« Storm, by Tim Minchin | Main | Masters Exams »

April 26, 2011

Why students don't learn what we think we teach

It's that time of year when professors reflect on what their students have learned in the past semester. While mulling over my own class on the design and analysis of algorithms, by chance, I was pointed at this excellent lecture by Robert Duke, a professor of music and human learning at the University of Texas Austin, on "Why students don't learn what we think we teach" (hosted at Cornell University Videos).

Duke is captivating, and he makes a clear argument that students don't learn what we think we teach because they're too busy learning what we're actually teaching, which is, often, that precision is more important than understanding and that grades matter. The solution, he argues, is to teach, over and over, the things that we actually want our students to remember after the semester is over. And, that we should not defer learning about "The Good Stuff" until after they've suffered through boring prerequisites. Instead, we should teach the good stuff first and teach what we really enjoy.

This seems like pretty good advice, and I hope to be able to improve my courses by striving to follow it. The difficulty, of course, is trying to figure out what exactly are the most important ideas we want our students to remember after the course is over. Duke argues that many professors don't themselves have a clear idea of these things, which of course makes it difficult to teach in a way that emphasizes them. For quantitative fields like computer science, I think we sometimes convince ourselves that detailed problem sets are necessary to teach the quantitative (mathematics or programming) skills the students will need for more advanced work. But how much of what we really care about depends on those? For students destined for the computer industry, or for non-computer industries, what do we as computer-science teachers really want them to remember, even if they never write another line of code in their lives? Food for thought, for sure.

Tip to Larry Hunter.

Update 27 April 2011: While I'm on the subject [1], here's another fascinating video on teaching, this one a TED talk by Salman Kahn, the former hedge fund manager who started the Khan Academy by first doing short YouTube videos of simple math lessons and then expanded it to a fairly comprehensive-looking system for learn-at-your-own-pace education. Given how little time (and money) university professors and school teachers are provided to tinker with such novel course structures, it doesn't surprise me one bit that it was a hedge fund manager who developed this idea. I would very much like to introduce some of this stuff in my university courses, but the prospect of developing the videos from scratch is more than a little daunting!


[1] For long-time readers, you've probably noticed that I've only posted videos and comics for the past two months or so. This is because I'm a new professor and new professors as a rule, true to the warnings I was given a year ago, struggle to find time for anything other developing their courses and learning how to teach. I don't find it ironic that I was hired as a professor on the merits of my ability to do things that my job now systematically prevents me from doing. Instead, I find it maddening. Putting that aside, last week, I was mulling over writing something about the recent pieces in Nature on the brokenness of the PhD system (here and here; to be honest, the articles are not as insightful as their titles suggest), which is, I think, intimately related to the ideas posed above about teaching.

That is, universities are not really in the business of providing job-training for the private sector, but increasingly that is what both the private sector and the students want universities to do. On top of that, university professors generally have very little experience outside the academy and we're hired mainly on the merits of our ability to conduct and communicate novel research to our peers, who are mainly other university professors and program managers at federal agencies. It's a strange situation, frustrating on all sides, perhaps. The strangeness is, I think, driven by pretty deep societal structures related to the priorities of the people with money (which is basically students and the federal government) and to the way employers select among applicants (academic degrees are signals of legitimacy, competency and prestige, which are easier to sort on than actually taking the time to evaluate a person individually). These things are so deep they're not going to change much anytime soon, and from my brief experience so far, seem likely to only get more pronounced. The question then, is into what shape will that pressure push universities, over the long run?

posted April 26, 2011 04:46 PM in Teaching | permalink


I think Robert Duke is a good professor. He teaches well. I have met him ones.

Posted by: names of birds at April 30, 2011 11:23 AM

I really enjoyed this article. Something like this cane up in the Austin American Statesman, my home town paper, and there was an article about some of The University of Texas professors were paid large sums of money for "research" but spent very little time teaching. I agree that if they are so good to acquire money for research that they should be using their skills teaching freshman courses.

Posted by: Bird Watcher at May 6, 2011 01:55 PM

Hey, good to find somneoe who agrees with me. GMTA.

Posted by: Vinnie at May 7, 2011 03:26 AM

Thanks for this blog! I've thought a lot about how to teach students algorithmics and what really matters to me - and what can be tested to determine the grade... Actually, I was quite honest to my students and told them that what I want them to learn is something that I cannot easily test but that our society wants us to give grade to students (there is an interesting argument about the role of universities in societies). So, my deal with them is that I will hammer in my main points in the lecture, and that the exam will be on those parts that can be tested (and what this is I make quite clear to get rid of the guess working). So, what are the important points? Here is a provisionary list:

1) What is the problem you want to solve? State it clearly. This is half the way.
2) Learn some basic tricks and techniques to solve algorithmic problems (this is the stuff we can test).
3) Understand other people's algorithms (testable).
4) Come up with your own algorithmic idea (not testable in exam, much work for teachers to test for > 100 students)
5) Learn that the resulting algorithm might not be hard to understand but that getting there is difficult (not testable, but we can make them experience it by giving homeworks).
6) Learn that some problems are not solvable in real time and get an intuition for those (not well testable).
7) Go back to 1 and change the problem to be solved if that is still acceptable (e.g., approximation instead of optimality).
8) Get an intuition to where to look when encountering a new problem (not really testable, maybe).

Anyone up for discussion? I'd be glad as there are many things for us to consider: how to deal with all the stuff available on the internet when teaching classic algorithms? More theory or more practice? Which algorithms and why? Encourage team work or not?

Posted by: Nina at May 17, 2011 02:24 PM

Where I'm from (Eastern Europe) it's much, much easier to see that. As a matter of fact, students here aren't learning what they should be learning because they are too busy learning what they are taught.

Posted by: quantum at May 21, 2011 09:16 AM

Hi Aaron, Thanks for posting the lecture by Robert Duke. It is a brilliant lecture and i also thank Cornell University for taping and making the lecture available to the larger public. What impressed me most about Prof. Duke was his ability to traverse so many different academic interests with such ease. There was an implicit thread through his talk about philosophy of science, philosophy of mathematics and i would say even philosophy of social sciences. A genuine example of liberal thought and embodiment. The lecture is like a solid Rock show for me...it keeps ringing in my head for a few days and makes me smile!

Posted by: Arvind at June 17, 2011 01:46 PM

I liked this article and agree to teach what we like - all the good stuff. Very good concepts by the way.

Posted by: toffee candy at June 26, 2011 12:24 PM