Group model building: a dialogue
In the late 1980s, academics on both sides of the Atlantic started to develop structured techniques for eliciting knowledge from expert and client groups as part of System Dynamics modeling. The goal was to democratize or “open up” the process, so that the ensuing model would have a stronger basis in observed reality, and so that its results would gain more support. The details of the practice, including formal roles and scripts and the name group model building (GMB), emerged largely from the University at Albany and Radboud University in Nijmegen; and before that from the model reference group concept developed in Norway.
It’s 30 years later, and GMB (also known as participatory modeling, mediated modeling, collaborative modeling, or heterogeneous problem solving) has become a widespread phenomenon in SD. I’ve had various thoughts about the value of GMB and decided to go directly to one of its originators, George Richardson, to ask him some questions about its current and ideal practice.
JH: Do you think GMB has lived up to its promise?
GR: Yes and no. On the one hand, much progress has been made in refining the process, and over the years I’ve known of many good modeling projects that employed GMB. On the other hand, I’ve seen work advertised as GMB or participatory modeling that does not involve formal models at all. It teaches people about causal-loop diagrams and then turns them loose to draw feedback pictures with little guidance and no testable model. Contrast this with our GMB work at Albany, where we have always built formal simulation models to help the group’s strategic thinking.
JH: Does this mean GMB has strayed from its initial purpose?
GR: Group modeling was created as a way to enrich and strengthen the modeling process, not water it down or simplify it. It wasn’t supposed to be a way to shortcut the hard work of modeling, nor to open the door to using qualitative models or maps instead of testable, quantitative models. We need to distinguish clearly between these things. Group mapping is not the same as group modeling.
JH: What do you see as the “hard work of modeling”?
GR: Gathering data, reading published studies, talking with top experts, developing robust equations, testing and refining hypotheses. The whole iterative process that allows us to get closer to reality and improve the reliability and usefulness of our models. Group modeling doesn’t replace those things; in fact, it adds depth and difficulty to them.
JH: Why should we want to add to a process that’s arduous to begin with?
GR: Here I come back to the primary purpose of GMB, which I see as broadening a model’s base of support to improve the chances of successful policy implementation. SD modelers have always puzzled about implementation, especially as our applications spread from business to government and public policy. GMB offers the opportunity to broaden the conversation to all constituencies, not at the end of the modeling process but at the beginning and all the way though. Done right, the clients can defend all of the substantive conclusions of the work because they are their own conclusions! It’s a way to build more understanding and get more buy-in for the model’s findings.
JH: But isn’t GMB also supposed to improve the knowledge elicitation process?
GR: It’s true that group modeling can make the knowledge elicitation process more efficient and effective, but if that’s all we wanted, we might have stuck with something more individualized, like written surveys and structured interviews. These are proven methods for gathering information and ideas, and avoid the problem of groupthink. But they do very little to build understanding and support, or to test assertions, which are the real strengths of putting everyone together in the same room as we do in group modeling.
JH: Thanks very much, George. I hope, as you do, that GMB returns to its initial purpose and approach.