No Model Is Right, But Some Models Are Useful

In other words, I love ModSim.

Today, we started in the auditorium, but then we went to the studios where we were assigned to groups for our first project. The groups are of four people each, and the ModSim instructors (and probably NINJAs) assembled the groups using the surveys we turned in about our experience in various areas. I'm in a group with three guys: Roland, Guillaume, and Geoffroy. (Yes, that's how it's spelled in French, at least traditionally. Promise.)

There were four ideas that the instructors had given us for our projects, and my group chose the suggestion about education. We're looking at the percent of children in Uganda who finish primary school and how social inertia impacts that. In other words, we're going to make and analyze a lot of graphs involving percent of children finishing primary school, number of educated adults, drop-out rate, and time. Today we were just sketching out our ideas -- where do we want to go, what's a basic stock and flow diagram, and what sorts of graphs do we expect will be on our poster?

The general advice for modeling is to start simple. It's hard to know what the right balance is. How simple is too simple to actually do useful work? But a model can also be too complicated to even really begin to get something with which to do useful work. I think our group encountered those problems today. What needed to be a stock? What could just be a source or sink? Upon what did all of these flows depend? What factors did we want to consider? What data did we have -- because after all, if there's no data, we can't really build the factor into our model?

At the end of class, we were talking with one of the other groups who had chosen the education topic. We had a lot of the same ideas, but their proposed model was rather different from ours. That was interesting and, I thought, pretty cool. It also led to Roland stating the quote in the title.
That's really the point of ModSim, I think. Actually, yesterday and today when we were in the auditorium, one of the profs spent a long time talking about the point of ModSim, in various ways. ModSim is about the process of modeling and using models to do work. Along the way, it has a lot of what the professor called disciplinary content, things like common models, difference equations, and differential equations, but that's not really the point of the class. We'll leave the class with a good grasp on those, but that will be because we've used them for our own modeling projects.

At the end of class yesterday, we quickly filled out a survey and left it for the teachers. Evidently one of the most common questions was "How simple is too simple?" Surprise, surprise, considering it actually came up as a difficulty in our work today. The professor told us that in many ways, answering that question is the idea behind ModSim. As we go through the class, that's what we'll figure out.

Because really, no model is right. There are too many variables in real life to ever model. But some models are useful. We just have to figure out which ones.