Siler, S., Rose, C., Frost, T., VanLehn, K., & Koehler, P. (2002). Evaluating knowledge construction dialogs (KCDs) versus minilessons within Andes2 and alone. ITS2002 Workshop on Empirical Methods for Tutorial Dialogue Systems (pp. 9-15).

In this paper, we will compare student learning of basic physics concepts when they engage in natural language dialog, specifically, in Knowledge Construction Dialogs (KCDs) with student learning when they simply read about the physics concepts (minilessons).  Because KCDs require increased student activity through incremental question-answering, we expected students to learn more from the KCDs than the minilessons.  We conducted a study in which we compared student learning from problem-solving in Andes2 with either KCDs or Expository minilessons.  No learning differences were found.  We then conducted a follow-up tudy in which students either worked through 30 KDCs or read 30 corresponding minilessons, and found higher learning gains for students in the minilesson condition.  However, because these results may have been influenced by technical difficulties experienced by students int he KCD condition or student fatigue in the KCD condition, we conducted another study in which we reduced the number of KCDs and minilessons in the intervention phase and fixed the problems leading to server crashes.  In this study there was no difference between conditions in student learning; however, more subjects need to be run before we can make any conclusions regarding the relative effect of student learning from KCDs and minilessons.  Finally, we discuss possible explanations for the null results and plans for a future study.

For a PDF full article version, click here (526KB).