Perhaps I can add to Terry's argument, by describing my viewpoint as a cognitive scientist in design research:
What you do when developing cognitive models in AI is basically that you try to figure out what people do and then capture that in a theory expressed in an computer model. What you then do is that you run your model in a computer program on a test material and see if your model produce the same behaviour as people exhibit. If you model produce the same behaviour, including erroneous actions, you are likely to have a functionally correct model of what people do.
The upside about this way of conducting cognitive science is that you need to develop a very explicit theory, which you actually can model in detail. The downside is that you need to have a very narrow focus. This means that it is best suited for research problem where you know exactly what the relevant problem is, otherwise you end up with modelling something you actually *can* make a detailed computer model of, rather than doing research on what is really relevant.
I have personally never used cognitive modelling as a research approach, so someone who has worked with cognitive modelling (Don in the 70'ies and early 80'ies?), might correct me if I'm wrong. I guess I need to (re-) learn these things if I am to take over as director of studies for the cogsci undergraduate programme here in Linköping...
MATTIAS ARVOLA, Ph.D.
Sr. lecturer in Interaction Design.
Linköping University and Södertörn University.