What this results in is a system which can respond to reasonable yet unanticipated questions put to it by the student. This result contrasts sharply with the much more rigid interaction (relatively speaking) of traditional CAI systems, in which all potentially meaningful questions and responses from the student must be anticipated by the courseware author and explicitly included in the program. Traditional CAI is frame-oriented, whereas ICAI is knowledge-oriented at a much finer level - that of the knowledge elements themselves. A second powerful feature of many ICAI systems is the inclusion in the system of a component which can analyse student input expressed in natural English and which can translate this input into an internal form recognizable by the system's knowledge base. This natural language processing ability (itself an active area of research in AI) opens up the human-computer interaction well beyond the short-answer expectations and menu-driven interactions of CAI.