The problem of causation in large systems

Like most things on the internet this week this blog post is inspired by the riots in the UK. Pundits, politicians and various plonkers have been throwing around explanations for 'why' this series of explosive riots and social disturbances happened. Some people blame the heat, others the effects of austerity policies on the social cohesion of poorer inner city areas. Some point to demography, others to unemployment, and even multiculturalism as key factor in setting the scene for an uprising. David Cameron's speech today pointed the finger clearly at the decay of modern society as he saw it.

Irresponsibility. Selfishness. Behaving as if your choices have no consequences.

Children without fathers. Schools without discipline. Reward without effort.

Crime without punishment. Rights without responsibilities. Communities without control.

The breakdown of the family and of established social structures looms large behind the Prime Minister's speech, and behind that, an older narrative about welfare-scrounging single mothers. But let that pass, it's a red herring in my opinion.

Mr Cameron pledges changes to the implementation of policy around 'the family'--whatever that is--and a range of measures designed to foster ownership of schools, communities, and the nation itself. It is all very worthy, and to a certain extent I'd applaud some of the sentiments, but as there are no concrete measureable policies, let alone evaluations, at hand, then we can do little but applaud fine words as just that. Words.

But I'm not sure the Prime Minister has the causality right. Causality is a funny thing in nonlinear systems. You can say X causes Y with some confidence in linear (or quasi linear) systems. But causal factors get difficult to discern when systems are complex, and complicated.

Like the real world.

It's very hard, even knowing the dynamics of any system, to assign causal status to one variable on another. But we don't know the dynamics of the real world, UK city-level system. I don't think we can know those dynamics in fact. Here are hundreds of potential causes. So rather than point out single mothers (or fathers) stuffing chips into unwanted children's' mouths and saying, without evidence, that they represent family breakdown coupled with poor feedback effects from society and particularly from the correctional and coercive parts of society, and that they are what's causing the moral breakdown, I'll just say I don't know what caused the riots.

In fact I'll never know. And neither will anyone else.

But in the rush to do something, to place blame somewhere and institute a commission or a national whateverthefuck or take people's social welfare benefits away (it's normally the poorer elements that get a kicking, isn't it? No mention of rioting middle class kids having their college entitlements removed or their taxes increased, or whatever), are they just blindly assigning causality in this situation and moving on?

What if those assigning that causality like Mr Cameron are wrong?

They will do more damage than they know, and they won't even know how bad they've made things. That's the nature of a complex system. It unfolds in an unpredictable way.

In 1945, the mathematician and all around genius Norbert Weiner wrote a beautiful paper, sadly pay gated, on the role of models in science. Weiner and his co-author Rosenblueth had the essential idea that our model of the world (X causes Y; Family breakdowns lead to violence on the streets) is built to simplify. Herbert Simon wrote in the 1950s on the problems of 'identifying' a model as being useful or not. The legendary theorist of causality, Judea Pearl, has a lot to say about causality.

They all say, essentially, it is very hard to assign causality, you should be really careful when trying to do so. Essentially you're trying to do science when you do that carefully enough.

But this is not science, this strange thing that is happening across the water from us. This is something else. Science tells us we need evidence to assign causality. And we don't have evidence, for all the harrowing stories of uprooted homes and businesses and endless hours of cctv footage. This isn't science or even a pretense of science. This is some weird PR generated witch hunt dressed up as social engineering within a large system.