Thursday, March 13, 2008

How to tackle complexity


Stuttering is a very complex disorder, and I want to talk about strategies to control this complexity in order to understand the underlying dynamics. They are mostly inspired from methods or tricks that I learned from fellow physicists.

- Always remember that every phenomenon no matter how complex is ultimately an interaction of atoms! There are no ghosts or paranormal forces or magic at work. There are better explanations for why something happens, but sometimes it is damn difficult to achieve understanding. In some cases, it might be (nearly) impossible, like in fluid dynamics, climate models, social interactions, but nowadays computer simulations are very helpful in reproducing complex phenomena whose internal dynamics we do not understand theoretically.

- Do not talk to people who advocate a "holistic approach" or "to view every person as an individual" consciously or unconsciously acknowledging defeat to complexity. They are like the ancient tribes assigning a God for every phenomena they do not understand, or the 18/19th century scientists inventing concepts like elan vital as the fluid that makes matter come to life, like the calorific fluid that transfers heat, or the ether to give classical physics alive. They are only stating the obvious that we already know, namely that stuttering is a very complex disorder. Very rarely, do they have to say anything deep. We must strive for better explanations and not give up.

- Construct theories up to a first order approximation first. For example, we conveniently talk about males and females, but there are some people who are different physically or genetically. Or, let's take planetary orbits. At a first order approximation, they are circles, at second order they are ellipses, at third order they are ellipses with some correction due to the other planets, and at fourth order, these corrected ellipses have corrections from general relativistic effects.

- Cut out sources of complexity that are not fundamental to the disorder. Here, I would put psychological and social consequences of stuttering in this group. I am purposefully ignoring them, not because I think that they are not important in other areas like therapy but because they add considerable complexity without adding much insight into the underlying causes. A clearer example is hair loss. There is a wide range of complex reactions in males and females and their environment, but they tell us nothing about hair loss itself.

- Throw out any atypical cases. In stuttering, we need to "get rid" off people who started stuttering after age 7, who have undetectable dysfluencies, who have other disorders, and even who are females in the case of brain scans. This relates to the first order approximation in that we need to first try to understand the typical cases. Clearly, this atypical group might well be a source of inspiration and explain the typical behaviour better.

- Look at every possible dimension you can think of independently. Here are a few: brain imaging, self-reports, genetics, therapy, phenomenology, linguistics, child development, and so on. But, it is clear that a full picture only arises once you look at all dimensions at the same time and look for the interactions between the dimensions. Nevertheless, you should first understand each dimension independently.

- Concentrate on more quantitative dimensions where you can get clearer and more OBJECTIVE results. Genetics is the king here, because the DNA is well-defined. For example, we know due to twin studies that genes have a big influence, but how it happens we have no clue. Still, we can identify genes and bypass the enormous complexity.

No comments: