Application of intelligent tutoring systems in ill-defined domains

Herbert Simon distinguished between well-structured task domains (e.g., math and physics) and ill-structured ones (e.g., legal reasoning or intercultural competence). ITSs have been very successful in well-structured domains, but ill-structured domains pose a challenge for ITS development: in these domains, the correctness of answers often cannot be determined objectively. I am involved in three projects that try to broaden the range of application areas for ITSs by focusing on ill-defined domains:

• In the Hypothesis Formation project, my colleagues Kevin Ashley, Niels Pinkwart, Colin Lynch, and I showed that argument graphs, combined with on-demand feedback in the form of prompts for reflection, and suggestions to improve the graph, helps beginning law students learn argumentation skills, although the effect seems confined to the lower-ability students.
• In a project dealing with causal argumentation, Matthew Easterday's PhD research, a lab experiment showed that causal diagrams help students interpret texts about complex policy issues.
• In a project dealing with intercultural competence in 2nd-language instruction, Amy Ogan (one of my PhD students) showed that the video clips from feature films spur richer discussion among students, and lead to better cultural learning, if the clip is paused at a key moment, and the student is asked to predict what will happen next.

Combined, these three projects provide strongly-suggestive evidence that ITSs can be effective in ill-structured domains.