Especificação formal de um sistema de apoio à descoberta.

Abstract

In the context of agent based systems, a rational agent may be seen as a system appearing to the user as being capable of reasoning, handling interactions, beliefs, as well as incomplete, imprecise and erroneous knowledge. The rational agent SAID is conceived under those ideas, and is caractherized by its skill of producing révisable knowledge, and also by its skill of explaining its decisions. This work introduces a formal specification for a system for aiding discovery based upon rational agent SAID. Such a specification means a starting point to an implementation that will provide, from one hand, experiments of learning from examples algorithms in a variety of domains, and, from other hand, an aiding tool for generation and validation of scientific theories. From a widespread point of view, one should say that research in the direction of knowledge generalization is suitable because the large amount of information people handle nowadays is increasing. In such a sense, this work meets actual investigations both in Knowledge Acquisition and Machine Learning field.