Arguing that the evolution of intelligent tutoring systems better reflects the recent theoretical developments of cognitive science than traditional computer-based instruction (CBI), this paper describes a general model for an intelligent tutoring system and suggests ways to improve CBI using design principles derived from research in cognitive science rather than behavioral psychology. Differences between the two types of instructional systems are explained, and four major components incorporated in all intelligent tutoring systems are identified: (1) the user interface, which provides the means for two-way communication where the learner is engaged in some activity while the system is interpreting that activity in order to make a meaningful response; (2) the learner model, which is a representation of the errors or misconceptions that commonly occur when learners are exposed to the content; (3) the expert model, which is typically a database that represents the knowledge of an expert in the domain; and (4) pedagogic knowledge, which receives the results of a comparison of the learner's present knowledge state with knowledge in the expert model and makes decisions about what, when, and how information is to be communicated to the learner. Each of these components is then discussed in the context of design principles and general guidelines for designers of CBI derived from research in cognitive science and intelligent tutoring systems. It is concluded that, in order for developers of CBI to integrate these findings into their practice, a new perspective, which will require different assumptions about learning than the instructional design models currently in use, is necessary. (36 references) (BBM)