Tuesday, January 15, 2008

What Part of Evolution Don't You Undertand?

We're coming up on the Darwin Bicentennial, hooray! In honor of the occasion, I'll throw out my own little explanation of evolution for the sake of anyone dazed and confused by the massive amount of misinformation wandering the Internet.

Evolution is the name of a process. It is what happens when the following are all true for some period of time.
  • There is a population of individuals, and these individuals have traits.

  • The traits influence the success of the individuals in reproducing.

  • There are limited resources available for reproduction.

  • Not everyone that wants to reproduce will get the chance.

  • Chances depend on traits.

  • The traits can vary when individuals reproduce.

If we compare the population after some long period with the initial population, the frequency of the traits will have changed. Traits that helped reproduction will have increased their frequency at the expense of alternatives.

There is nothing in this that limits evolution to biology. The whole point of a field like Genetic Algorithms is that I can set up those conditions in software and evolve data. This is a very successful part of computer science, with many commercial applications.

Of course, it does happen that evolution applies to the biology that we see around us. If you don't think it does, you'll have to point out which of the above conditions doesn't hold in the real world.


Anonymous said...

Um, heritability?

David vun Kannon said...

I assume you mean "where is heritability in your outline?" rather than, "I don't understand heritability." ;-p

I've tried to cover that with saying that traits can vary during reproduction. "Heritability" is the name of the scale for measuring to what degree is a trait heritable. Variance can push you up or down that scale.

This definition also allows for Lamarckian evolution. While we know that in the real world, the experience of the somatic cells does not feed back to the genetic information in the sex cells, we can write GA code that does allow learning to influence traits passed at reproduction.

I hope that helps explain my choice of wording. Thanks for commenting!