Complex system

From The Jolly Contrarian
(Redirected from Complex adaptive system)
Jump to navigation Jump to search
The JC’s amateur guide to systems theory
Index: Click to expand:
Tell me more
Sign up for our newsletter — or just get in touch: for ½ a weekly 🍺 you get to consult JC. Ask about it here.

Everybody has a plan until they get punched in the mouth.

—Mike Tyson

You can’t eliminate the risk, so focus on managing it

A traditional risk manager — that is, one managing complicated systems and not complex ones[1] — will be conditioned to using control techniques to anticipate and eliminate all risk.

In a complex system this is not just hard; it is impossible. One must instead depend on local managers making spontaneous, provisional decisions in real-time to address the situation as they see it and under conditions of significant uncertainty. This is not a suitable application for chatbots: here expertise and, even more importantly, experience are essential qualities when making those decisions.[2] A complex system is not totally random — in that case, any action would be as good as any other — so some control is possible, but it is not possible to prescribe in advance what that action should be.

Therefore plan, but not with an expected outcome in mind. Plan for the unexpected. Have band-aids, a Swiss Army knife, some duct tape and a towel with you. Try to imagine how things might unfold, and watch them as they do, adapting as you go.

When a man throws a ball high in the air and catches it again, he behaves as if he had solved a set of differential equations in predicting the trajectory of the ball. He may neither know nor care what a differential equation is, but this does not affect his skill with the ball. At some subconscious level, something functionally equivalent to the mathematical calculations is going on.

Richard Dawkins, The Selfish Gene (1976).

If you have ever wondered why science nerds tend to be unco-ordinated, wonder no more.

You cannot brute-force compute a wicked problem, like catching a ball,[3] but you can still catch a ball. Don’t think, “punch all the variables into a machine and run round to the resulting co-ordinate and stick your hand out.” You don’t have nearly enough information to even make the calculation. Instead, just run towards the damn thing, watching it, adjusting as you go.[4]

Now, compare catching a ball with predicting any future event — be it the expected local weather in Lissingdown,[5] or the level of the Eurostoxx, six months from now. The further in the future the event, the poorer your snapshot prediction will be, however sophisticated your apparatus. Now, ball-catching isn’t that wicked: none of the factors at play in the fight of a cricket ball are especially suggestible, or possessed of independent moral agency, after all. Weather predicting is more wicked, and stock markets are properly, fire-in-a-crowded-theatre wicked, but the principle remains the same. The nearer you are to the event, the better your guess will be.

Calculating an exact parabola from initial conditions — even if you have good approximations of the necessary data inputs to hand, which you won’t — will give you a rough vector and distance, but the range of potential trajectories will be far too great to to ever actually catch the ball. Likewise, the prediction of rain in Lissingdown a month in advance is is highly speculative: we may have average rainfall data, but as to whether it will rain or not at a given moment, who can say?

Where the gaze heuristic helps us is by constantly, and cheaply, refining and updating that initial prediction as the event gets closer and better information becomes available. The skill lies in the experience and judgment in making those on-the-fly adjustments: that is, expertise. Likewise, with a weather forecast: my guess is as good as yours whether it will be raining 6 months from now; but I can be certain as I look out my window to an angry Lissingdown sky that it will still be raining 5 minutes from now.

This is hard for a complicated systems guy. Complicated systems you can brute force, and you can predict how they will behave. You can pre-bake solutions, making them more simple. In complex systems you can’t: need to keep your options open and be prepared to shift, adapt, re-evaluate, and toss out whatever you might have concluded before now. Philip Tetlock’s “Superforecasters” are complex systems thinkers. Baseball players are complex systems thinkers. Richard Dawkins, whom I like to imagine was dyspraxic,[6] is a complicated systems thinker.

If a complex system blows up, “complicated” risk management systems can get in the way

Frequently complicated system risk attenuators can, in fact, aggravate risk situations in complex systems. Alarms going off make it harder to hear; multiple alarms increase panic and obscure each other; an obligation to follow prescribed safety routines can impede quick and surgical response to traumatic situations. There are times, therefore, where you want to throw your checklist out the window. [7]

References

  1. Open question — a gaping open question, like when your goalie has come up for a corner — is why a traditional risk manager is managing what is undoubtedly a wicked environment using tools suitable for a tame one. But it was ever thus: Black-Scholes option pricing model, which is predicated on a normal distribution, can’t work with The tail events and whose failure in those circumstances led directly to both the LTCM collapse and the Great Financial Crisis, is still widely used today, after all.
  2. Needless to say, this is not what our management consultant friends, who advocate down-skilling and offshoring, want to hear.
  3. Ohh, but catching a ball isn’t a wicked problem! I hear you cry. For hard-determinist, reductionist types maybe, but if you have ever pondered the odd lack of tenured physics professors in the national cricket team you may, like the JC beg to differ. The JC’s celebrated experiments with the proverbial crisp packet in St Mark’s Square may help explain.
  4. A study a while back found professional baseball players, while excellent at catching moving balls they were allowed to run towards, had a lot more trouble predicting where those balls would land when made to stand still.
  5. May the lord bless and watch over Ronnie Barker.
  6. largely because he was trying to solve differential equations instead of running after the ball, of course.
  7. I know, I know — try telling that to the chap who landed his plane on the Hudson thanks to his unflappable compliance with cockpit checklists, right?