Learning to satisfice
The other day in class, I was walking students through the design of a recursive procedure that they had done on a recent homework. We spend some time developing a correct procedure The procedure had some repeated code, which I tell students to avoid . So we rewrote the procedure to avoid the repetition. Then I observed that it was calling another procedure on every recursive call and that, since that procedure was expensive, we should consider how to avoid those repeated procedure calls. Along the way, students had suggestions and questions on other improvements.
When we were done, I asked the students what they’d learned from the exercise. One of them said something like
It seems like you can always make another change to improve a procedure. Know when to stop. So I wrote the following on the eboard .
Learn to satisfice.
Satisficing is a decision-making strategy or cognitive heuristic that entails searching through the available alternatives until an acceptability threshold is met.
Yup. Stop when your procedure is above an appropriate acceptability threshold.
Alternately, stop looking for topics to muse about when you find an acceptable one.
 Upper-level students know to keep their code DRY . The introductory students are more likely to repeat code, and I’ve been working on helping them understand that repeated code incurs not only computational costs, but maintenance costs.
 Don’t repeat yourself.
 The eboard is my electronic whiteboard. Basically, I open a terminal window at the start of class, start vi, and type on the computer rather than write on the board. That lets me keep a nice record of each class that students can refer to later .
 There are some questions about whether this approach is best, since it leads some students to stop taking notes. There’s evidence that note taking helps you learn. Good students still take notes and rely on the eboards for additional detail.
 Wikipedia may have taken the definition from
Colman, Andrew (2006). A Dictionary of Psychology. New York: Oxford University Press. p. 670. ISBN 0-19-861035-1.
Version 1.0 of 2017-11-09.