The impetus for this notion was produced by an observation about the normal effect of spending a class period reviewing for a quiz: the top third of the class is bored, the bottom third has its condition converted from vague anxiety to confirmed panic, and only the middle third actually profits. Not good. To this the Clarke Solution is to divide the class into small groups (four or five students) and charge each group with producing a set of problems which adequately covers the relevant material--i.e., a proto-quiz. The contract the instructor makes is that at least one problem from each group will be put directly onto the quiz (with possible cosmetic changes if needed for clarity.) David's experience has been that the quizzes are generally harder ones than he would have written, and the grades are appreciably higher than on his. My brother-in-law was not convinced that the quiz was any harder, but it was also not easier--and the median was 95%.
David takes the process one notch further and reports that the day after the quiz, which is normally the only day that is even more depressing than the review day, is much enlivened by everyone's desire to find out just who produced that crazy problem #9. Myself, I tend to dodge that issue altogether by handing out solution sheets and forging ahead, but it's an intriguing thought!
A brief note from elsewhere on the Pew front: we just had rather an intensive day and a half meeting at the Sheraton with representatives from all seventeen clusters which have received PFF grants (four others in our category and twelve with smaller grants.) The first morning's worth was a report to the Council of Graduate Deans, which was finishing off a several day meeting, and which is a co-sponsor of the PFF. The rest was a discussion among ourselves. I'll try to bundle together a bunch of the interesting reports and ideas for another newsletter, but meanwhile it seems worth mentioning that one of the best received pieces of the whole thing was a discussion by a panel of graduate students, including our own Brian Hopkins. Definitely a Good Thing!
The Clarke quiz idea is one I've used a couple of times with some success as well. A variation of it for a bigger stake, perhaps a midterm, is to take a list of potential questions, perhaps some developed by students and others contributed by the boss, and have student groups adopt a question. Each group develops a model answer and presents it to the class. Come exam time a subset of these are actually on the exam. Everyone has seen a solution, but then each one needs to present their own on the test without the benefit of notes.
I don't do a lot of this, but the results are actually kind of surprising. Some students produce the best timed test work they have ever done. They say it is the first time they really knew what kind of answer was good. The ones who still do badly have no excuse. And sometimes it actually makes them turn around and change some behaviors.
In any case it gets the groups talking to each other about what constitutes a correct solution, and how to express it succinctly--not an altogether wasted use of time I think.