…or why you can’t make your Nanna’s banana cake they way she does.
My Nanna makes an amazing banana cake. It’s light, fluffy and moist. It has a very thin, delicate, dark brown crust and is pale yellow inside. My mum has tried for years to make this cake. She has a recipe, written down by my Nanna, that she follows. The cake that the recipe makes is not the cake that comes out of my Nanna’s oven, even though my Nanna says it’s the recipe that she follows. My mum’s cake is not the same colour and has a different texture. From the same recipe!
It’s a basic principle of “expertise” that experts don’t follow explicit rules when they’re acting as an expert. Instead they act “intuitively” and they make decisions in a blink. The action of an expert and the decisions they make are guided by their recognition of the salient features of a situation. Gary Klein calls this the “recognition primed decision model”.
Novices, on the other hand, do follow explicit rules. It’s a characteristic of a novice’s interaction with a system that they will follow rules to achieve an end. This works when you’re learning or when you’re doing something simple.
When an expert explains something to a novice, they’ll often use a rule of thumb or some other aphorism that explains how to acheive some end. “Keep your eye on the ball”. “Keep your head down”.
When I’d work on cars with my Dad I’d ask how tight a particular fastening should be. “Finger tight”. (My dad has never owned a torque wrench.) I’d always get him to check my bolt tightness, until one day I was confident enough to just do it finger tight myself. How tight is finger tight? I’ve no idea.
Experts talk in this sort of gobbledygook because they’re not following rules. They’ve completely internalised the process they follow and they quite literally can’t explain it. If you get an expert to write down the rules they follow, at best, you get a set of rules that allow competence and little more. The differences that make the difference in an expert’s performance are so subtle and so tacit that even they are unaware of them.
This was a problem encountered often in the days of much excitement over expert systems. You’d get a doctor in a room and ask him how he diagnosed, say, lung cancer. You’d encode the rules and steps he gave into software and then you’d pronounce that you had a machine that could diagnose lung cancer. Except you’d rapidly find that your machine was wrong sometimes, not just missing diagnoses but misdianosing. Because the doctor was explaining the rules, rather than being able to explain his tacit knowledge.
Nanna’s recipe is a description of a competent cake, so when you follow the recipe, all you get is a competent cake. If you want an excellent cake, you’d need to stand and watch my Nanna make a cake, and even then you’d probably not get it, because as a novice, or even competent banana cake maker, you’d be looking for all the wrong things. You’d need to stand and watch, and watch and watch and watch, for ages until you started to see the differences that made the difference.
And that’s the problem with Nanna’s recipe. It’s not that my mum is a poor cook — far from it. And it’s not that the recipe is a bad, or even just inaccurate, recipe. It’s that my Nanna is too good at cooking that Banana cake.
