kasku.net/blog/2022/11/26

Learning

Lectures

Andy Matuschak's Why books don't work also talks about why lectures don't work. However, I don't endorse any other part of the article.

Of course, good lecturers don't usually believe that simply telling their audience about an idea causes them to understand it. It's just that lectures, as a format, are shaped as if that were true, so lecturers mostly behave as if it were true.

This explains the dissonance between the fact "everyone knows" that lectures aren't a good way to learn, in general, and the importance people give to them in educational institutions.

Another facet of this is that group instruction generally doesn't deal well with a diverse group of people. Lectures can't deal well with an inefficiency like "70% of the people present already understand this part of this concept", and that's if it's even possible for the lecturer to find that out. Ultimately the author is right that most justifications for lectures are post-hoc rationalisations.

However, Gilbert Strang's linear algebra lectures are sufficiently well optimised that it's probably the best way to learn about the topic. It feels incredibly close to how efficient having an expert explaining it to you personally would be, primarily because he understands the learning process so well and has a strong ability to empathise with learners and solve the problems that they have. The best teachers always make it feel like they can read your mind.

But most lectures aren't this good. I wonder why not. Maybe because most lecturers aren't this good — I think in a lot of cases lecturers don't really have formal instruction in teaching, which I'm sure is a contributing factor. Still, I think we should be honest about the limits of how good lectures can be, both in the sense of how good the best possible lecture is (I conjecture that the above ones come close), and in the sense of inherent problems with the format meaning lectures can only be so good.

Principles of Learning

Morton Ann Gernsbacher's Why internet-based education? "illustrates five ways that Internet-based higher education can capitalize on fundamental principles of learning".

Most of the individual points make sense, so I'll leave you to read it on your own rather than quoting all of it and saying "yes", but I'll quote the following, because the fact that this is the opinion of the author isn't obvious from reading the rest of the essay:

Indeed, it is likely that any medium will lead to more successful pedagogy if it capitalizes on fundamental principles of learning.

which means it's not quite "advocating" for internet-based education in general, which would be subtly different somehow.

We've known about these kinds of principles for a while now. They aren't new, but in general, higher education still follows patterns established hundreds of years ago, which means that as time passes the gap between what we do and what we know is a good idea is widening. An interesting thought experiment is imagining an optimal system of instruction based on what we know about how learning works, without the preconceptions of the structure of existing systems. As an exercise, compare this "objectively optimal" system to a real one, and judge whether there are good justifications for the differences.

Another important fact that educational institutions ignore is that despite people often applying their expertise and knowledge of one field to a different one, deliberately providing an environment for this "transfer of learning" to occur is extremely difficult. A deprecated example of failing to do this is the idea that learning latin will make you more intelligent in general. Unfortunately, in some ways we haven't advanced very far from this. Language instruction is a field where you can easily test that people who've been taking classes for several years often aren't proficient in the language at all. However, you'll still hear people justify the system of education, like "it's about having a solid base", which I would call a post-hoc rationalisation if it even sounded like it made sense.

The reason it doesn't make sense is because the idea that you can get good at producing and understanding a language by doing something sort of adjacent to producing and understanding a language isn't true. At best, memorising words and grammatical rules lets you supervise the language acquisition process as it happens naturally. Ultimately, activities which resemble speaking and using a language will help people learn to speak and use the language, and different activities will be only marginally useful, and certainly do not constitute a "solid base", since the extent to which they are useful is essentially as a supplement.

We know how language acquisition works, but language instruction fails to provide a good environment for language acquisition to happen in, which means people don't acquire the language. I've made lots of generalisations, and I'm aware of exceptions to the above, but what I've described is what characterises language instruction, at least where I live.

Indeed, I was lucky enough to have such an exception as a teacher. The contrast really drove home how low the quality of the average language course was. The teacher in question was one of the mind-readers that I mentioned earlier. However, he wasn't some super-genius or anything. He just repeated a few strategies that he knew were effective, and made sure to make use of every minute, which added up to hours of extra practice compared to other classes. Unsurprisingly, most of what we spent time doing was speaking and using the language — right from when our class were complete beginners. It's almost as if the idea that you need a "solid base" before you can start using a language is incorrect.

Sometimes people act like only a chosen few can be good teachers if I bring up these kinds of examples, and I don't think that's true at all — it's possible to improve at teaching, and mythologising people who had success using a simple strategy only acts as a barrier to improvement.