Matsuda, N., Cohen, W. W., Sewall, J., Lacerda, G., & Koedinger, K. R. (2008). Why tutored problem solving may be better than example study: Theoretical implications from a simulated-student study. In Proceedings of the International Conference on Intelligent Tutoring Systems.
Abstract: Is learning by solving problems better than learning from worked-out examples? Using a machine-learning program that learns cognitive skills from examples or by being taught, we have conducted a study to compare three learning strategies: learning by solving problems with feedback and hints from a tutor, learning by generalizing worked-out examples exhaustively, and learning by generalizing worked-out examples only for the skills that need to be generalized. The results showed that learning by tutored problem solving outperformed other learning strategies on the test scores that were measured on each problem solving step as the average ratio of the correct to incorrect rule applications. The advantage of tutored problem solving was mostly due to the error detection and correction that was available only when skills were applied incorrectly. The current study also suggested that learning certain kinds of conditions to apply rules only for appropriate situations is quite difficult. That is, learning how to perform mathematically valid operations is easier than learning when to apply rules.
