Analogical Learning with Multiple Models

Abstract

When students are learning from multiple analogies, they must relate and integrate several analogical models, in a fashion consistent with what they already know from examples in the domain they are studying. This analogical combination process often involves relating causal or structural models at several different levels of abstraction. Debugging such composite conceptual models thus requires attention to mappings between levels, as well as analogical mappings at a single level. We are studying the processes involved in this kind of learning by a combination of protocol analysis and the construction of computer models.