"Coisas que as pessoas sabem": computação e territórios do senso comum

Abstract

The discipline of artificial intelligence has been able to build computational systems with remarkable abilities concerning some difficult tasks for which a human intelligence would otherwise be required. One of those systems, a long term and large-scale enterprise known as CYC, aims to represent and to make usable by computational means the common sense knowledge. Common sense is, in this situation, non-expert knowledge of the world around us, of the type people put to use in their day-to-day affairs. The creators of CYC make use of a number of implicit assumptions about that common sense, such as, in the first place, that it may in some way be formally represented. In this article, we examine this project in order to make visible some of these hidden assumptions that we regard as important. We attempt also to show how common sense knowledge expressed in CYC carries with it worldviews and notions of what constitutes valid knowledge that derive from the perspective of its developers.