Abstract
In this study, we discuss a new direction of unsupervised learning and concept formation in which both domain knowledge and experimental evidence (data) are considered together. This is a reflection of a certain paradigm which could be referred to as knowledge-oriented clustering or knowledge mining (as opposed to data mining). We offer the main concepts and in selected cases present algorithmic details. The distributed way of forming information granules which is realized at the level of individual locally available data gives rise to higher order information granules (type-2 fuzzy sets, in particular).