An interesting point, which I think probably will begin to recur in many ID contexts, is the distinction between an instructional design strategy and a design strategy. In their response Mike and Paal indicted their agreement that it was probably SD as an instructional strategy more than as an instructional design strategy. But in other (private) comments, Mike made the good point that if SD is appropriate for organizing and representing the content domain, it is in part an ID strategy, or at least the distinction between ID-strategy and I-strategy becomes fuzzy. This is a good point and the one which I think we will see recurring in other discussions. SD is less an instructional strategy (like feedback, or questioning) and more a content creation and organizing tool. Other content creation and organizing tools, like expert systems will probably present the same issue. What are the implications of such distinctions? I suspect the implications are for metacognition (of the students) and the way we design cognitive strategies as a part of learning environments. My suspicion (though I have not thought about it much) is that the purer instructional strategies are not as good prospects for the cognitive strategy portion of a complete learning environment, while those tools that are somewhere between ID-strategy and I-strategy are better candidates.
In their reply to some of Chet's comments [Hedden, 17 Oct 95] there was discussion of whether individuals ought to be the unit of analysis, in contrast to aggregates of individuals. I think it was being discussed with regards to business simulations. This goes to the heart of the distinction between continuous simulations and discrete simulations. Continuous simulations are solved with differential equations while discrete simulation employ statistical and probability functions. But which one is appropriate is not always clear cut. Take the example of population dynamics, human or otherwise. Obviously, populations are discrete (we don't have halves of people in reality). But population dynamics simulation are almost always dealt with as continuous simulations and with calculus rather than statistics for the underlying math models. Why? Because the numbers are generally very large (millions or billions of people) and the time period for discrete change (births) is tiny (seconds) in comparison with the time frames real population change occurs over (decades). So it makes more sense to treat most population simulation as continuous, that is, the unit being counted is millions of people, which is pretty continuous. Now, in a large business simulation with the overall time frame being years, this will probably be true as well. But if you are doing a small business simulation over a short time frame (months) it might be reasonable to model the simulation as a discrete simulation. It all depends on the number of units, the time increment size, and the overall time frame size.
Last, just a point of useful information. High Performance Systems (the makers of) have just instituted a World Wide Web site at http://www.hps-inc.com/.
Thanks to Mike and Paal for their article. I would encourage others to look into SD more. Though a somewhat difficult topic, I think it's worth the effort.