2025-04-14 マックス・プランク研究所
マックス・プランク進化生物学研究所の研究により、プレイヤーは過去のすべての報酬を記憶せずとも、短期的な記憶（2～3回分）だけで協力戦略を効果的に進化させられることが判明。進化的ゲーム理論を用いたモデルで、直近の相互作用に基づく行動調整が、長期記憶と同等の協力率を生むと示された。これは、生物進化やAI、ロボティクスにおける協力行動の理解と設計に新たな視点を提供する。
＜関連情報＞
- https://www.mpg.de/24529633/players-can-succeed-with-less-memory-in-game-settings
- https://www.sciencedirect.com/science/article/pii/S0165176525001375?via%3Dihub
あなたより記憶が少ないのは仕方ない？繰り返し加法ゲームにおける最善の反応 Can I afford to remember less than you? Best responses in repeated additive games
Franziska Lesigang, Christian Hilbe, Nikoleta E. Glynatsi
Economics Letters Available onlin:e 30 March 2025
DOI:https://doi.org/10.1016/j.econlet.2025.112300
Highlights
- We study repeated additive games among two players with finite memory (finite recall).
- Suppose one player reacts to the opponent’s actions during the last n rounds.
- We prove the opponent has a best response that only depends on the last n-1 rounds.
Abstract
In this paper, we study best responses in repeated additive games among two players. A stage game is additive if each player’s payoff is the sum of two components, and each component only depends on the action of a single player. We suppose one player’s strategy depends on the co-player’s last ｎ actions. Then we prove that the other player has a best response that only depends on their own n−1 actions. That is, for an important sub-class of games and strategies, players can achieve maximum payoffs even with less memory than their opponent.
