So far, we’ve focused on brains, and implications of how brains work:

Now we’re going to move up the abstraction stack a bit, and talk about how brains do stuff, and especially how they learn to do stuff. Our discussion will use concepts like patterns and metacognition, created by learning researchers:

Although these concepts emerge from the way head meat works, it’s easier to understand and use the ideas if we think of them as existing in their own right.

In particular, we’ll focus on deep learning. It’s a catch-all term for learning that goes beyond memorizing. Things like critical thinking and problem solving. Deep learning is particularly important for skills courses.

# Deep learning and memorization

Doing tasks requires memorization, of course. For example, someone asks you whether black labs are on average heavier than dalmatians, and gives you a data set about dogs. If you remember t tests and p levels, you’re on the way. Having to look everything up all the time would be quite inefficient.

Memorization vs deep learning is a false choice. It’s a matter of balance. Given that students have limited time to learn, a better comparison is:

- Lots of memorization and a little doing, versus
- A little memorization and lots of doing.

Teaching for memorization is what we’re used to. It’s easy and cheap to test memorization with multiple-choice quizzes. But is that the best way to help people learn skills?

# Skills for RL

(RL: real life. That’s what the kids say these days.)

Suppose Stew Dent take a course on spreadsheets. What would we want Stew to be able to do, six months after the course was over?

Obviously, we’d like him to be able to create basic spreadsheets. He’d have to know spreadsheet mechanics, like using relative formulas, formatting numbers, using functions, etc.

Is that enough? No. Nobody’s going to say, “Hey, Stew! Make a spreadsheet.”

They’ll say, “Hey, Stew! Here’s product defect data for the last three years. Figure out if things are changing, good or bad.”

Stew has to answer a question about a data set. He has to know what analyses would answer the question, and how to do those analyses. Only then can he put his knowledge of spreadsheet mechanics to work.

Stew has to know how to think about his own work. How does he know that he’s making progress? How does he know when he’s finished? How does he check his work? If he finds a mistake, does he know how to correct it?

Shallow learning means knowing just the top part. Deep learning means knowing both.

# Deep learning research

Let’s talk about how skilled brains solve problems, and how we can help students skill up their brains.

*None* of the ideas in this section are original to Cyco. Cyco just makes it easier to use these ideas.

## Add new comment