Reasoning about knowledge is useful in artificial intelligence and distributed systems theory, particularly when we look at systems composed of many interacting "agents." This dissertation investigates various questions concerning knowledge that are specific to multi-agent systems, and develops theories that cope with them.Many existing multi-agent logics fail to make a clear distinction between the agents themselves and the names for agents that occur within the logic. This is restrictive and prohibits a general treatment of, for instance, anonymous agents, names with varying or unknown denotation, agents with many names, named groups of agents, and relative (indexical) reference. We examine the principles involved in such cases, and give simple propositional logics that are expressive enough to cope with them all. We also give a new and powerful first-order modal logic for knowledge. Although this logic is motivated by practical examples, it is also relevant to some well-known philosophical concerns, such as indexical descriptions, de re knowledge, and quantifying-in. The logic adopts a new and nontraditional approach to such subjects, because this seems to be required by our use of epistemic logic as a descriptive tool for computer science applications.The dissertation also investigates systems of message-passing agents. We argue that it would be useful to augment the standard theory of ascribed knowledge with a logic and semantics for ascribed message content. The theories we develop in this dissertation have the important feature that content, like knowledge, can be ascribed to concrete messages in existing message-passing systems. The goal is to be able to reason about real systems in high-level terms, using knowledge, content, and the interaction between them.
展开▼
