Prof. Amir M. Sharif in Complexity in Finance discussion asked:

What does complexity mean for you?

When we talk of complex systems we normally mean systems or interacting components that have an impact and relationship between each other – and generally exhibit non-linear, unpredictable, sometimes volatile, ever-changing and emergent behaviour.

But thats the textbook – what are your views?

________

To me, a better way to view complexity is describing what we see in the real world, the physical subjects that are what scientists try to imitate with their models.  In nature and in the economy complexity is quite often observed as a property of the unexplainable organization of systems of nature that work by themselves.

It’s what nature uses somehow to do things quite simply with vast assemblies of independent parts. I think the real question is how scientists can ask better questions them and construct better imitations. Science has a curious handicap for doing that, though.

The generally accepted scientific method does not actually have a way to refer to readily named and observed individual systems except by referring to our own models. So science can’t even point to the physical “things” which models are intended to represent.

Conventional science only has a way to define “data” points and models. So, we are left trying to make models of something without having a way to directly refer to it as a physical thing with its own organization and behavior.

So, my approach to studying complexity is first to make a scientific way to locate and refer to them as things, using something they do as a whole, such as grow. Having a way to refer to the whole, I then look to divide up the whole into its functionally related parts, mapping out which things respond to and use which things.

What you find is parts that act on their own and both individually and in larger units accumulatively learning as they go. That begins to expose the very fundamental difference between models and reality.

Models are invariably definitions of fixed relationships between controlled variables. You can’t define them any other way because models can only describe deterministic relationships, and can’t have independent parts. That is utterly unlike the physical systems we see behaving in quite organized but ever changing ways.

That complex systems then need to be viewed as learning processes, of swarms of learning individuals, seems to become the key to a productive new way to study them. Then data showing change over time represents their learning curves, and is quite sensitive information about the openings and challenges the learning processes within the system are finding.

Wherever there is evidence of a reliable regularity It does let you make a useful model, but then are always aware that that regularity is temporary. You then keep an eye out for diverging patterns from the regularities you had seen before, to watch for how the complex system is changing.

pfh