This paper has four parts. First, there is a brief overview of AutoTutor’s pedagogical strategies that embody a construc- tivist approach to instruction. Second, student assessment data are presented from both controlled laboratory experiments and classroom use of AutoTutor. Third, the architecture of the Auto- Tutor intelligent tutoring system is presented. Finally, a descrip- tion is given of some new advancements in the system, including a version of AutoTutor on conceptual Newtonian physics that has interactive simulations.