One common criticism of automated "Skinner-box"-like teaching machines (e.g. in Watters's book) is that they're fascistic, inhumane, etc. In the context of K-12, that's definitely true, but I think the stronger criticism is that they don't really *work*, even on their own terms!
Conversation
That is, even if a student mechanically does exactly what the Skinner box (or Khan Academy exercises) asks them to do, the resulting understanding is usually brittle, shallow, and short-lived. *Also* the experience is often awful, but that seems unimportant if it doesn't work!
2
15
It's funny—when I was working on K12 edu, what really bothered me about teaching machines was the fascistic, anti-creative bit.
Now that I'm working on expert learning, I have a different perspective: if such a machine truly worked, I'd *love* to use one for topics I care about.
2
1
21
The voluntarism makes all the difference for me. In the context of a coercive learning environment, the *affect* of the teaching machines really bother me; but if I'm just trying to efficiently learn topics I need for projects that matter to me, then sure—whatever works best!
4
20
Questions I'd like to understand better:
Intelligent tutoring systems seem to produce more flexible, durable understanding. Is this true? If so, what differences from a Skinner-style machine make it so?
3
8
I think the major differences are not between the technology but the underlying theories of learning, i.e. behaviorism vs. cognitivism. Under the sway of behaviorism those machines didn't even attempt conceptual instruction.
1
1
You ask about Direct Instruction, but at least Direct Instruction had visuals (forgetting the human teacher)! Compare these videos of the two methods and imagine trying to learn multiplication entirely via text interaction.
youtube.com/watch?v=jTH3ob vs
1
1
A contemporary math learning system like Cognitive Tutor is capable of prompting for self-explanation and asking students to analyze mistakes, things that neither DI nor behaviorists would deem useful. sciencedirect.com/science/articl
1
1
Incidentally, when thinking about why it's hard to do something like Cognitive Tutor for expert-level knowledge, I think it's essentially a curricular problem. We don't know the learning trajectories, the best way to think about component skills, the useful questions to ask, etc.
1
2
Replying to
Yes, that certainly seems to be part of the problem. Hard to make the "knowledge model."