Normative, Subjunctive and Autoepistemic Defaults
Normative, Subjunctive and Autoepistemic Defaults
复制标题
规范性、虚拟性和自我认知默认值
DOI：
--
复制
发表时间：
1992
期刊：
影响因子：
--
通讯作者：
Craig Boutilier
中科院分区：
文献类型：
--
作者：
Craig Boutilier
我们探讨了非单调推理理论与使用模态表征和条件表征的信念修正之间的关系。我们证明了控制信念修正和条件默认推理的逻辑是相同的，通过表明规范条件和虚拟条件，在早期的工作中在模态逻辑CO*中定义，具有完全相同的形式真值条件。这只能通过严格遵守我们配方中的拉姆齐测试来实现。因此，我们必须尽力避免众所周知的琐碎结果。我们解释了这种形式的相似性，区分两种推理类型的实际考虑因素，以及默认推理如何被视为信念修正的特殊情况。然后，我们证明了自认知逻辑可以在这个框架内推广，并讨论了自认知缺省值、规范缺省值和虚拟缺省值之间的关系。
We explore the relationship between theories of nonmonotonic inference and belief revision using modal and conditional representations. We demonstrate that the logics governing belief revision and conditional default reasoning are identical by showing that the normative and subjunctive conditionals, defined in a modal logic CO* in earlier work, have precisely the same formal truth conditions. This can only be achieved by a strict adherence to the Ramsey test in our formulations. Hence, we must take pains to avoid the celebrated triviality results. We explain this formal similarity, the practical considerations that distinguish the two types of reasoning, and how default reasoning can be viewed as a special case of belief revision. We then show that autoepistemic logic can be generalized within this framework and discuss the relationship between autoepistemic, normative and subjunctive defaults.