Thursday, May 30, 2013

Review are In, 2013

Teaching reviews are in!  I'm happy to say students were more forgiving than last year.  But also, I notice in the reviews the effects of 4 significant changes from last year.   (The best is saved for last.)

1.  Students knew it was my last year teaching the course.  I think they were nicer to me than normal because of it.  (Pity points!)

2.  For the last several years, students have complained that the midterm coincided with "Housing Day", the day the freshmen find out about where they'll spend their later years, and apparently it's a big party day.  This year, I was able to move the midterm.  (For didactic reasons, I assure you -- some of basic material in my course is now covered in an earlier class, saving me a lecture early in the semester.)  Students really appreciated that.  (I maintain that my midterm being the Thursday before finals precedes the advent of "Housing Day" -- someone put Housing Day on the day of my midterm, not the other way around -- but students have not sympathized with this reasoning.) 

3.  This was the first year students had the advantage of taking that new class, CS20, designed to give them more background on CS mathematics.  (We finally have the CS "Discrete Math" class we haven't had but have probably needed.)  I think this helped students, especially at the lower tail, and probably somewhat helped review scores.

4.  The final change may arguably be the most important.  In the past I've given longer assignments over usually 2 week periods;  something like 7-9 problems.  At the urging of one of my experienced TAs, who both wanted the grading split up more and thought the students would prefer it, I broke up the problem sets, so they were due weekly, and were usually 4 or 5 problems.  The feedback from many students was that they liked this approach better (obviously not from direct experience with the class previously, but from what they had heard from other students).

To me, this remains counterintuitive.  The students were getting the same problems either way, so the splitting only added an additional constraint on them.  Instead of having eight problems over two weeks, they were forced to do the first four in week one and the next four in week two.  But, clearly, for psychological reasons many students want (need?) that constraint.  As some have explained to me, they aren't going to start the assignment until they're close to the deadline, so the additional deadline prevents them from becoming overloaded and overstressed by a longer assignment.  Perhaps, beyond the psychology, part of the issue may be student collaboration -- more frequent shorter assignments introduces constraints that probably help encourage scheduling of working together.   

I worry about the time management skills of Harvard students.  Or I suppose for many it's just the way they live -- their schedules constantly packed full with deadlines serving as the basis for their priority scheduling.  I hope they experience a different lifestyle at some point.

However, lessons learned for whatever undergrad class I teach next!  Short weekly assignments.  And be sure to avoid student (non-academic) events when setting up the midterm in the class schedule.

Finally, I still suspect my reviews would be non-trivially better if they happened after grades were out.  Students often think they're doing worse than they are -- they don't see how much the curve helps them.  For example, one senior, after the final, came up to me worried that he/she did poorly enough that he/she would fail the class.  I asked how he/she had done over the semester, and said it seemed very unlikely, but I'd send mail after grading the final.  The student got a C and was in absolutely no danger of failing.  The horror stories of CS124 have been somewhat exaggerated over time -- perhaps all the more reason a "refresh" is in order.  

Anyhow, thanks again to this year's CS 124 class -- for those of you who aren't graduating, I hope to see you around, and for all the students, if you've read this far, I hope you'll send me stories when you find whatever you learned in CS 124 to be useful to you. 



Russ Ross said...

I can attest to the impact of #4. I teach our CS121 equivalent, and had problem sets that were each one or two weeks long. After a couple years I split the assignments up into pieces so that there was a deadline at the beginning of each class, i.e., two deadlines per week for a Tuesday/Thursday class. The result was a significant increase in completion rates for the problems.

In addition to the time management issue, I saw another psychological factor at work. With a two-week assignment, students were often willing to give up after completing a few of the problems, leaving the harder problems untouched (or at least blank on what they turned in). They seemed to feel that they had at least made an honest effort by completing part of the problem set. When I switched to smaller, more frequent problem sets, they had to make an effort on each problem to feel that same sense of justification.

As an example, consider a problem set with some "warm up" problems followed by an undecidability reduction or an NP-completeness reduction. It was all too easy to do the warm ups and skip the "hard" problem at the end. When a single deadline had nothing but that reduction, more students gave it some substantial effort.

An added bonus was that the resulting problem sets were small enough that we could review each one at the beginning of class--minutes after most of them had submitted their work--without using up too much class time.

Most of my students are at the other tail of the distribution from yours, so it is interesting to see similar dynamics at play.

Paul Beame said...

I noticed precisely the same effect wrt weekly vs biweekly assignments in my algorithms courses. At the undergrad level I assign weekly assignments, but teaching it at the grad level for the first time in many years, I thought biweekly ones and less of a drumbeat would be appreciated. I assigned the same volume of work in the grad course (albeit a bit tougher) and I was inundated with workload complaints. Students did start a week before the due date rather than 3 or 4 days prior that they might have done with weekly assignments, but I think the effect of having so much more work pending at one time was very stressful.

rrenaud said...

Predictibly irrational has a whole chapter dedicated to your point #4. The basic story is that people will generally procrastinate way too much and that Harvard students are people.

Arjun Narayan said...

This paper, by Dan Ariely (author of Predictably Irrational) is the relevant citation to rrenaud's comment. The abstract neatly states the point that intermediate deadlines help tremendously.

Anonymous said...

Yay Michael you're the best!!!