Sunday 24 February 2013

Complexity Misconceptions

I have just watched this TED talk:


It's about a work by the speaker, James Glattfelder, and collaborators which was published in PLOS ONE. I have not read the work, but it seems a fair study of economic networks with interesting conclusions. That's not exactly what I want to criticise. I would like to make some observations about the talk itself.

Glattfelder starts the talk by criticising the fact that millions of dollars are invested to understand the fundamental workings of the universe and still we know very little about human interactions, specially, of course, economic networks.

I've heard that criticism many times before and from many friends. It comes more out of some resentment than from a more objective point of view. The resentment is understandable as many physicists would argue that the study of complex networks is not really physics. In my opinion, that is just a word game that divides everyone and where everyone loses.

About the fact that we understand the "fabric of reality" better than human relations, as the speaker says, is actually not that surprising. It's naturally easier to isolate systems at the fundamental level and make experiments with them than doing the same with biological or social entities. They are way more complex and, therefore, way more difficult to understand. I don't see any reason to be surprised by that.

The second thing I would like to point out is the fact that he says that the "usual" physics approach fails for complex systems. That in physics we use equations, but in complex systems we need to use networks. To be very honest, I really cannot understand what he is talking about. I worked with networks also and equations are used as much there as in any other area of physics. By the way, you can see a picture full of equations in his talk. You cannot even say that physics doesn't use networks, or started to use them only recently, because the prototypical problems in statistical physics and condensed matter are all studied in networks.

I think there is a huge misconception with respect to complexity science that the speaker is spreading here. He's selling complexity science as something completely different from what has been done in science since ever, which is not true. Complexity science is more like a culmination of bringing together related techniques from many different areas in an encompassing framework, a highly interdisciplinary one, but it is still the same kind of scientific idea. Complexity science is still based on extracting equations from data. Of course it involves algorithms due to their complexity, but algorithms are composed by equations. 

The last thing he says as if this was a very recent discovery is that complexity can emerge from simple rules. I must, once again, remind everyone that this is known since the 19th century, although these were not the exact terms used at that time. In fact, Boltzmann was the one who first tried to do something like that. He knew that collective phenomena like phase transitions (change of states in matter, like melting or evaporation) should necessarily be the result of atoms and molecules interacting by the very simple rules of Newtonian Mechanics (that was before Quantum Mechanics, of course, but the idea remains basically the same). In fact, you can browse a more than one hundred years literature in Statistical Mechanics and find out many of the ideas that are attributed to modern complexity science.

In summary, I think that it is an interesting work, but that the beginning of the talk is a bit misleading. I do not blame a researcher by being excited about his field, but people tend to exaggerate many times. It's good to be always a bit skeptical.

No comments:

Post a Comment