15

Aleven, V., Mclaren, B. M., Sewall, J., & Koedinger, K. R. (2009). A new paradigm for intelligent tutoring systems: Example-tracing tutors. International Journal of Artificial Intelligence in Education, 19(2), 105-154.

Background

Example-tracing is an alternative to the popular Model-tracing framework which is used to design intelligent tutoring systems. Intelligent tutoring systems are automated learning environments that help users understand a concept through problem-solving. Example-tracing involves creating user behavior graphs in order to track their progress when solving a problem. The developer of the system creates a graph that maps out possible user steps in solving the problem. The system judges a user’s work as correct, incorrect, or suboptimal. If the user does a step incorrectly the system provides a hint on how to perform the step. If the user completes a step suboptimally, the system will provide a hint on how to best perform the step. The user can choose to change their behavior or ignore the hint. If the user ignores the hint, they chose to take the path with the suboptimal step. Once locked into a path, the user must complete the problem along that path. Each incorrect and suboptimal answer possibility has a hint within the user behavior graph. An example of a user behavior graph is provided in the text (Alvene et al, 2009).

Example-tracing ITSs, intelligent tutoring systems, were developed as a way to author tutoring systems without the need for programming. The article’s authors believe “that efficient authoring tools for non-programmers are an important condition for making ITSs widespread” (Aleven et al, 2009, 106). The example-tracing model is more adaptable and does not require the extensive technological background to design. Instead of programming authors of these systems use drag and drop functions to create the interface and problem-solving model(Alvene et al, 2009). There is a possibility that teachers would be able to design intelligent tutors for their classes.

Key Points

• Example-tracing paradigm
  • User behavior graph to track the way a user can solve the problem
    • Generalizable
    • Mass Production
  •  Feedback in the form of hints in every step
  •  Easier to author
    • More people able to author ITSs
    • Content experts can be more involved

Design Principles

The example-tracing paradigm employs some of the same design principles as the model-tracing paradigm. Both rely on immediate feedback and a problem-solving context. In example-tracing, the developers create a user behavior graph for a problem.

When developing the user behavior graph, the graphs can be generalized which allows the system to do complex evaluations of user work. The author can make steps optional or reusable which gives the system more flexibility. Through generalization, an author can give a step a range of acceptable answers or create paths that can skip steps or loopback to different steps in the problem-solving process. Generalization can provide a more personalized learning experience for the user with the multitude of paths available to them (Alvene et al, 2009, 128-129).

Example-tracing authoring tools also allow for user behavior graphs to be mass produced. The author can reuse steps from user behavior graphs in other similar problems(Alvene et al, 2009, 132-133). These features make authoring ITSs easier and more affordable which could lead to a proliferation of the technology.

Discussion Questions

1. Does usability suffer with the use of example-tracing ITS authoring tools? Are authors with no design or programming experience equipped to create systems that are suitable to their users?

Additional Resources

1. Aleven, V., Baker, R., Wang, Y., Sewall, J., & Popescu, O. (2016, April). Bringing non-programmer authoring of intelligent tutors to MOOCs. In Proceedings of the Third (2016) ACM Conference on Learning@ Scale (pp. 313-316). ACM.
2. Blessing, S. B., Aleven, V., Gilbert, S. B., Heffernan, N. T., Matsuda, N., & Mitrovic, A. (2015). Authoring example-based tutors for procedural tasks. Design recommendations for intelligent tutoring systems, 3, 71-93.