Tuesday, October 18, 2005

Cohorts

We’ve started a small alternate-delivery degree program this year, and so far, it’s going pretty well. The students enrolled seem to like it, and the numbers, while small, are viable. We picked the first few courses to offer based on what seemed to make sense for an entering group, and we mostly got it right (though transfer credits meant that some students who wanted to start in September already had credit for the first few courses we ran).

Now, the students who started in September are asking, reasonably, for the actual course schedule for the next few years – partly so they can plan, and partly, I think, so they can force us to commit to running everything they need, when they need it.

Conceptually, I agree with them. But anytime you try to figure out which courses everyone will need when, you start imaging students in cohorts. And they aren’t.

For scheduling purposes, it’s lovely to imagine that first semester students need these five courses, second semester students need these next five, and so on. When you’re talking about the weekday classes, the numbers are large enough generally that this illusion can go uncorrected. But when you have a fairly small group to start with, thin-slicing will kill you; if we only have 25 students when they’re all together, splitting them into three or four groups would make the program a money pit.

The catch is that students stubbornly refuse to behave the way we think they will. We can lay out a lovely, logical progression of courses, and students will bring in random transfer credits, fail a course here and there, ‘stop out’ for a semester, put off math until the bitter end, etc. They just won’t move in unison.

There’s no real reason they should, of course, but it makes scheduling a course rollout a lot harder, and effectively raises the bar for the minimum size population we need to make a program viable.

At my previous school, I went through this fire drill several times, and never found an elegant answer for it. Surveys don’t work, since students often don’t remember which courses they’ve already taken (incredible, but true), or, by a weird dyslexia, will badly miscategorize them (psychology and physics often get transposed, which I still find bizarre). Predictions are ballpark, at best, and usually pretty inaccurate when the sample size is small and the program new.

Any ideas out there? I’m getting pressure to lay out the next round of courses, so this isn’t just a theoretical exercise.