在动态认知逻辑中融合观察与内省

Merging Observation and Access in Dynamic Epistemic Logic

一直以来,我们使用逻辑系统来描述数学的证明、结构的计算以及语言的意义。近年来,逻辑系统却越来越多地被用来研究理性行动者的很多方面。例如,如何接受单一的信息,多主体间的交流行为,以及更为一般的受目标驱动的主体间的互动。特别是,对观察和交流中的信息流的研究,大家使用所谓的知识更新、信念修正和偏好改变的动态认知逻辑。当新信息进来时,这些逻辑使用信息的"语义意义"作为被更新状态的选择范围。然而,同样重要的是,理性主体的行动也会基于其他信息,譬如,由推理和反省得到的信息。对这些问题的研究实际上是对信息在更为语法的意义上进行理解的,即,把信息看作是可以由主体阐明的东西。也许有些奇怪,尽管在这一领域已经存在不少的研究方案,但是大家对什么是信息,信息的关键机制是什么等问题更少有一致的意见。可以看出,对"信息"的意义在逻辑中确实有很多不同的理解。本文基于可能世界的语义,给出一个一致的信息模型,同时也赋予可能世界语法的"可及通路"。这样,我们就能把外部的"更新信息"和内部的"阐明信息"放在同一个动态逻辑系统中。特别是,我们提出了两个基本的信息行动:纯粹的基于观察的更新("单纯的看")和把不明显的知识变为明显知识的"知觉实现"。我们阐明为什么这些行动是自然的,同时我们也提出了一些新的研究问题。其中,很多问题探讨如何使其他的逻辑传统,包括信念修正理论、情景语义学和弗协调逻辑等适用于信息-驱使的理性行动者的图景。

Logical systems have long been used to describe mathematical proof, structured computation, and linguistic meanings. In recent years, they are also coming to be used increasingly to study rational agency in its many aspects, from picking up single pieces of information to multi-agent actions of communication and goal-driven interaction generally. In particular, information flow through events of observation and communication has been studied using socalled dynamic epistemic logics of knowledge update, belief revision, and preference change. These logics use the ＇semantic sense＇ of information as ranges of possible options which get updated as new information comes in. But equally importantly, rational agents also base their actions on information from other sources, such as inference and introspection. The latter is the area of more syntax-oriented senses of logical information as something which can be ＇elucidated＇ by agents. Perhaps surprisingly, there is much less of a consensus on what this information is, and what its key mechanisms consist in, though there are many competing proposals in the logical literature. Thus, logic is really a field with many different senses of ＇information＇. In this paper, we try to get clear on these issues by presenting one unified model of in- formation, based on semantic ranges of possible worlds, but endowed with syntactic ＇access＇ to these worlds. This allows us to integrate external steps of ＇updating information＇ and internal steps of ＇elucidating information＇ into one system of dynamic logic. In particular, we propose two kinds of basic informational action： pure observation-based update （＇bare seeing＇） versus acts of ＇conscious realization＇ which turn implicit knowledge into explicit knowledge. We show how this is a natural fit, which also provides many new research questions - many of them having to do with fitting further traditions into this picture of information-driven rational agency： including belief revision theory, situation semantics, and paraconsistent logics.