Within intelligent tutoring systems, instructional events are often embedded in the problem-solving process. As students encounter unfamiliar problems there are several actions they may take to solve it: they may explore the space by trying different actions in order to 'discover' the correct path or they can request a hint to get 'direct instruction' about how to proceed. In this paper we analyze experimental data from a tutoring system that provides two different kinds of hints: (1) interface specific hints that guide students attention to relevant portions of a worked example, supporting student discovery of next steps, and (2) procedural hints that directly tell students how to proceed. We adapted a method of sequence clustering to identify distinct hinting strategies across the two conditions. Using this method, we discovered three help-seeking strategies that change due to experimental condition and practice. Further, we find evidence that the most successful strategy increases in use more quickly in the discovery condition and that students in this condition achieve mastery more quickly.

Tenison, C., & MacLellan, C. J. (2015, June). The Impact of Instructional Intervention and Practice on Help-Seeking Strategies within an ITS. In Proceedings of the 8th International Conference on Educational Data Mining, Madrid, Spain. (Paper available here)

Poster can be downloaded (here)

  • In-vivo study of 763 high school students
  • Hidden Markov modeling
  • K-mediods clustering
  • Mixed-effects modeling
  • Christopher J. MacLellan