Chapter 14. Complex, Adaptive Systems

Ollie picked up the phone in his office. It was Raj on the other end. “Hey, buddy, are you looking at the email I just sent you?” he asked.

“Not yet,” Ollie replied, scanning his inbox for an email from Raj. It was hard to miss, with a subject like “COMPLEXITY!!! SCIENCE!!!” Ollie opened the email and saw a wavy diagram labeled “Cynefin.”

“Dude, I’ve been researching this complexity stuff that we’ve been talking about, and I found a Harvard Business Review article from 2007 by David Snowden and Mary Boone that introduces this framework called ‘Cynefin.’ It’s a way of thinking about complexity. And this framework is not just philosophy—it’s based on complexity science, and it’s practical, it’s usable.”

“Whoa, Raj, slow down. I’m looking at the diagram, but I’m not sure what to make of it yet.”

“All right. What we have here, my friend, is a framework that buckets systems into one of the five domains: obvious, complicated, complex, chaotic, and, finally, unknown.”

“I see the first four... Where is the unknown one?”

“It’s right in the middle. It’s also known as ‘disorder.’”

“OK, got it. Go on.”

“Let’s start with obvious systems. Imagine a fast-food chain restaurant. The deep fryer station. There’s a very well-understood process of how to make fries reliably. It’s documented, and it’s highly optimized. There’s a best practice, which can be taught and followed by anyone with only a little training. That’s how you can get the same fries in any franchise, from Venice, Italy, to Venice, Florida. The process for making fries is highly constrained; the cooks don’t get to choose the type of potatoes—they come frozen, precut, and are made from specially grown potatoes. There’s even a playbook for when things go wrong: Fries too soggy? See page 39, and turn up the heat by 5 degrees. No room for creativity here, just follow the process.”

“This is making me hungry, Raj.”

“Buddy, the following will make you even hungrier. Now imagine you’re in a fancy, five-star restaurant. You’ve got a kitchen full of highly experienced chefs, and on this day they, too, are making fries. But this time, the system is a lot less constrained. The chef went to the market that morning and bought some blue potatoes in season, and some organic leeks. She picked up some truffle oil, too, and is now in the kitchen making fancy pommes frites. She’s following a basic recipe, but there’s a lot more creativity and a lot more experience required. You know what else she’s doing that no one at any fast-food restaurant does?”

“Making a living wage?”

“Ha! Yes, and she’s also tasting the frites along the way, adding a little salt, or a little more scallions. So in this domain (the complicated-systems one), you still have a recipe to follow—what you might call a ‘good practice’—but you need a whole lot more experience in order to operate here, especially when things go wrong. That’s why this one is called the ‘domain of experts.’ Also, complicated systems are far less constrained than obvious ones, but both are still within the realm of ordered domains.”

“OK, I’m with you so far.”

“Now, my friend, imagine you’re watching a cooking competition show. In this particular episode, the 12 participants are given some very loose constraints: ingredients—which feature potatoes—and one hour. Each of the chefs has a lot of freedom to cook whatever and however he or she likes. There are no recipes that they must follow, and they have to improvise a great deal. If we zoom out to look at the entire show kitchen buzzing with activity, we see a dozen parallel experiments going on. Even if this is an elimination challenge, for most of the contestants, these experiments are safe-to-fail experiments. They won’t die, even if they get eliminated from the competition. We expect some of the dishes to be mediocre, some of them possibly downright bad, and maybe one or two standouts. But no one knows in advance who’s going to win, and previous experience is not that valuable here.”

“Potato ice cream might just be the best or worst idea in the world.”

“You’ve got the idea about the kinds of things that could come out of these complex, adaptive situations. OK, so now, let’s move into the chaotic domain, which, like the complex domain, contains unordered systems. Imagine you’ve unleashed a gang of toddlers into the same kitchen. No constraints. They’re running around with knives, throwing things around, locking one another in freezers, and of course, playing with gas stoves.”

“A toddler with a sous-vide machine! What a nightmare!”

Raj laughed. “These chaotic systems are, luckily, very short lived. Someone usually steps in and sets some constraints, moving the system into a more ordered domain.”

“‘No running with the knives’ might be one of the constraints.”

“Indeed. In both the complex and the chaotic domains, there’s no best practice—not even good practice. We can’t predict how things will go in advance, and we might not even be able to establish causality in retrospect. The point is, we have to act differently than we do in the more ordered domains, if we find ourselves working within these unordered ones.”

“This is pretty cool, Raj!”

“Yup. So within which domain do you think we most often operate?”

“Certainly not the obvious one, even though there are small armies of consultants that will sell you a ‘best practice’ for pretty much anything. I’d say we’re mostly in the complicated or complex domains. And when we have outages, almost certainly in the complex or chaotic ones.”

“Right. Assuming you’re working within a complex system is probably a safe default. Most of the time we don’t know how our systems are working, and whether the changes we’re making will work or backfire. But we can experiment in safe-to-fail ways, and find out more about the constraints of our systems. We can develop heuristics about how to act within them. And, with some luck, we can move these systems into the more predictable and ordered domains by adjusting the constraints. This is something that humans are quite good at.”

“This is really fantastic, Raj! Great find. I think it will really help ground all the discussions that we’re having about the recent outage.”

“I also think it’ll help us further convince Roger that we need to change how we deal with outages and the people involved with them. This helps us build credibility.”

“Have you sent it to Roger yet?”

“No. But I’ll get on his schedule right now.”