John Anderson, Carnegie Mellon University
Abstract: We have been looking at the learning of a difficult class of word problems and the insights required to master these problems. We call these problems part-rate problems. They involve learning a set of rates for different parts of a larger whole. These rates can be things like speed, payment, weight, etc. Students have difficulty appreciating the abstract communality among all members of the class and the abstract principles for solving all such problems. Based on earlier research, we have created a word problem tutor that uses a tabular representation to illustrate the abstract communality in the content of all of these problems and the common principles for solving these problems. In parallel with research on this tutor we have been using fMRI data to identify regions that are active when people get insights. We will present data from two laboratory studies of use of the tutor. These studies show clear markers of insight in predefined anterior cingulate, prefrontal, and parietal regions. We will also present data from an ongoing study with the tutor. We will describe the implications of these data for developing an improved version of the tutor.