Sunday, October 21, 2012
Mad Scientists and Marshmallows
It was a nifty question, and I probably should have expected it. But since the question came out of the blue, my answer did, too.
I’d love to see a culture in which faculty use their academic freedom to experiment. In my ideal setting, they’d be working together -- and separately, as appropriate -- to keep trying different approaches to helping students succeed. That could mean different teaching techniques, different scheduling ideas, different course content, novel uses of technology, or whatever; the one thing it absolutely would not mean is doing the exact same thing year after year. I would love to see faculty as a group of mad scientists, innovating gleefully.
This story about the marshmallow experiment came out at about the same time, and I think it offers a useful nuance. Most of us know the classic marshmallow study in which young children were left alone for several minutes in a room with a marshmallow. They were told that they could eat the marshmallow, or, if they managed not to until the adult returned, they could have two. The kids who exhibited enough self-discipline to hold out for the second marshmallow wound up having better lives by a host of measures.
Apparently, researchers at the University of Rochester replicated the study, but with a twist. They had some adults come through with the second marshmallow, and others seemingly forget. Then they ran the experiment again with the same kids. Unsurprisingly, kids whose trust had been violated the first time were much less likely to defer gratification the second time. It’s one thing to wait for a payoff; it’s quite another to wait for a broken promise. The study suggested that kids whose home lives are chaotic will have a harder time in school, since they will have a harder time believing that delaying gratification will result in a payoff. At home, the promised marshmallow never comes. Why would school be different?
It occurred to me that, in a sense, I’m hoping that faculty will wait for the second marshmallow. I’m hoping that they’ll use their autonomy and academic freedom to experiment, rather than to coast (or fulminate). Which requires a certain faith on their part that there will be some sort of payoff, and that they won’t be punished if an experiment fails.
With people who are relatively new to the college, it’s easier to set a certain expectation. But with those who’ve been here longer, through various administrations, it can be hard to get past old, forgotten marshmallows. Habits learned early are hard to shake. That’s why the marshmallow study matters.
In the meantime, here’s hoping that enough security will lead to gleeful experimentation, rather than just digging in. The marshmallow parallel works in two directions, after all.
That said, it's hard to tease out the simple effect of "disappointment depletes willpower reserves" from "fool me twice shame on me" in their model.
There's also the ethical issue of experimenting on students. How to you get informed consent, and yet still conduct proper experiments (with controls and other proper experimental protocols*)?
*And I'll note that most 'educational experiments' published in the literature don't have sufficient controls to be considered valid, at least to any science that understands basic statistics.
I guess you'd like to visit some parts of my college. (Notice that I said "some". Other parts are not perceived as being supportive of the risk of innovation.) Our biggest weakness is that failures are not discussed as openly as success, yet any scientist will tell you that you learn as much or more from failed experiments than from successful ones.
Most of those developments have been undertaken by tenured faculty with the support of long-term Deans, for all of the reasons alluded to in the comments and the
"I'll note that most 'educational experiments' published in the literature don't have sufficient controls to be considered valid, at least to any science that understands basic statistics."
That would be why basic statistics are rarely enough when human beings, with all of their variability, are part of the equation. Experimental physics is easy by comparison, yet even there you will find some highly sophisticated statistical techniques in use.