Thursday, July 30, 2009

Le prisonnier, le gardien et la visiteuse

"Chaque mercredi la demoiselle parfumée me donne un billet de cent couronnes pour que je la laisse seule avec le prisonnier. Et le jeudi, les cent couronnes s'est sont déjà allées en bière. Quand l'heure de la visite est passée, la demoiselle sort avec sur ses vêtements élégants la puanteur de la prison; le détenu retourne dans sa cellule avec sur ses vêtements de galérien le parfum de la demoiselle. Moi, il me reste l'odeur de la bière. La vie n'est qu'un échange d'odeurs."
Italo Calvino, Si par une nuit d'hiver un voyageur, Penchée au bord de la côte escarpée.

Monday, July 27, 2009

"Biology Is Just a Dance" by Brian Goodwin

Here are some interesting extracts from an on-line article by Brian Goodwin published in Edge who past away recently:

"Will biology join up with physics, take on its flavor, have this notion of rules, organization, regularity, order? The new movement is transforming biology from a historical science, which is what it is at the moment, the objective of Darwinism being to reconstruct the history of life on Earth. Well, that's not the style of physics. Physics is about laws, the principles of organization of matter. We're doing the same thing in biology; we're looking for the principles of organization, the dynamics of the living process. Once that's understood, you're in a position to say, "Ah! History followed such and such a course in expressing and revealing the subtle order in this particular type of organization of matter we call the living state." Thus, the first thing is to understand the living state."

"The small-scale variation and the detailed adaptation of organisms to their habitats are very well explained by neo- Darwinism, but the global problem, the large-scale evolutionary problem, is unsolved. How do you get evolutionary novelty? Emergent order? The difference between squids and fishes and penguins. That's what the science of complexity is beginning to address — to demonstrate how emergent qualities can develop out of complexity, so that you get the emergence of order. The difficulty is making the theoretical work connect with the biological evidence. Most of the modeling currently done on computers is still very abstract, and there's not a lot of detailed evidence as to how that translates into what actually goes on in organisms."

So, what are the main tangible results so far?

You can read the entire article on the Edge website.

Monday, July 6, 2009

Humble models

David Orrell got his PhD from the University of Oxford on the modelling of nonlinear systems. Although he got it only in 2000, he earned some authority and describes us in his book, The future of everything, his point of view of the failure of present models to predict correctly anything, from the weather to the economy.

His main argument is composed of two points. First, he notices that natural systems are like some theoretical systems that are called automata systems: they are systems based on a set of local interacting rules. Among the three classes of automata systems, one is composed of uncomputable systems: there is no way to speed up the calculation and the only way to know the future of the systems is to run the model.

The problem, which is his second point, is that we are not and may never use the right set of rules. All known models have some kind of parameterization of the processes that are not modelled -because we do not model from the atom to a society. The additional difficulty is that models of natural systems are like natural systems, full of feedbacks, which make them highly sensitive to any parameterization. They don't even have to be chaotic to be completely wrong:

"By varying a handful of parameters within apparently reasonable bounds, we can get a single climate model to give radically different answers"
David Orrell, The future of everything, Chapter 8.

And thus, we might never be able to predict the future as Laplace dreamed of:
"Lack of predictability is a deep property of life. Any organism that is too predictable in its behaviour will die. And in an unpredictable environment, the ability to act creatively, while maintaining a kind of dynamic internal order, is a prerequisite. The balance of positive and negative feedback loops, when combined with the computational irreducibility of life processes, makes the behaviour of complex life forms impossible to accurately model. The problem is not that such organisms are erratic, but that they combine creativity with control. House plants are quite stable (they tend to stay in their pots and don't suddenly walk off to join the forest), but it would still be impossible to predict the exact effect of moving a single plant from a shaded spot to a warm greenhouse, based only on a detailed understanding of its biochemistry. If we can't do it for a plant, we can't do it for a planet. Life, it seems, evolves toward rich, complex structures, which defy simplistic analysis."
David Orrell, The future of everything, Chapter 8.

Thus, should we even bother to try to predict? The answer is yes because although the models are wrong, they are one way to try to predict the future. What the authors try to put a term is on the confidence, and at times arrogance, of modellers. They should be the first to recognize that their models are not perfect and, on top of it, are not that objective at all -the models are full of assumptions that are, after careful look, just a set of subjective views of the world hidden behind technical terms. Thus, the author would like some kind of balance: between the objective ways to predict the future and the subjective ones:
"Objectivity and subjectivity must be in balance, and inform each other, just like the positive and negative feedbacks loops that characterize living systems. We will choose to protect nature only if we value it -and not just as an object, but because it is alive. The only way we will respect it is if we understand that we cannot control it.
In non-linear, complex systems, change often happens abruptly, like water turning to ice. Extreme change is normal. This makes prediction difficult, but it also holds out tremendous hope, because it means that a sudden change in course can be expected. Such change often comes from the bottom up, rather than the top down [...]. Unlike deterministic mechanical systems, we have a choice; we can determine our own destiny. We are not slaves to the initial condition, our genes, or the efficient market. We are unpredictable, and that's not a bad thing.
The science of complexity will not build a better GCM [General Circulation Model], and neither Gaia theory or earth system science. Their stories are more of humility than of human ingenuity. But if we as a species are standing at a precipice, it is better that we see the world feelingly than be completely blinded by our mental models; that we know what we do not know. Creativity often emerges from a state of uncertainty. Grasping for illusory knowledge by over-modelling our environment is therefore part of the problem.
Mathematical models will always be indispensable. Like language, they are a way to understand the world, and organize and communicate our thoughts. They help us perform hypothetical experiments, explore possible scenarios, and expose fragilities. Most of all, they help us comprehend what is happening now."
David Orrell, The future of everything, Chapter 8, italics are mine.

Thus, modellers, keep doing the good work. But please, drop the certainty and try to be more humble.