Simple or simplistic?

My dad is a shrink, the kind of professional who will make you talk about your childhood to try to help you solve your issues. While I was growing up, I often listened to him talking about fascinating (and disturbing) concepts such as sublimation, death drive or transitional space. For some reason, it always made a lot of sense to me, and I guess it helped me realize how incredibly complex humans and human interactions are.

I am a physicist, so I like my models as simple as possible, but not simpler. I believe that a model shouldn’t have 5 parameters if 3 will do, not only because that makes it harder to handle and explore, but also because it makes it somehow less elegant.
And that wouldn’t be a problem in my research life, if it wasn’t for the fact that I abandoned the so called orthodox path some years ago, to enter the study of complex systems, social networks and the such. So yes, I face a dilemma (pun intended) when modeling these social systems and ignoring pretty much everything that make us humans.

For example, I very much appreciate the effort (and I understand the need to simplify), but I just don’t agree with a model for human interactions based on the maximization of a utility function, nor with the restriction of human responses to a very limited set of discrete, fixed-no-matter-what strategies.
Of course, when applying this approach to a very restricted and controlled scenario, it might be enough to reproduce an observed behavior. Don’t get me wrong, that is something! But recent sociological experiments about games on networks, by Anxo Sanchez, Yamir Moreno and collaborators, contradicted pretty much all dogmas blindly followed until now, when it comes to humans having to chose their strategies. Better more accurate models are desperately needed, obviously.

Another example could be the excellent talk I attended to a few months ago, about extracting and classifying the underlying emotions from text messages. It was a commendable effort, and probably one of the finest works I have come across on the subject, but still, the two-dimensional plane used to characterize emotions was naive at best.

And on top of everything else, there is irrationality: we all know that there is an important (major, I would dare say) irrational component to all our decisions, but when not ignored completely, is usually modeled by adding random noise to the dynamics. That is probably because it is the best we can do, but I can’t help but thinking that irrational and random are not synonyms. Not even close.

Of course, we still have a long way to go until we are able to model social system as accurately as we can model magnetism (if ever), and oversimplifying is usually both necessary and inevitable, but I guess my concern is that we may be happily settling down for a simplistic model, instead of keep going towards more elaborate ones.

- Julia Poncela Casasnovas