A few tidbits that caught my eye on computers and education. First, the inspiring Conrad Wolfram adding his powerful voice to an argument that is now 30 years old…whether we should still be wasting the massive amount of time we do teaching students to calculate by hand using algorithms they memorize, but neither understand nor will ever use.
One thing he doesn’t get into…rote methods are certainly inappropriate for this computerized age because computing has fundamentally changed both mathematics and will change education, although not without a fight, but what’s more–there is really very limited evidence they worked in the past. As somebody who has spent a lot of the last 20 years doing work in math education, the nostalgia for our parents’, grandparents’, and and great grandparents’ modes of teaching and learning math would be funny if it weren’t so sad. If they are around, go ask them what it was like, and then reflect whether you’d go back, and also what it did for them? (For the mathematically elect it worked, everybody else it was a dead letter).
Wolfram’s ideals–problem solving, real world problems, real engagement with computing as part of thinking, and assessments that make this visible, are mine. But the question does remain, what do you do in class as a teacher? What does it actually look like? Pedagogy–although there is much talk about scientific testing thereof–often seems like pseudo science. For one thing, real tests on actual students raise ethical questions, time/longitudinal work is hard to do, and controlling for variables is logistically impossible. (I went to seven schools before I was 14, just for one factor, I also grew up with writing teacher father and journalist mother. I wasn’t a very tractable English student, and would have stood for being a lab rat even less.
What I’ve wondered about for years is whether you could take educational data and create a model of an educational experience in a computer, including necessary data on the students, the teachers, the environment, and the content, and then feed in various pedagogical approaches and see the results? You could run as many trails as possible speeded up as you want, and watch these simulations to see what happen.
One problem with this is that learning (as opposed to scoring tests) has been very poorly modeled in computers until recently (and even now it’s still pretty primitive–the not exactly earth shattering discovery that students take many different paths to solving problems is enough to get you tenure at Stanford apparently, although doesn’t seem like something a grade school teacher takes more than a week or two to notice her first year!).
A piece in the Post by Joel Achenbach on computers that that learn the way humans do. It too is pretty primitive in comparison to my automated learning simulation lab idea, but at least it suggests that there is some plausibility that such a thing might happen.
By the way, on an unrelated note: The Post is reading better than it has done in years. I attribute it, right or wrong, to Marty Baron, and certainly not to Bezos, beyond his hiring of Baron.