Machine learning models of problem space navigation: the influence of gender

Ron Stevens, Amy Soller

We have developed models of how problem spaces are navigated as male and female secondary school, university, and medical students engage in repetitive complex problem solving. The strategies that students used when solving problem-solving simulations were first classified with self-organizing artificial neural networks resulting in problem solving strategy maps. Next, learning trajectories were developed from sequences of performances by Hidden Markov Modeling that stochastically described students’ progress in understanding different domains. Across middle school to medical school there were few gender differences in the proportion of cases solved; however, six of the seven problem sets showed significant gender differences in both the strategies used (ANN classifications) as well as the in the HMM models of progress. These results were extended through a detailed analysis of one problem set. For this high school / university problem set, gender differences were seen in how the problems were encoded, consolidated and retrieved. These studies suggest that strategic problem solving differences are common across gender, and would be masked by simply looking at the outcome of the problem solving event.