Analogy and models

I’ve gone on a bit about the value of mental models in instruction (and performance). (I guess this cements my position as a representationalist!) My interest isn’t surprising, given my background. But someone recently pointed out to me an aspect that I hadn’t really commented on. And, I should! So here’s some thoughts on analogy and models.

The initial callout was me talking about models, and communicating them. In particular, I’ve mentioned a number of times the value of diagrams. Yet, someone else pointed out that another useful mechanism is analogy. And this rocked me, because of course! Yet, I’ve neglected this mention.

As context, I’ve been a fan of mental models for thinking since I got the gift of a book on them from my work colleagues as I headed off to grad school. Moreover, I did my PhD thesis on analogy! I broke down analogical processing in a unique way, and looked at performance. finding some processes could be improved. Then I tried training on a subset, and achieved some impact.

Analogy is, by the way, a useful way to communicate models. What’s important in models are the conceptual causal relationships. If there’s another, more familiar model with the same structure, you can use it. For instance, the flow of electricity in wired can be analogized to the flow of water in pipes. Another, flawed, model is saying that the orbit of electrons around a nucleus is like the orbit of planets around a sun.

So, why have I been blind to the use of analogies? Perhaps because I’m so familiar with them that I just assume others see the possibilities? Or maybe I’ve just got a huge blind spot!  Still, it’s a big miss on my part.

When you want learners to ‘get’ models (and I think we do), you can present them as diagrams. You can have people embody them through things like Gray’s gamestorming.  And, of course, you can use analogies. We have to be careful; empirically, most folks aren’t good at generating them, they focus too much on surface features. Yet, what’s necessary is sharing what cognitive scientists call ‘deep structure’, the important relationships that guide outcomes. People are good at using given analogies, but don’t always recognize them as useful unless prompted.

If, and it’s not a given, we have a familiar structure that happens to share the relationships of the model we’re trying to communicate, we can make an analogy! Though, there are nuances here too. For instance, Rand Spiro found that, when developing an understanding of muscle operation, a progression of analogies was needed to develop the final understanding!

Still, we shouldn’t ignore the possibilities of analogy. Some have argued that we fundamentally understand the world by bringing in prior models to explain. Which isn’t hard to countenance in a ‘predictive coding’ view of the world, that we’re actively trying to explain observations. Wrong models are typically an explanation for misconceptions, using  the wrong model in new ways. We have to diagnose and remediate those understandings, because folks don’t tend to replace their models, they patch them. Giving good models a priori, via analogy or otherwise, is a good remedy.

Analogy is a feature of our cognitive architecture and formal representations. It’s a useful way to communicate how the world works, when possible. Like with all things, of course, the nuances matter, but analogy and models are tools we have to facilitate understanding, if indeed we understand them. So let’s, eh?