Both Chet Hedden and Clark Quinn put a case for modeling as an approach to researching educational technology. This appears in my paper as Hari Seldon's "psychohistory" approach. But I'm pleased to note that both of their suggestions involve qualitative methodology in addition to mathematical modeling. I really don't think we can treat a human learner in the same way as we treat a weather system or tectonic plates. One of the supposed differences between rocks and people is "free will," which implies that even the most sophisticated models will not apply to all cases.
[quoting Hedden, 19 Mar 97] A significant difference between what geologists do and what human subject researchers do is to be found in the sophistication of the instrumentation and the theory that supports and is supported by that instrumentation. I suggest that we need more sophisticated instruments--if not like the remote sensors used to sample the depths of the earth or the surfaces of distant planets, then perhaps more like those employed in biomedical research. And we need to build new theory that is based on measurement, observation, and modeling of subjective experience within complex social, political, and technological environments.
I would not agree that sophistication equals better measuring instruments. There are other ways of being sophisticated (e.g., I believe that poets are sophisticated). The important factor, as far as I am concerned, is the level of discipline one puts into collecting and dealing with the data. Any data. Not just numbers.
[quoting Quinn, 20 Mar 97] Kurt Van Lehn, in his AI-Ed (93, Edinburgh) keynote, talked about a methodology that included:
Given a domain and activity...
- take protocols of performance on that activity
- divide subjects into good and poor performers
- find processes that distinguish between the two
- build model and simulate the processes
- verify that the model predicts differences
- design interventions (most people start here)
- test the interventions
His claim was that this was the way to develop effective interventions. He has done this with Micki Chi, et al's, self-explanation effect and Kate Bialecyzk (sp) and others have found instructional improvements therefrom.
And Diana Laurillard gave us the "conversational model" in which actions could be defined as: Discursive, Adaptive, Interactive and Reflective. This is a model with its roots in both Socratic dialogue and Ference Marton's Phenomenography.
To my mind the difference between mathematical modelers and qualitative modelers is their distance from the subject--their willingness to get their hands dirty or to wade through the "slimy swamp." Mathematical modelers (like laboratory researchers) see virtue in maintaining a distance from their subjects. I don't know him, but on the basis of his prose style, Kurt Van Lehn seems to fall into this category. Qualitative modelers like Laurillard are first and foremost teachers themselves who perceive and appreciate complexity.
It is my firmly held belief that the search for simple solutions, and this includes developing models of human behavior, is like growing a delicate hybrid flower a sterile room. It looks great in its glass case, but when you take it out into the street it withers and dies.