In this paper, a general overview of the components which have characterized the development of intelligent tutoring systems (ITS) over the past fifteen years is provided. Accompanying the overview of each component is a discussion of limitations which, we feel, restrict the extent to which ITS technology can be useful as an instructional delivery vehicle and as a tool which can be used to learn about the processes which underlie teaching and learning. These limitations, however, can be compensated for by altering how and for what purposes ITSs are developed and implemented. Our goal in writing this paper, in addition to discussing the problems associated with present ITS approaches, is to present a set of suggestions which we feel can guide the development and implementations of ITSs such that their potential for useful instructional tools can be enhanced and extended. We argue for the development of ITSs which: (i) progressively and as much as possible reduce the a priori restrictions that are placed on learners as they learn new content. Technology should empower the learner as a learner, enabling him/her to uncover the “mysteries” of new knowledge. Technology should not rob learners of the joy of discovery the “aha!” experience. It should, however, facilitate the integration of new findings into existing cognitive frameworks, and provide opportunities for learners to examine and expose misunderstandings and misconceptions about how the “universe” operates; (ii) enable educators to reliably determine and report what is being learned and mastered against some set of standards which exist independent of the learning environment; and (iii) provide opportunities for educators to learn about how learning is occurring and intercede (real-time if necessary) in ways that can alter and improve (either temporarily or permanently) the environment within which student interactions are occurring.