"Belief change in rational agents (with application to action formalisms, ontologies, and inferentially weak systems)"

在人工智能的知识表示(KR)中，代理的知识通常是不完整的，可能是不准确的，并且是不断演变的。无论代理是物理领域中的机器人，还是网络上的软件代理，还是处理医疗信息语料库的推理代理，都将是这种情况。KR中的一个关键问题是管理这种不断变化的信息，无论变化是来自于被告知一个新的事实，还是作为一些执行的行动的结果，还是来自对环境的感知。这项拟议研究的重点将放在如何管理信念的改变上。虽然在关于行动和信念修正的推理领域已经开展了许多工作，但仍有一些关键问题没有得到充分解决或没有得到充分理解。例如，我们所拥有的信念修正模型是否完全表征了信念变化的所有方面，这一点并不清楚；此外，最近才有研究涉及特殊用途KR语言的变化，例如涉及Horn理论或本体论的语言。在前一个赠款期间的工作集中在信念改变的基础上，重点是迭代改变算子和推理弱系统中的改变，以及在偏好和非单调推理的相关领域的工作。在接下来的赠款期间，我将继续致力于信仰改变的基础工作，但也将讨论与信仰改变在现实环境中的应用有关的三个具体相关项目。第一个项目(与来自SFU、希腊和澳大利亚的同事)将检查描述逻辑中的信念变化，特别是一方面使用用于表示医学本体的逻辑，另一方面使用所谓的语义网。第二个项目(与来自多伦多的一位同事)将提供一个完整的高级规范，该代理能够结合有关已知操作效果、被告知某事和传感器信息的信息，目标是能够修改其知识，无论不正确的知识来自不正确的信息、易出错的传感器或失败的操作。第三个项目(与稳定部队的一名同事合作)将解决大规模(海上)和当地(家庭辅助生活)环境中的情况认识问题。

In Knowledge Representation (KR) in Artificial Intelligence, an agent's knowledge will generally be incomplete, possibly inaccurate, and evolving. This will be the case whether the agent is a robot in a physical domain, or a software agent on the web, or a reasoning agent dealing with a corpus of medical information. A crucial problem in KR is to manage such changing information, whether change comes from being told a new fact, or as a result of some executed action, or from sensing the environment. The focus of the proposed research will be on managing belief change. While much work has been carried out in the areas of reasoning about action and belief revision, there are crucial issues that have not been fully addressed or that are not fully understood. For example, it is not clear that the models that we have for belief revision fully characterize all aspects of belief change; as well, research has only recently addressed change in special-purpose KR languages, such as those involving Horn theories or ontologies. Work in the previous grant period focused on the foundations of belief change, with emphasis on iterated change operators and change in inferentially weak systems, along with work in the related areas of preferences and nonmonotonic reasoning. For the coming grant period I will continue working on the foundations of belief change, but also will address three specific interrelated projects concerned with the application of belief change in realistic settings. The first project (with colleagues from SFU, Greece, and Australia) will examine belief change in description logics, specifically with logics used for representing medical ontologies on the one hand, and the so-called semantic web of the other. The second project (with a colleague from Toronto) will be to provide a full high-level specification of an agent able to incorporate information about known action effects, being informed of something, and sensor information, with the goal of being able to revise its knowledge, whether the incorrect knowledge came from incorrect information, fallible sensors, or failed actions. The third project (with a colleague from SFU) will address situational awareness in large-scale (maritime) and local (household assisted living) settings.