Rambles around computer science
Diverting trains of thought, wasting precious time
Wed, 19 Jul 2006
Thoughts on writing supervision exercises
As an undergrad I often bemoaned the fact that Computer Science lecturers at Cambridge, unlike their counterparts in Natural Sciences, did not generally provide exercises to accompany their lectures. Instead, most lecturers expect supervisors to set Tripos questions. I found this unhelpful, and still do, for reasons I will now explain.
For one, relevant Tripos questions are often few in number, either because a course is relatively new or because the syllabus has recently undergone significant changes. Secondly, even when a good number of questions are available, they often leave some parts of the course untouched -- but there is no guarantee that the upcoming year's questions will do the same. And finally, Tripos questions are often discouragingly difficult, or infuriatingly tedious, for a student who is learning the material for the first time.
All this has led me to conclude that questions which are good for examining a course are typically not good at helping students learn. (Note that I'm sidestepping the question of whether most Tripos questions actually are particularly good for anything.) So, as a supervisor, I've been having a go at writing some questions of my own. It's difficult! But nevertheless, I feel that the results are mostly half-decent as first attempts. The questions themselves are linked from my teaching section, but the following is a list of goals which I try to ensure are satisfied by the questions I write.
- They should be complete and balanced with respect to the course's contents. For courses with heavy factual content, this might not be possible, but at least every concept from the course should be covered by the questions.
- They should be minimal. There is no good in setting two or more questions covering the same conceptual material if one would do. Most undergraduates are busy enough without having redundant questions to answer, and in most Computer Science, conceptual understanding is more important than practice. (There are, of course, a few exceptions in the more mathematical areas. I'm also definitely not trying to understate the importance of practical classes for programming, hardware design and the like.)
- They should be free of bookwork and other tedium. To generalise wildly, students are mostly interested in the material, but are easily distracted. It's easier to learn if the experience of learning is interesting. This doesn't mean that bookwork might not be important, but the really important factual material can be picked up along the way, and most of the rest is best nailed down during exam preparations.
- They should require understanding, not just knowledge to answer. Knowledge is easy to pick up, but filling the gaps in understanding is what supervisions are most valuable for.
- As a consequence, they should mostly be answerable briefly. There will be some exceptions among the more discursive topics, but generally an understanding (or lack thereof) can be demonstrated succinctly provided that questions are carefully written.
- They should be as interesting as possible. Use of interesting real-world examples and analogies can help.
- They should be motivating, in that they should provide enough guidance and internal structure to prevent the students from being intimidated or from giving up too easily.
- They should be challenging. Balancing the need for interesting and motivating questions is the intention that a student who is comfortable with the supervision exercises should not find Tripos questions unmanageably difficult.
Clearly, some of the above goals are mutually reinforcing, while others are mutually conflicting. Striking the right balance is difficult, and one reason why good questions are more likely to be the result of a collaborative, incremental process than of a one-time effort. On that note, if you have any comments on my questions, please contact me!
[/teaching] permanent link contact
