AI research in qualitative physics and causal models suggests new approaches to teaching people about science and engineering. We have been investigating the form that such models need to take to be effective within intelligent learning environments. The subject matter we have focused on is understanding how electrical circuits work, but the approach can be generalized to other subjects. Two key hypotheses have emerged from our research. The first is that in order to understand a physical system, students need to acquire causal mental models for how the system works. Further, it is not enough to have just a single mental model; students need alternative mental models that represent the systems behavior from different but related perspectives, such as at the macroscopic and microscopic levels. The second hypothesis is that in order to make causal understanding feasible in the initial stages of learning, students have to be introduced to simplified models. These models are then gradually refined into more sophisticated mental models. The questions addressed in this article are: What are the properties of an easily learnable, coherent set of initial models? What are the types of evolutions needed for students to acquire a more powerful set of models with broad utility?