For the first time ever, I am teaching without tests in my classes.
"Heresy!" you may be thinking. CS is a mostly objective field; how can I get away with not testing my students? And I have to admit that when the idea first came to me, last summer, I really struggled with it--the culture of testing is strong in my field and strong in my department. How could I make this work and still fairly evaluate my students?
The answer is slightly different for each of my classes.
The upper-level course is easier. What I want my students to demonstrate can be done more naturally through a series of projects rather than an arbitrary in-class or take-home exam. This means that I have to be a bit more deliberate in designing the projects, and that I have to provide more opportunities in class for the students to try out and practice the material. But this in turn has forced me to be more creative about how I structure class meetings, as I've discussed before.
In the intro course, I've actually just substituted weekly quizzes for the 2-3 tests I typically give. When I mentioned to my colleagues that I'd be doing this, they all said "are you crazy? isn't that much more work?" And I worried about that too. But one thing I repeatedly noticed in previous intro courses was that the 2-3 tests model just did not work for the students. It didn't reinforce the concepts I wanted them to learn; it didn't indicate conceptual problems until too late in the term, past the point where some extra practice could have righted the sinking ship; and in some cases, the tests were not fair evaluations of the students' understanding of the material (one bad day could torpedo someone's grade, or shake their confidence to the core). So I thought, why not use an evaluation model that reinforces the idea that learning CS is a gradual process in which topics build on previous topics. Quizzes seemed better suited for that. Plus, it has not been as much work as I originally thought--a nice bonus!
Good things about weekly quizzes vs. tests:
1. It's easier to make up one question at a time than 5-6 in one sitting.
2. It's easier to focus on one particular concept in each quiz.
3. It's easier to remember how I covered a concept in class and structure the quiz questions accordingly.
4. Oddly, grading has been easier. Mentally, I find myself saying "hey, they're only quizzes, so I'll just sit down and grade them and get it over with". And even though I love asking open-ended questions, I've found that I can grade them pretty quickly.
5. Students are much less stressed about the weekly quiz than they typically are about tests.
Challenges in moving to the quiz model:
1. It does take up class time.
2. Makeups are a nightmare. I have a pretty strict no-makeup policy, because otherwise it becomes too unwieldy.
3. I do have to set aside time each week to make up the quiz and then grade the quiz. This comes out of my class prep time. (But it's probably not much more than the time I'd spend making up and grading tests, so I'll call this a wash.)
It's still too early to determine if my no-tests experiments will prove worthwhile. But I'm interested in seeing if there's any noticeable change in grades (or quality of work), and also what the students ultimately thought of this experiment. Assuming all continues to go well, though, I can see myself doing something like this in future classes.