In this paper, we present a computational framework for the detection of conceptual discrepancies in human-computer interaction. The framework is developed as a basis for the generation of feedback utterances at the ontological level. In our method, presuppositions are extracted from the user's message, expressed in a common vocabulary, and compared with the system's ontology, which is expressed in type theory. Discrepancies are detected by the system if it notices type conflicts, particular inconsistencies or ontological gaps. Depending on the kind of discrepancy, the system generates a particular feedback message in order to establish alignment of its private ontology with the mental model of the user.