On the Formation of Empirical Theories in Poorly Formalized Domains

The paper presents the problems of generating new knowledge using artificial intelligence in open empirical domains, where formal apparatus is largely absent and the procedures for forming theories are heuristic in nature. The issues of presenting data and knowledge, forming fact bases, and choosing procedural solutions are considered. Approaches to specifying the algebraic operation of similarity—a key element of plausible reasoning using the JSM method for automated research support—for various structures are characterized. It is emphasized that effective solutions to these problems can only be proposed under the conditions of interdisciplinary interaction.