Wednesday, 15 September 2010

Cannonical Casts

Imagine you're starting from scratch with the goal to model human behaviour scientifically.  How would you go about it?  Would one aspect of that behaviour be easier than another to model?  Would you want to build mathematical models?  Would you build models only of certain dimensions of human behaviour?  For example, economic models, political models or ethnological ones?

I'm asking this so that I can understand the move in classical economics which resulted in the more or less formal model of the optimising individual, the rational economic agent, Homo Economicus.  I want to see what drove the original economics to postulate such a model of economic behaviour, especially in light of subsequent criticisms of those models and their effects (from the merely performative to the destructive) on our world.
 These criticisms started to gain traction in the 1930s, as a consequence of the great depression, a worldwide phenomenon, and have again received attention following the more recent credit crisis.

I want to articulate the main elements of this approach, as I see them, and identify what is still good about the approach, what may change.  The best place to start is by imagining the simplest possible useful economic model.  

We will always in any model need actors who interact with each other.  My first question : how many different kinds of actor?  My second question : how constrained are the interactions?  What I'd like to achieve in thinking about these two questions is an framework which simultaneously represent classical/neoclassical (Jevons, Walras, Fisher), institutional (Veblen), Austrian and  Keynesian approaches.
The kind of framework I have in mind is the cellular automaton.  I think here we can strip down all the main kinds of economic approach as variants of cellular automata. 

This is not to say that cellular automata are the best, or even useful ways of modelling economic phenomena.  What I am saying is that thinking of all of the approaches to economics as fitting into this framework might be a useful exercise in understanding their relative strengths and weaknesses.

My contention will be that with cellular automata, there are two equally essential structural parts - one defining the repertoire of behaviours of the actors, the other defining the range and complexity of signals which can be emitted and received by these actors.  All forms of economic modelling can be re-framed as kinds of cellular automata, with varying kinds of actor, varying kinds of signal.  The signalling architecture of these models corresponds to a kind of institutional constraint.  Jevons and Walras were right to begin their analysis with the simplest kind of model of economic behaviour - one where the only institution in place was a competitive market, processing and re-distributing price signals to all participants instantaneously and evenly.  Those very participants are following the same rulebook - they're a family of mathematically identical little John Stuart Mills - brought up on a strict 
regime of Benthamite utilitarianism. 

However, economics is performative, and this initial model, with its most basic institutional element, together with a doctrinaire utilitarianism, began to have efefcts in the real world.  This kind of performativity can act as a drag on the speed with which new economic world views can be synthesised.  But before I take this investigation any further, I want to read some more on the intellectual transformations which take us from what is essentially the socialist agenda of Auguste Comte and J.S. Mill to the more familiar model of the optimising, self-interested economic agent.