Steve Alessi's [Oct 95] comments on our paper initiated a very excellent series of comments. We wanted to make a few responses in this note to both Steve's and Chet's concerns.
First, Steve is quite right that system dynamics is difficult to teach. Some of the references provided in the response to Barney also make mention of this. I do not believe that all subject matter and learning situations are appropriate candidates for SD models. However, there are many complex and dynamic situations in which learning can be enhanced by using system dynamics methodologies, in terms of deeper understanding of the relevant phenomena as well as in terms of more engaging involvement of learners with the material. These cases are reasonably documented in Roberts, Andersen, Deal, Garet, & Shaffer's (1983) Introduction to Computer Simulation: A System Dynamics Modeling Approach, in Mandinach & Cline's (1995) Classroom dynamics, and in Morecroft & Sterman's (1994) Modeling for Learning Organizations.
The subject domains typically include economic systems, ecological systems, and various engineering processes. Students range from graduate students to elementary school age children. Results have indeed been varied and not especially well documented, as Steve already indicated. We never intended to suggest that System Dynamics was an appropriate instructional strategy in all situations. We do believe, however, that many of the steps required in constructing system dynamics models are appropriate for a great many domains, and we did try to indicate some of those steps (e.g., making assumptions explicit, finding and citing evidence in defense of assumptions, formulating representations of causal relationships among stocks and flows, and so on.). It is not simply that making ones assumptions explicit is a good habit for learners to acquire. It is more specifically that one needs to identify and make explicit assumptions in a very particular in system dynamics in order to create a meaningful model. It is this latter practice that is of particular interest to us.
With regard to Steve's comment about being confused about whether we were discussing an SD-based instructional design strategy (principles that could be used to guide the design of a wide variety of learning environments) or an SD-based instructional strategy for learning in a particular domain (ecology). He concludes that the latter was our real focus. Steve's insight on this issue is remarkable. We are of two minds on this issue, and we have debated it between ourselves. We remained unresolved about where the emphasis (in terms of interesting and beneficial educational research) should be placed. We gladly welcome and encourage additional discussion of this particular issue.
Chet's remarks [Hedden, 17 Oct 95] were also quite insightful, and we shall make a few rejoinders here. The observable changes to which we refer as indicators of learning are not exactly "equal to" acquisition of knowledge. We regard acquisition of knowledge as a somewhat fuzzy and ill-defined process (we say this knowing that the computer scientists out there may now take out a contract on our lives), which is relevant to many of Chet's later comments. We would prefer to say that the observable changes that would serve as markers of learning are the results of knowledge acquisition and representation processes that occur dynamically in individuals.
Chet's remarks about knowledge as stocks raises some interesting concerns. If we say yes, knowledge can be modeled as a stock in SD, then it is likely that the discussion may become yet another version of the constructivist-instructivist debate--and this would detract from potential insight that may result from trying to model knowledge in a learning community as a stock. Please recall that we have already indicated that stocks may represent continuous things (we called them substances) as well as discrete, countable things. We even mentioned that we could represent perceptions as stocks. That should make it obvious that we think Chet is on to something here--others (John Sterman and Tarek Abdel-Hamid, for example) have been thinking about this issue. For now, let us merely remark that involved in this discussion is an assumption about the appropriate unit of analysis for learning. Those who want to introduce learning about system dynamics (so as to improve complex decision-making and problem-solving skills) in business and industry might well argue that individuals are not an appropriate unit of analysis, as would be most typical for cognitive and developmental psychologists.
Chet's remark about the choice of species between the question of greater interest is also interesting. We regarded that learning goal (learning how populations of two species fluctuated) as poorly defined exactly because we did account for very many of the critical attributes of learning environments introduced earlier in our paper. One of those characteristics would be who are the learners; another is why someone would be studying this topic. Since we wanted to provide a generic discussion to sketch how things might be done, we chose to allow readers to fill in their own species (that increases the odds of having a personally meaningful example). This same strategy can be applied in a course, if the goal is to learn about the kinds of relevant and dynamic behavior that is involved in ecosystems, rather than about particular species.
Well, we are only about half way through Chet's remarks and this response is already quite long.
We just wanted to indicate that we were indeed here and that we are very much appreciating the thoughtful responses that are being posted.