ThoughtStorms Wiki

Something I wrote about PersonalResponsibility and ComplexSystems here :

I have problems with the "all blame rests at the point of failure" because I think it is a misleading theory of the world. In particular it leads you to make the wrong predictions and inferences.

Maybe a medical analogy can help. Take AIDS, which systematically destroys the patient's immune system. Then, one day, the patient dies of pneumonia.

If the doctors simply say "OK, another pneumonia case" or sit round devising new strategies to treat the symptoms, they'll never make any progress actually curing the disease itself. Before you can cure AIDS you have to recognise that a problem exists beyond the immediete, apparent cause of death.

This is obvious to doctors. The body is a collection of systems, and many diseases are the result of problems in one place which cause symptoms in others.

And that's a standard feature of all complex systems.

Complex systems are non-linear : small inputs have large effects, large inputs can have small effects, positive inputs can have negative effects, and vice versa.

Let's take another analogy, in a domain that's currently well studied. Traffic congeals into a traffic jam, not because these particular drivers on this bit of road screwed up, but because at this speed and density, a slight fluctuation in their driving behaviours, (one driver puts his breaks on a touch early) triggers a phase-change in the state of the traffic as a whole. Suddenly everyone is going slower.

Now you might want to try to "blame" that driver who triggered the phase-change, but you must consider a) you can't tell who he is, b) he did it by doing something he and every other driver do, unproblematically, dozens of times an hour, just in a crucial place and time, and c) neither he, nor anyone else could have recognised that the point where he did this, was the crucial point.

I'd say that the concept of "blame" here is inappropriate. But, worse, "find the guilty person who breaked 5 seconds too early" is a totally useless strategy for trying to cure traffic jams. It won't work, because tomorrow the traffic jam won't be caused by the same person, and might not occur due to an action at the same time / place. It might even be caused by someone breaking 5 seconds too late.

Nevertheless, that doesn't mean you can't study traffic and improve its flow. But you do so by ignoring the "individual responsibility", and by looking at the overall statistical behavior, modelling alternative scenarios, modifying the context (an extra lane here, removing a junction there etc) and seeing what works.


No Backlinks