The Rational Mirror: Learning how to Explicitly Organize Messy Concepts

Concept learning is a fundamental meta-cognitive capacity of humans. We show in the paper how concepts may be learned by humans engaged in conversations with a challenging artifact, a system we called “Rational Mirror” (RM), that shows the human how his/her partial theory is incomplete, contradictory, paradoxical so that it is to the human judgment to modify it until an adequate theory is constructed. We show with a magnifying lens, step by step, how the RM operates in a toy domain (the popular concept of an arch) and describe a real world application of the same in the domain of Law. Finally, we frame the ideas in a discussion that separates the Symbol Level from the Knowledge Level in conversations with RM, documenting therefore the current state of our generic theory for knowledge construction, called Phi-Calculus.