Thinking as a Hobby

Get Email Updates
Email Me

Admin Password

Remember Me

3478491 Curiosities served
Share on Facebook

Does Evolution Favor Complexity?
Previous Entry :: Next Entry

Read/Post Comments (3)

Laelaps has a great article on whether or not evolution is directed or guided toward a particular goal. His conclusion: no. I'd agree, given the definition of "directed" he is using:

The term "direction" alone is ambiguous, so for the purposes of this essay I'm going to take it to mean that direction means that a process is imbued with some sort of purpose or progresses towards an end point, even if that end point is transitory and leads to the continuation of the process in a new direction.

The article reminded me of a number of issues.

First, it touches on the issue of whether or not evolutionary theory can make testable predictions. In my first class in the doctoral program here, I had an argument with the professor about whether or not evolutionary theory makes predictions. He said that evolutionary theory was explanatory, describing fossil evidence and the current state of life on earth, but due to its random nature it cannot make observable predictions, except of the very general kind.

Also in the evolutionary dynamics class I took last year, we talked about the issue of whether or not evolution is directional. For my class project, I specifically looked at differences in speciation/extinction rates and complexification rates in two sets of evolving populations, ones that had a specific selection pressure and ones who didn't (i.e. they were chosen to survive randomly, not based on how well they did something).

I used a direct encoding, which means that one gene encodes for exactly one trait, in this case either a neuron or connection in an artificial neural network. Real genes don't work this way, but for the purposes of the model, the simplification is useful.

Basically, all individuals start out with 9 genes. There are random mutations which add and delete features from the network, with rates biased in favor of increasing features. The following graph is from a run with a random fitness function, and relatively low mutation rates:


For each generation, blue bars indicate the range of complexity for the population; black dots indicate the mean fitness for a given generation; red triangles indicate the complexity of the individual with the highest fitness.

I did 20 runs for the project, both random and with selection pressure, and the graphs tend to look very similar. The main thing to notice is that there is a fan-out dynamic to complexity (here measured as the number of genes). There is a lower bound to complexity, but simpler individuals persist reasonably well, while the maximum complexity grows. In these kinds of graphs, you often see big dips in the maximum complexity, reminiscent of extinction events, where a lot of complexity is lost in the population due to either random death or competition.

As a very simplified model, these kinds of dynamics mirror those of biological organisms. When someone who doesn't understand evolution asks "If we evolved from monkeys, how come there are still monkeys?" they could very well ask the same about any less complex organism with which we share ancestry, all the way down to single-celled organisms.

This kind of graph answers that question very nicely. If you start with simple individuals, mutations are going to increase the range of complexity over time. So you will see more and more complex individuals, but the simpler ones will continue to stick around. The evolution of complexity is a fan, not a ladder. More complex organisms on average don't replace less complex ones; they branch out alongside them.

This trend in evolutionary dynamics certainly seems like a meaningful prediction to me. We would expect to see an increase in the mean range of genotypic and phenotypic complexity, although over the short term, extinction events will temporarily decrease that range.

So in this simple sense, evolution does have a direction: an outward spread, not an upward slope. However, since that trend can be evoked in a simulated population, even without any kind of environmental pressure, it seems clear that there is no need to invoke a guiding intelligence to explain what's going on.

Read/Post Comments (3)

Previous Entry :: Next Entry

Back to Top

Powered by JournalScape © 2001-2010 All rights reserved.
All content rights reserved by the author.