Centipedes Leap into the Quantum Realm
Kaytki Chakankar, Xinhui Tang, Yiguo Zhang
TL;DR
This paper applies quantum game theory to the multi-round centipede game by adapting the Eisert–Wilkens–Lewenstein protocol into a CTZ framework, representing each round as an entangled qubit and encoding decisions as local unitaries. Through analytic expansions and Qiskit simulations, it identifies two quantum Nash equilibria at $[0,0,0]$ and $[\pi,\pi,\pi]$ that yield cooperative payoffs of $(2,2)$, suggesting a degeneracy where distinct quantum strategies produce identical outcomes. The results demonstrate that maximal entanglement and quantum interference can suppress last-round defection ($P_{3d1}=0$) and promote sustained cooperation, aligning better with observed human behavior than classical backward induction. The work introduces a general conjecture about quantum enhancements in sequential two-player games and lays groundwork for scaling quantum game-theoretic analyses to more complex strategic interactions and real hardware.
Abstract
The centipede game is a two-player non-zero-sum game. Each turn, a player can choose whether they want to take or pass a growing reward. The classical, rational solution of this game shows defection in the first round, when in reality, players cooperate much more often. Inspired by prior work employing quantum strategies in the prisoners dilemma, we showed that when similar quantum mechanics principles are applied to the centipede game, it leads to two new quantum Nash equilibria that are superior to the classical solution. Furthermore, by implementing our algorithm on Qiskit, we confirmed that leveraging quantum strategies, rather than strategies like backward induction, to solve the centipede game provided better payoffs for both players and more accurately modeled the games real-life outcomes. Ultimately, we propose a generalized conjecture for similarly structured quantum games.
