Douglas Hofstadter, of Godel Escher Bach fame, gave a lecture at Stanford this evening. I happened to hear about it, and convinced DocBug to go with me (which worked out great when I didn’t allow enough time for traffic because he was able to save me a seat despite a standing-room-only crowd). Hofstadter’s a great speaker. Interesting and personable, with entertaining stories to illustrate his points. Plus he used hand-drawn transparencies on an overhead projector. Old skool, dude.
I was intrigued by the title of the talk, “Analogy as the Core of Cognition”. I thought I might agree with it, but I wanted to hear more about where he was coming from. I liked one story he told to illustrate what he meant: his one-year-old daughter was playing with a dustbuster. Press the button, it made noise. Press the button, it made noise. Then she found another button on the dustbuster. She pressed it. It did not make noise (it was the one to open the dustbuster and empty the filter). And then she looked at her father as if to say “Yo, what’s up?!” (note: not actually how he phrased it). Making the analogy from her own experience, she had expected all buttons on this object to make noise.
Hofstadter’s theory is that all thinking and cognition comes back to this same core use of analogy to one’s own experience: “This is like that other thing, so it should behave similarly”. I think I may agree with him. I had started a series of posts a couple months ago (part 1 and part 2) that was exploring that cognitive process of how we fill in information that we do not know by assuming it was like our previous experience (I need to go back and finish up that series of posts). Along similar lines, Hofstadter made the point that there is no cognitive difference between a single memory trace and a category or concept. As soon as we have a single cognitive element, we can relate other cognitive elements to it. He used the example of the statement “Maybe there are two or three Einsteins in the audience”; there was only one Albert Einstein historically, but he defined a genius “category” into which others can now be placed.
One thought-provoking point was on the process of chunking, where we group things together, and then group the groups, and so on up the chain (e.g. labradors and retrievers and poodles are grouped into “dogs”, and then dogs and cats and monkeys are grouped into “mammals”, and mammals and reptiles are grouped into “animals” and each category becomes more abstract). Hofstadter said that we build up these groups from specific examples, and so we can always deconstruct the groups by looking inside. What struck me about this point was that there are times when I am asked to look inside an abstract concept and I feel like I have to construct its constituent members on the fly, rather than deconstructing the concept from the way I originally made it. But when I started to think of examples from my own experience, I realized that they were all abstract concepts that I had not devised (e.g. somebody once asked me to explain some economic principle that I’d quoted from The Economist). It was somebody else’s abstraction, which is why I had to struggle to construct its constituent members. This process of being able to use somebody else’s abstractions (or analogies as Hofstadter would have it) is interesting, especially in the sense that we can use them even if we did not construct them ourselves. But I’m not sure where I’m going with this, so I’ll stop now.
Another point I really liked was that he emphasized that these analogies are entirely in our heads. They may relate to things in the outside world, but the analogies are between our mental representations of those things. We sometimes project these mental analogies to the outside world, but it is not the world’s fault when they don’t apply.
He mentioned the idea of “unlabelled concepts”, but then skipped over it because he was running out of time which was a disappointment because I’m intrigued by it. The example he used in passing was that there are certain experiences he had had, but completely forgot about, because there was nothing for the experience to relate to and no easy way to label it. But when a new experience evoked the old one (e.g. his one-year-old’s disappointment with the Dustbuster reminded him of one of his own childhood disappointments), the analogy tied the experiences together. I’ve definitely felt this experience myself, when a bunch of unrelated concepts finally align into a structure and everything locks together into an “Aha!” moment. I tend to think in architectural structures rather than bilateral analogies, but I think it’s the same idea.
One minor quibble I had is that he stretched the idea of an analogy so far that it was unclear what he meant by it at the end because he used it differently in different situations. But since I’ve done the same thing myself, I’m sympathetic.
It was an interesting talk. I liked that he took a radical position, and tried to defend it. I’m not sure he entirely succeeded, but he got me thinking, and that’s a good thing. The talk reminded me that I need to do more things that make me think, whether going to good talks, or reading books, or talking to people that challenge me. I let myself get too lazy and set in my thinking habits. And I really enjoy stretching my mind – I could feel my mind revving up trying to take Hofstadter’s ideas and relate them to my own experience and figure out if they made sense to me or not. More thinking. Less lazing about. Since I’m currently unemployed, I have no excuse for not getting back to blogging about abstruse philosophical concepts again.
Bug and I chatted for a bit afterwards, and I brought up my difficulty with the construction of chunking, and he asked me a really good question, which is how do we decide which analogies to make? We have a network of analogies in our mind, and we get a new concept (whether an experience or an object) – where do we hook it into our network and why? I said that it was probably a matter of feature recognition – what other things already in our minds did it most closely resemble? Bug pushed back and said, okay, what are the features and how do we measure similarity/resemblance? He suggested that it comes down to our own experience. We make analogies based on the ways in which we’ve perceived and experienced things, like Hofstadter’s one-year-old with the dustbuster.
One thing I thought of on the drive home was that this network of analogies is one of the reasons I sometimes get tangled up when I’m trying to explain something. I’ll relate it to one thing, and then when that explanation doesn’t seem to work, I’ll relate it to another thing. And they’re all connected in my head, but the connection may not be obvious to somebody else who does not have the same experiences as me, so I end up confusing them more (I was chided recently to “keep my story straight”). I need to remember to make the connection between viewpoints more evident, especially when switching. Of course, if the person I’m talking to believes there’s only one “true” way of looking at things, changing viewpoints will only confuse the issue. But that’s a separate personal jihad.
Okay, this is totally scattered and incoherent, but I’m going to post it anyway. Editing is for wusses.