本人运用动态认知逻辑的方法研究狼人杀博弈中多主体的知识和信息和主体 之间的信息互动。在狼人杀博弈中，好人玩家需要根据自己的身份牌以及听其他 玩家的发言来推理其他人身份牌的信息，依此来推理和判断在场的疑似狼人并将 其投票出局;而狼人玩家由于已经知道队友的身份，他们则需要听发言来推理和 区分好人玩家中的神玩家和普通村民玩家，在夜间行凶时专注于杀害神玩家或普 通村民玩家，在白天发言时撒谎隐藏身份并且转嫁嫌疑，煽动好人玩家帮助自己 把目标好人玩家以狼人嫌疑犯的身份推出局，用尽可能少的轮次淘汰更多的好人 玩家以获得胜利。因此，该博弈中知识，信念和撒谎扮演了重要的角色。本文首先介绍了认知逻辑基础，包括多主体知识逻辑、多主体信念逻辑以及 二者的融合，还包括动态认知逻辑。这些基本逻辑理论为分析博弈中的知识和信 念以及信息的动态变化?供了逻辑分析工具。这一部分为分析狼人杀博弈中多主 体信息互动?供了逻辑分析工具。本文分析了狼人杀博弈，并由于该博弈复杂度 高，行动种类多而对它进行简化为仅有三名玩家和三张身份牌的“三人杀”，并且 分析了其中包括的知识和信念等信息。本文运用动态认知逻辑分析了狼人杀博弈中不同主体的认知行为，这些认知 行为对博弈模型的进化起到关键作用。这些行为是分析博弈中多主体之间信息互 动的关键因素。本文所采用的动态认知逻辑理论是主体宣告逻辑，可以用以分析 真实宣告、虚假宣告等不同的认知行为。在狼人杀博弈中，主体有不同的类型。因 此，本文所运用的逻辑理论也建立于不同主体类型的基础之上。在技术方面，本 文还分析了反映狼人杀博弈中知识和信念变化的归约公理。

Dynamic epistemic logic is applied in this paper to study multi-agent knowledge and information in the Werewolf Game and the information interaction between the agents.In the Werewolf Game, villager players and god players need to reason about other people’s identity cards based on their own identity cards and other players’ speeches, so to infer the presence of suspected werewolf players and vote them out; and having known the identities of teammates, werewolf players need to listen to speeches to reason and distinguish between god players and villager players. At night, they focus on killing god players or villager players given the winning condition for werewolf players is to kill either all the villager players or the god players. They lie and stay hidden while speaking during the day, also convince god and villager players to help them vote out the “good guys”as werewolf suspects, with as few rounds as possible to expel more “good guys”so they can win. Therefore, knowledge, beliefs and lying play important roles in this game.This paper first introduces the basics of epistemic logic, including multi-agent knowl- edge logic, multi-agent belief logic, and the fusion of the two, which also includes dynamic epistemic logic. These basic logic theories provide logic analysis tools for analyzing knowledge and beliefs in games and the dynamic changes of information. This part pro- vides a logic method for analyzing the interaction of multi-agent information in Werewolf Game. This article analyzes the Werewolf Game, and given the high complexity of the game and the variety of actions, we simplify it to a “sanrensha”with only three players and three identity cards, and analyse the knowledge and beliefs included in the game.This paper uses dynamic epistemic logic to analyze the epistemic actions of different agents in the Werewolf Game. These epistemic actions play a key role in the update of the game model. These actions are the key factors in analyzing information interaction between multiple agents in the game. The theoretical dynamic epistemic logic used in this paper is agent announcement logic, which can be used to analyze different epistemic behaviors such as truth announcement, false announcement and etc. In the Werewolf Game, there are different types of agents. Therefore, different agents type is also discussed in this thesis. In terms of the technique, this paper also analyzes the reduction axioms that reflect the changes in knowledge and beliefs in the Werewolf Game.