2007-01-05

What is Emergence?

A couple of weeks ago I went to the lecture "Qu'es-ce que l'émergence?" by Hugues Bersini, as a part of the Seminar series "Penser La Science", in honor of Ilya Prigogine.

"Emergence" has been a widely used and abused term, with the drawback that there is no agreed definition of it. As Jack Cohen puts it: "Emergence is like pornography. I cannot define it, but I know when I see it." Well, we can say that emergent properties are those that are not present in the components of a system, but are a product of their interactions. Phil Anderson gives the nice example of gold: its atoms do not have temperature, malleability, conductivity, colour, etc. that a piece of the metal has. Thus these properties "emerge" from the interactions between atoms.
Similarly, cells are made of non-living molecules, but their interactions generate properties that we call "living", so we can say that life is an emergent property.

What Bersini discussed in his talk was the following: is it suitable to call emergent engineered systems? His answer is no. Thus, complex structures in cellular automata (such as the Game of Life) or swarm behaviours in robots would not considered emergent. This is because the desired behaviour is defined by the designer, even if there is surprise involved. I find this a bit anthropocentric, since this would imply that the same system, e.g. a bacterium, would be considered emergent if it evolved, and not emergent if it was genetically engineered... For Bersini engineered living systems are not emergent, and I am a bit unsatisfied with that restriction.

I must say that my French is not tip top, so maybe I misunderstood something, but after the talk I spoke a bit with Bersini and the discussion clarified some things. In principle, you can call anything emergent, since anything can be seen as a system, thus emerging from its components (some may say that elementary particles cannot be divided, but the same was said about atoms and later about protons... the problem is that you need increasingly higher energies to detect smaller particles... but anyway). But the question is more "when is it useful to call a system emergent". (We reached the same conclusion with self-organization). And for Bersini, it is useful to call biological, physical, and economical systems emergent, but not engineered ones... If this is the case, I'm still not sure, but I agree that we should explore this pragmatic side of science. For self-organization, my criterium is the following: it is useful to call a system self-organizing when its state/phase space is too large to be explored completely, or it changes in time. In other words, if you have a complete description of a system, you don't need self-organization. But if it is too complex to be comprehended, or too dynamic to be determined, then you can use self-organization to find "desired solutions" within the immense/dynamic state space. Could it be the same for emergence???

Post a Comment