Field of Science

What's so great about modeling?

Not my kind of model, really.
Warning! For those who thought this was about fashion: it ain't.

I have a model of speciation that I have been working with for a while now, and I think it's good. And it's a model. With this model I think we can learn something real about nature, and I would even go as far as saying that it's a better model in some respects than previous models. It includes certain aspects of the problem that simpler models didn't take into account, and it is simpler than some other models that are great, too, but they're sometimes so complex that it isn't easy to discern what goes on inside them. In fact, this model originated in an attempt to mathematically describe what goes on in Avida, which is an artificial life model used to study evolution.

Why do we make models of things? Why do some scientists spend time making and analyzing mathematical and numerical models of systems and processes in nature? Can't they just use the natural systems and look at the processes out in nature? After all, if we're trying to learn something about nature, that seems to most direct route, rather than building a model on paper or on a computer and work with that instead.

I am interested in evolutionary processes, such as they take place in nature. That means, for starters, that the models I build have to contain elements similar to those found in nature. As an example, inheritance has to be done in a way that resembles how real living organisms inherit genotypic information from their parents (or parent, in case of asexual organisms). That means I can't go around having little critters in a computer inherit what their parents learned in their lives, because that's not the way genetic material is inherited out in nature (Lamarck thought so, but we now know better). But one could argue that if we build a really, really good model, then it wouldn't be different from the thing we're modeling, and then we might as well just study that thing directly in nature. So the model has to be somewhere in between - simpler than reality, but still realistic. And the advantage of making such a model is that
  • we can do controlled experiments to study isolated processes
  • we can repeat the exact same processes as many times as we want
We only have one sun, one biosphere, one mexican gulf with crude oil pumping out at the rate in excess of 5,600 cubic meters per day, and one species of California Red-Legged Frog (Rana draytonii), and we can't afford to tamper with them, even if we are able. With numerical models we can, and processes that in nature take a very long time can be modeled on much shorter time-scales, provided the models are simple enough (sometimes they get so detailed that they take longer than the real thing). Additionally, we can hold parts of the model constant, so that we can study other variables in separation. For example, the evolution of a population can be studied without the interference of other populations and environmental effects - which is usually impossible an/or unethical with living organisms.

But more to the point, modeling is how we understand things. Theories, such as the one about evolution, can be viewed as a model. The whole obnoxious, sidetracked debate with creationists about the theory of evolution being a just a theory, and have not yet been proven and elevated to a law, and all that crap is really quite misplaced. If we instead introduced the notion of a model, and that models cannot ever be proven, in part because they are just models, and models are never 100% accurate, or they would be identical to the system that it attempts to describe, then I think perhaps some people (other than committed creationists, of course) would get it, and realize that the evolution is just a theory argument is vacuous.

In April I gave a talk at this great Early Career Scientists Symposium at University of Michigan. Spot Wally seven times (not counting reflections)! Rich Lenski have the plenary presentation on aspects of his lab's long-term evolution experiment with E. coli, and because I am occasionally asked and frequently think about the issue of choice of numerical models to study evolution, and whether we can expect to learn anything about organic evolution from them, I asked Rich how he feels taking discoveries about evolutionary processes in E. coli and inferring how evolution proceeds in other organisms. The problem is that E. coli, like any particular numerical model, could be highly specific (as in, evolution being "just one damn thing after another"), and lessons learned from it not necessarily transferable to other organisms. It's not that I don't think it's a good idea to study E. coli, or any other particular organism, but that the people who do need to be careful about their conclusions about evolution in general in the same way as people using numerical models must. Unicellular, asexual microbes may not be the best model of how evolution proceeds in mammals, for example. Rich's answer was that he agreed, and that... (shit, I can't remember the rest now - so much for jokes without punchlines).

And to make the missing-punchline punchline even worse, I have to admit that I was going to write a fair bit more about the model I am working on, but truth be told I really can't make myself do it. See, I am working on finishing the manuscript that will be submitted to some journal, and I don't have the energy to also blog about it. I promise to blog about my own paper once it's out in some form, of course, though that might take a very long time. Sorry.


  1. 'ts alright. We'll still be here. :)


  2. Yeah right! How do you prove a mechanism to be effective? You show that it works...

    BTW. the argument that you have to model to be able to generalize is a very good one, we incorporate that in the next grant!

    Cheers Arend (proud coauthor)


Markup Key:
- <b>bold</b> = bold
- <i>italic</i> = italic
- <a href="">FoS</a> = FoS