If you’ve been a student or faculty member at an American college or university in the past twenty years or so, you’ve almost certainly run across student course evaluation surveys. They’re different in detail, but the key idea is always the same: toward the end of the term, students in every course are asked to fill out a questionnaire, usually a bubble sheet, assigning numerical values to various aspects of the course and the professor’s teaching. Most schools also provide some option for free-form written comments as well.
These course surveys, particularly the numerical scores, figure highly in evaluations of faculty for things like merit pay, tenure, and promotion. And yet, almost everybody in academia agrees that they’re highly flawed, easily gamed, and totally inadequate to the real task of evaluating faculty performance.
A passing mention of course evaluations in yesterday’s links dump (see also Matt’s front-line reporting) prompted some suggestions of alternative ways of doing course evaluations. So I thought I’d throw this out there more explicitly:
Suggest some practical ways of improving the standard student course evaluation process.
The constraints are that you have to provide some legitimate avenue for student feedback about the quality of the class, that the scheme has to be legal and ethical, and it has to be something that is not orders of magnitude more difficult or expensive than the current bubble-sheet systems (we do 20+ interviews of randomly selected students for tenure and promotion reviews, which is undoubtedly more accurate, but not remotely feasible for regular evaluations).
Suggestions can be minor tweaks (I personally favor the figure skating system, where you throw out the highest and lowest scores before calculating the average; other people do mid-term evaluations, and use them to adjust the course on the fly), or major overhauls (scrap the whole thing, and base promotion reviews on RateMyProfessors chili-pepper ratings). I’d love to hear some new ideas about how to make the process work better.