Discovering Strategy Switching with Mixed Membership
Thursday, February 7, 2013 - 3:30pm

There is ample evidence that in a wide variety of settings, as people try to solve problems they switch strategies. For example, a student taking a test might solve one arithmetic problem using one method and then use a different method on the next problem. More importantly experts use different strategies than novices. Both experts and novices switch strategies, but experts use a different mixture of strategies than novices. Existing psychometric models do not model strategy usage,and cannot capture this critical dimension of expert-novice differences. Mixed membership models, e.g., latent Dirichlet allocation, are built to model the case where individuals have partial membership in multiple profiles. I present several simulations and an analysis of an addition assessment that demonstrate that mixed membership is feasible for not only modeling strategy usage, but also discovering the strategies that are present in the data. Finally, I characterize the cases where mixed membership models can capture strategy switching (using one strategy and then subsequently using another) vs. strategy blending (simultaneously combining two or more strategies.)