Metaphors, Holistic Learning and Digression

I just read an interesting article on Lifehack on the relationship of metaphor and learning. It happened to spark off a fairly detailed stream of thought, so I figured I'd blog about it. Give it a read first, it's fairly short.

Now, when I first read the article, I immediately liked it, because I've actually done a bit of thinking on stories and metaphor in a more personal way, applied to how I tend to remember things. Anyone who has had a long discussion with me (or who reads this blog) will notice that I have a very strong tendency to digress all over the place. This isn't just something I do while blogging. In most discussions, I'm not just explaining my point of view - I'm also feeling out new connections as I think. It's not a conscious decision - I just do it naturally, out of some subconscious drive or something. I'm the guy who always goes "This reminds me of..." and then I come up with something that no one else in the conversation has been reminded of. I actually realized how weird this was when a friend pointed out several years ago that I always got reminded of incredibly unrelated things during any discussions. I had no idea that it sounded odd to other people until he told me.

(And that's not the only thing - I didn't realize my eerie ability to detect spelling errors instantly by merely glancing at text until my dad noticed me doing it and pointed it out. Maybe I'll blog about that some time. There's another thing that I didn't know about until someone pointed out, but since it deals with stories and memory, I'll talk about it in a bit more detail, but not in this paragraph. And that's because this parenthetical paragraph is really another digression...)

One of my programming team coaches pointed out last year that I often quoted something from some book I had read, and usually did so with extremely high levels of accuracy and detail that you wouldn't normally expect from someone making a mostly offhand comment. I suspect those details are unnecessary and boring, so maybe he was giving me a hint...

At the time, I thought about it for a moment, and agreed with him. Later, I remembered that conversation again, so I thought about it a bit more. The best explanation I could come up with was that all those things I quoted often formed some kind of coherent narrative, or at least part of one. And as the article that inspired this post points out, stories are really easy to remember. On top of that, I've been reading very intensively since I was maybe five (clearly I'm obsessive), so I process stories better than less trained people, so to speak.

Why are stories easier to remember, though?

I suppose there are several reasons, but the central theme of them all is the same - the human mind is optimized for dealing with social complexity. Just ask anyone who is addicted to a soap opera, and chances are they'll be able to narrate everything that happened from the beginning without a hitch. We take that for granted, but consider what an incredible feat that is. You have a terribly unimaginative plot, characters that all fall into a small number of classes whose members can be (and often are, for the purposes of relationship drama) interchanged, and a bunch of phenomenally dreary interactions between them, drawn from a pretty small set, on top of that.

And now consider the fact that despite all that, no matter how much you hate soap operas, just watch a few episodes, and you'll have near perfect recall of what happened a few weeks later.

Now try reading three chapters of a bad textbook, and see what you remember at the end of the month.

Works the same way for songs - people who can't remember stuff they've read will nevertheless be able to sing hundreds of songs with perfect command of the lyrics, and they don't find it surprising. (Ah, if only we found more things surprising, the world would be a better pla....er, never mind) This happens because of the rhythm and repetition, the way the words follow the music, and the associated emotional experiences. Our brains are extremely good at processing vivid sensory and emotional experiences, not to mention music. There you go - instant memory aid.

Seriously, though this sort of optimization shouldn't really be surprising. Closer to my field, think about recognizing faces. We do it all the time - hell, we can tell twins apart with relative ease if we're around them for a while. But getting a machine to do that is phenomenally difficult. One reason it's so hard is that we expect so much - we can do it effortlessly, so we expect great things from our computers. Doesn't work, of course - we've got specialized brain hardware that helps us out, over and above our complex visual systems.

And now observe how this digression of mine comes around to something useful. There was an episode of My Brilliant Brain about chess grandmaster Susan Polgar, which showed how parts of the brain normally involved in face recognition lit up when she looked at chessboards. Sounds like a pretty good explanation for the phenomenal chessboard memory exhibited by chess experts, doesn't it? Faces are pretty easy to remember because we notice certain features and how they fit together. It's a form of chunking, really.

Interestingly, if you set up chess pieces in some configuration that won't ever occur in actual play, chess players don't do any better remembering those than normal people. Presumably it looks like a picture of a face cut into weird pieces and reassembled in ways that don't make sense.

So it should be obvious now why it'll be easy to learn stuff that you can map into a narrative of some kind. But what about all the rest? How does one map a mathematical proof to a story? Some things are just too far removed from our experience to 'narrativize', if there is such a term.

This doesn't mean you should give up there - far from it. There are all sorts of other things our brains are optimized for - you can use them too. For instance, muscle memory comes in handy when you're solving the Rubik's cube. It never take me more than 4 minutes to solve any configuration (unless someone is cheating by switching stickers around ) and people who are watching always ask me how the hell I'm going so fast. Well, first off, I'm pathetically slow - the world record is 9.41 seconds, if I remember right. Secondly, once you learn a method for solving the cube, a bit of practice makes the component motions second nature, so you don't have to think about them anymore. That gives you a lot of speed, and it's no different from riding a bike or driving - stuff that you did consciously becomes automatic.

So that's one reason not to be discouraged about learning. Another reason is that you can always use a second-order effect to learn more easily. There are many skills that become automatic once mastered - stuff like basic algebra, for instance. The concepts involved can become the building blocks of more advanced concepts. So you expand the coverage your narratives have by embedding well-known narratives in them as atomic components. (Star Trek fans, remember "Darmok and Jalad at Tanagra" ) Mathematicians do this all the time - as Von Neumann once said, there are things in mathematics that you don't understand so much as get used to.

It's helpful to try and spot underlying themes for both explanation and understanding. Most fields build up complex concepts from simpler ones, but when you're trying to grasp a lot of stuff, for an exam or something, it's not that easy, and many times it's downright unnatural. It's like thinking about quarks and leptons - sure, they're the building blocks of matter, but when was the last time you saw a parent telling a kid about the subatomic composition of an elephant? He'll say stuff like, "It's big, it's got a trunk and big ears..." and so forth. Sure, a particular configuration of quarks and leptons makes up an elephant, but that's not the right scale at which to think about your friendly neighborhood pachyderm! (Bit of an odd neighborhood, that...) So it makes sense to see how the big pieces fit together, because they're easier to deal with, and then go down to the basics.

Computer Science people, think about Automata Theory. What's easier to understand - Turing machines as a big tape with symbols on it and something messing with them as they go by? Or the formal description with transition functions and tape alphabets and what not? Sure, the second is more precise and fundamental, but the first is closer to human experience, so you can reason about it more easily. So you start with students using the first explanation, and then develop the second one, which gives them time to tie the two together. Eventually you can bounce back and forth between the two, and even use the tape representation as a thinking tool (What if I have two tapes? Suppose I could only go forward...Maybe there are only a fixed number of symbols allowed....). You can look at the other automata in similar ways, and there are multiple ways to tie it all together - the Chomsky hierarchy, the languages accepted by each class, the differences and equivalences an so on.

It's a little more explicit in CS and mathematics than in other fields, but hopefully I've made my point.

So the next time I start boring you with something totally unrelated, see if you can spot the weird web of connections between all the things I say. You might find it enlightening, and more importantly (yes, I chose this word with a great deal of care), I probably would too, if I wasn't consciously aware of the links and someone told me about them.

And in conclusion, I shall mention that I was suddenly reminded of something Eddie Izzard used in his stand-up routine Dress To Kill. He's poking fun at the way he jumps from one topic to another seemingly unrelated topic. So when he starts talking about The Great Escape, he explains it like this.

"So you've got a bunch of British actors, and I'm British - link up there. Steve McQueen, action hero - I'm an action transvestite, link up there."

PS: I strongly recommend watching this guy's stuff - a funnier comedian I have never seen.

PPS: I just got reminded of something else that I've been musing about, which struck me as being related to this post - has to do with the AIs of the Golden Age scifi trilogy and how they think. I'll leave that for a later post.