Dealing with expected and unexpected obstacles
Fausto Giunchiglia · Journal of Experimental & Theoretical Artificial Intelligence · 1996
Generality and locality have been proposed as two crucial properties for systems formalizing common-sense reasoning. The first is the ability of representing knowledge in a way that makes it usable in a wide class of circumstances. The second is the ability of using only a subset of the potentially available knowledge, namely the subset which is held to be relevant in a given circumstance. These two properties seem to be one the opposite of the other, since disregarding part of the available information (locality) may lead to a loss of generality. In this paper, we argue that this is not the case, and propose a general methodology for combining generality and locality. This methodology is essentially based on the notion of context. As a case study, we propose a formalization of the Glasgow-London-Moscow example and its mechanization in an interactive theorem prover, GETFOL.