Extracting knowledge structures from observed data

How can we construct a knowledge structure for a given set of problems? A natural construction method is to collect a large set of response patterns to these problems and then extract the knowledge structure from the data. We describe an algorithm for the construction of a best-fitting knowledge structure from a given data set. The ability of the algorithm to reconstruct the correct knowledge structure from the data is tested in a simulation study. The results show that the algorithm is able to reconstruct the underlying knowledge structure with high accuracy if enough response patterns are available.