Teaching Ideas
I had an experience in class recently that bears on the issue
of how to teach and how to evaluate teaching quality. The question
arose of how to learn whether moderate drinking of red wine reduced
the number of heart attacks. I asked the class of twelve or so bus
econ seniors— an elite group within my elite business school– how
they would attack the problem.
The first answer was to do an experiment, having some people drink
wine and some not and seeing what happens. That was a good answer.
I went on to ask what I should do if I didn’t have enough money to do
a large-sample experiment changing people’s behavior and all I
could do was to use a survey to
ask them about their normal behavior. The answer was that I could
see if drinking was associated with heart attacks.
My next question was about a problem with the survey approach. Maybe
rich people drink more wine and have fewer heart attacks, but it’s
because they are rich, not because wine reduces heart attacks. How
could I deal with that problem?
After some thought, a student suggested that I could construct my
sample so I had equal numbers of rich and poor people, and another
suggested that I could look at the two groups separately. I said that
medical studies often were of only men or only women for the same sort
of reason, but that income was hard to deal with because it varies
continuously from low to high, not as a binary variable. I asked them
to keep thinking, and I used words such as “correlation coefficient”
and “variables” in the hopes of stimulating their subconscious minds.
I told them that this was an occasion for making a link with other
classes.
Finally, after a long time, one student did come up with the word
“regression”. I didn’t ask him to explain– I just put up on the board
that a regression could estimate the separate effects of income and
wine drinking using the formula:
Heart Attack Probability = alpha + beta*WineDrinking + gamma*Income
We then discussed what that meant, whether it would separate out the
two effects, how difficult it is to include all relevant variables,
and so forth.
The point of the story is that regression analysis is something all
these students had learned in their business statistics class and many
of them had learned in a bus econ econometrics class that in
essence studies nothing *but* regression analysis. Yet they couldn’t
make the connection. And if they don’t know when to use regression
analysis, what use is learning how to do it?
Teaching students how to recognize when a given techique is useful
is hard. First, there is the problem that students do not expect ideas
from one class to be used in another class, much less in real life.
Second, while it is easy to describe when a technique is useful, what
the student needs to learn is how to extract that technique from the
myriad of ideas in his memory when he encounters a problem outside
the context of the class. If I had reminded the students of how to
do regression analysis at the start of the class, they would (I hope)
have been able to think of it at the end of the class, because that’s
the pattern of instruction that they are used to.
Having a teacher who is wonderful at teaching students how to do
regression analysis may be much inferior to having a teacher who can
teach them how to remember to use it. How can we measure this?
Student
evaluations are of little use even in discovering how well a teacher
teaches the technique, and are entirely useless in discovering how
well the student can apply the technique during the next semester.
Regression analysis itself might do better for looking at technique-
teaching, if we are allowed to
require two teachers to use the same test and we compare performance
using enough semesters to get statistical significance. But to do
that for technique-choice teaching would require a surprise test
enough time after the class finished that the students would not know
what technique was being tested.
The instructor himself cannot do much better. It’s hard even to tell
how well students are learning a technique if you don’t talk one-on-
one and they don’t ask questions in class. (Usually the only way you
know the bottom students aren’t understanding something is if one of
the top students asks a question that shows he doesn’t understand.)
When it comes to knowing how well students are learning to recall the
ideas you are teaching, it is even harder.
What should we conclude from all this? Perhaps that we shouldn’t
devote much time to trying to evaluate teaching quality, because we
can’t. Perhaps even evaluating learning quality is not all that
useful, because all we are measuring is IQ and effort. But mostly,
this underscores the important of teaching critical thinking as
opposed to technique.