On Expertise and Process
My internet friend Nina has a really interesting blog about design. She’s a designer herself and she often posts about ideas that are really engaging. Recently she posted about trusting process and the times when the trust we have in our processes fails.
I got involved in the resulting discussion with some really smart people. I’m pretty pleased with one long comment I made, so I thought I’d post it here in its entirety.
Here’s the thing about process: process is for people who haven’t learned to develop instincts (or for situations where instinct fails).
The canonical reference for that idea is Hubert and Stewart Dreyfus’ “Mind Over Machine” (1986). Dreyfus H is a professor of philosophy at MIT and Dreyfus S is a mathematician. (Dreyfus H is a follower of Heidegger and that tradition.) Mind Over Machine describes (among other things) a hierarchy of expertise: naive; novice; competent; proficient; expert. In that hierarchy, only an expert uses their instinct. Everyone else follows various rules, frameworks, or processes. The mark of an expert, for Dreyfus, is that an expert acts in a way that *even they* cannot decompose into rules (or that they rules they say they follow are hilariously incomplete).
So, this has a number of applications. Patricia Benner, a nurse and nursing researcher, actually published Dreyfus’ ideas in 1984 in a book on nursing practice. In it she quotes a nurse telling a story about how they look after babies in the Neo-natal ICU. Some of the nurses “just know” if a baby is going OK or if the baby is in trouble, even if all the outward signs and all the diagnostic tools say everything is going fine. So they will hover over that baby or check it more frequently. Sometimes everything is fine, but sometimes it’s not and the extra care often saves a life. But if you ask the experienced nurses what it is they see, they can’t tell you any more than vague homilies. They have passed through a stage of rule-following.
BUT! There are times when instinct and expertise fails. In “Sources of Power” (1992?) Gary Klein, a psychologist and consultant researcher, often for the US military, tells the story of a bunch of firefighters trying to rescue a woman dangling from a highway overpass. Their previous experience led them to attempt a particular kind of rescue, using a particular sort of harness that was usually used on firemen on the outside of buildings. But the woman was slightly built and she was far too small for the harness. She slipped out of the harness and was only saved by a fireman on the ground catching her(!). In a debrief with Klein they figured out that they had other rescue tools that they could have used effectively that she would not have slipped out from.
Now, to finance. Some people have argued that one of the things that went wrong to cause the GFC was that the different investment houses had their financial models which had worked really well, until the time when they very much *stopped* working. Dreyfus would argue that the best a computer model can do is competence (this requires accepting that expertise has superior performance to competence) because a computer model is the very embodiment of rules. This is not to say that having the models was bad, simply that there times when something occurs that is not foreseen by the rules and those are the times when expertise comes to the fore. But, as the firefighters example shows, expertise is only as good as the experiences that built it.
So, is process-oriented thinking bad? A process-orientation works because, Dreyfus would argue, you simply cannot communicate the process an expert follows without converting it into rules. But that leaves the expert as expert and the person, or thing, following the rules as competent.
The vast majority of the time there’s nothing wrong with competence or proficiency. Competence is lovely. Give me competence or give me a leaky sink. And the best thing about competence is it’s easy to assess.
(Actually, this may or may not be true, when I was studying for my PhD one of my supervisors other students was a nurse pursuing a definition of competence which was surprisingly, or perhaps not, proving to be highly situated and contextual. But let’s leave the constructionist sociology aside and treat ourselves to a little reducitionism and say that competence can be tested.)
So, is process a hoax or is an emphasis on process false or useless? I say most likely not. Process is itself a model or a map and as has been pointed out by many others, the map is not the terrain. Process tends to break down under extreme scrutiny as people elide over parts that they have internalised. But it’s still useful.
Finally, back to Nina’s skiing. In the post she says, “I do it exactly as I planned in my head, carefully paying attention to my markers, the physical cues and things I need to do”. In our research we have found that experts plan well in advance and then execute, and we have seen this with designers and nurses and ordinary people who are well versed at using digital cameras and microwaves. (You’re thinking, “aha! he just said experts use rules!” Maybe. The difference is that if you ask an expert to describe what they do and then watch them do it, you’ll see two different things. They will gloss over things in their pre-description that they explicitly engage with while acting and they will ignore things in their action that they described in detail. Expertise is weird like that.) From this description I would be fairly confident to say that Nina is an expert skier.
But why doesn’t her expertise translate to racing? Well, I’m not a cognitive psychologist, but with that hat on I would say that maybe Nina’s competence at racing combined with her expertise at skiing sends her expertise at skiing back to competence. Maybe it’s hard to follow the rules of racing without engaging the rule-following part of your brain, so if you’re rule following about racing, you can’t switch to instinctive skiing, but you are stuck at rule following for skiing too. And then it’s like trying to drive a manual (stick-shift) transmission fast when normally you just drive to commute. You start paying more attention to *how* you use the clutch and shifter and pretty soon you’re going slower than you think you should be and you’ve crunched gears or missed a shift. Or something like that.