From 1999-2003, as I've mentioned elsewhere, I taught at the Beacon School in New York City. For three of the four years I was there, I taught the senior physics class, which was a great experience. I was the only one teaching physics, and it was not a Regents or AP course, so I had a ton of flexibility in designing the curriculum, assessments, labs, and so on. Beacon had a midterm and finals schedule very similar to other schools, in that students took their exams during a specified block of time, sitting at their desks, and working bell-to-bell in most cases. For my AP calculus class that I was teaching, this seemed a quite appropriate model, as I knew I was preparing the students for the AP test in May. For my other math classes, I also (at the time) felt the model was appropriate; students were being given a chance to demonstrate in that block how much information they had retained over the preceding semester.
But in physics, things were different. I never loved the idea of a 2-hour physics midterm or final, with every student having to manipulate the equations in the right way to get a specific answer. Such exams tend to fall to the level of "which equation do I plug stuff into to get the right answer" very quickly, but more importantly, also lead to many of the affective issues Kyle talks about in his post. So I had a strange idea, which amazingly enough I was allowed to do.
Here's what it was: About 3-4 days before the exam, I gave every student in the classroom the 20 or 30 exam questions (depended on which exam it was; 20 for final, 30 for midterm). I told them that they could work on the questions from the moment I gave them out, that they could work with others, and that they could use any resource they wished (including the Internet) to help them. Then, every student was required to schedule themselves for a 10-12 minute time slot, which could happen either during the normal exam time or during "outside" times (during my preps when I wasn't otherwise giving exams, for instance, or before or after school). This would be my time to interview them about how the questions had gone, actually doing the assessment.
When the student walked in for their session, I had a deck consisting of some cards - exactly one more than the number of questions given. The student picked 3 cards from the deck. Each card corresponded to a question they'd been given; the student then had to do the questions right there in front of me. I could help them as needed (the degree of help needed would, of course, affect their grade), or just listen while they explained their answer. The extra card was a "wild card" - if the student picked that one, they could choose which of the questions they wanted to answer.
In the year that I did this, I observed a couple of things:
- Students were much more relaxed about the "exam". They knew precisely what I was expecting them to know, and because of the random nature of the process, couldn't "game" it.
- With 20+ questions, it really became impossible to cheat, in the sense of just memorizing someone else's solutions to the questions. They really had to own the answers for themselves to excel.
- The students could, if time permitted, pick a fourth question to do, especially if they felt they'd not done as well as they'd have liked on one of the three required questions. This gave the kids more of a feeling of control over the process.
If you are interested in seeing a couple of the tests themselves, I have them available via Google Drive here and here. They're not particularly hard sets of questions, but they do a good job of getting the idea across about what topics I was covering during each semester.
In spite of this, I am giving real thought about how to resuscitate this model - somehow - in June. The model I used 10 years ago is very simple, and needs serious updating: a more specific rubric for evaluating the students' performances on the one-on-one, more interesting and authentic questions than I asked back then, and preferably a way to get more one-on-one time with each student (I'm convinced 15 minutes should be a minimum), as starters. But the more I think about it the more convinced I am that the model itself had value. What do you think?