Sushmita Gupta, Kazuo Iwama+1 more|arXiv (Cornell University)|Sep 14, 2015

Game Theory and Voting Systems13 references3 citations

TL;DR

This paper introduces and characterizes $P$-stable Nash equilibria in the Gale-Shapley matching game, where a woman's manipulation of her preference list must yield a stable matching under her true preferences $P$. The authors present an $O(n^4)$ algorithm to test whether a given strategy profile is a $P$-stable Nash equilibrium, ensuring manipulations do not create unstable outcomes vulnerable to further exploitation.

ABSTRACT

In this article we study the stable marriage game induced by the men-proposing Gale-Shapley algorithm. Our setting is standard: all the lists are complete and the matching mechanism is the men-proposing Gale-Shapley algorithm. It is well known that in this setting, men cannot cheat, but women can. In fact, Teo, Sethuraman and Tan \cite{TST01}, show that there is a polynomial time algorithm to obtain, for a given strategy (the set of all lists) $Q$ and a woman $w$, the best partner attainable by changing her list. However, what if the resulting matching is not stable with respect to $Q$? Obviously, such a matching would be vulnerable to further manipulation, but is not mentioned in \cite{TST01}. In this paper, we consider (safe) manipulation that implies a stable matching in a most general setting. Specifically, our goal is to decide for a given $Q$, if w can manipulate her list to obtain a strictly better partner with respect to the true strategy $P$ (which may be different from $Q$), and also the outcome is a stable matching for $P$.

Motivation & Objective

• To address the gap in game-theoretic analysis of stable matching where prior work (e.g., Teo et al.) did not ensure that manipulated matchings remain stable under true preferences.
• To formalize the concept of $P$-stable Nash equilibrium, where a woman’s manipulation leads to a stable matching with respect to her true preferences $P$, preventing further strategic exploitation.
• To develop an efficient algorithm that determines whether a given strategy profile is a $P$-stable Nash equilibrium, ensuring strategic manipulations are both beneficial and stable.
• To prove that for any woman and target man, the stable matching resulting from a $P$-stable manipulation is unique, which enables algorithmic verification.
• To provide a computational framework that integrates true preferences $P$ into the analysis of strategic behavior, aligning with standard game-theoretic assumptions where true and stated preferences may differ.

Proposed method

• Leverages Algorithm 1 from Teo et al. (TST01) as a subroutine to compute, for a woman $w$ and her stated list $Q(w)$, the set of men $N_w(Q)$ who can be matched to $w$ via $P$-stable manipulations.
• For each woman $w$, computes the set $N'_w(Q)$ of men she prefers to her current partner under her true preference $P(w)$, and checks all such potential matches.
• For each such man $m \in N'_w(Q)$, constructs a modified preference list $Q(m,I;w)$ that places $m$ at the top of $w$'s list, and computes the men-optimal stable matching $\mu$ for the resulting strategy profile $(Q(-w), Q(m,I;w))$.
• Tests whether this resulting matching $\mu$ is $P$-stable (i.e., stable with respect to the true preferences $P$) using the Gale-Shapley algorithm.
• Uses a key lemma proving that for any manipulation strategy $Q$ yielding a match with man $\overline{m}$, the resulting stable matching is unique, enabling deterministic computation.
• Employs a recursive exploration procedure in Algorithm 1 to efficiently explore all possible manipulation paths that could yield a better partner under $P$.

Experimental results

Research questions

• RQ1 Can a woman manipulate her preference list to achieve a strictly better partner under her true preferences $P$, while ensuring the resulting matching remains $P$-stable?
• RQ2 Is there an efficient algorithm to test whether a given strategy profile $Q$ is a $P$-stable Nash equilibrium, where no woman can improve her outcome via a $P$-stable manipulation?
• RQ3 What is the computational complexity of determining $P$-stable Nash equilibria in the men-proposing Gale-Shapley setting?
• RQ4 How does the uniqueness of the stable matching resulting from a $P$-stable manipulation impact the design of verification algorithms?
• RQ5 What is the role of true preferences $P$ in ensuring that manipulations are not only beneficial but also robust against further strategic responses?

Key findings

• The paper establishes that a $P$-stable Nash equilibrium is a strategy profile where no woman can improve her partner under her true preferences $P$ via a manipulation that results in a $P$-stable matching.
• An $O(n^4)$ algorithm is proposed to test whether a given strategy profile $Q$ is a $P$-stable Nash equilibrium, significantly extending prior work by ensuring stability under true preferences.
• The algorithm correctly identifies $P$-stable manipulations by verifying that the resulting matching is stable with respect to the true preferences $P$, thus preventing further exploitable instability.
• The key lemma proves that for any manipulation strategy that matches a woman $w$ to a man $\overline{m}$, the resulting stable matching is unique, which ensures determinism and correctness of the algorithm.
• The algorithm runs in polynomial time, with $O(n^3)$ time per woman using Algorithm 1, and $O(n^4)$ total time over all $n$ women, making it efficient for practical use.
• The paper demonstrates that a $P$-stable manipulation is not only beneficial but also secure, as it avoids the risk of further manipulation by others, unlike unstable manipulations that may be exploited.