Jerry Smith

Much has been made in both camps about Dawkin's algorithm which starts with random letters and produces by a process of random change modified by selection the phrase "methinks it is like a weasel"... The creationist objections are as follows:

1) The phrase is targeted for, but darwinian evolution evolution does not postulate a targeting process (real world: there is most definitely a targeting process, just no single targeted outcome)

2) The "transitional" phrases do not "work", but darwinian evolution relies on transitionals that are capable of survival.

In order to answer these objections, I would like to propose another, more slightly more grandiose, experiment:

In a double blind test, start with a short and meaningful phrase. Hand it off to your white-coat webmaster, along with a simple algorithm for copying the phrase and making occasional random changes.

The algorithm could specify a probability that any one letter could become another letter based on how close they are in the alphabet... 'a' becomes 'b' - .01%, 'a' becomes 'c' - .008%, 'a' becomes 'd' - .006%, etc. In addition, there would be a chance that any sequence of letters the original length of the phrase could be copied over and inserted at any point in the phrase, at (.01%*length of sequence)/length of phrase.

Ask the white-coat webmaster to run the algorithm on the inital phrase, generating 100 (nearly identical) copies of the phrase and place each one on a web-site, along with the original. 50 subjects will be assigned to view each web-site and vote on which phrase is more relevant to their own life, between the new phrase and the orgiginal. The algorithm is then run again, however the number of copies made of each phrase will be a function of the number of votes each phrase received, with no more than 100 phrases posted back to the web-sites for review.

After 7.8 blue-million iterations, the final products are examined. If we find our algorithm is at least as good as any given author of Harelquin romance novels, then we can put to rest forever the idea that random changes coupled with selective survival pressure cannot produce positive changes in overall information, with viability at each step of the process.

If we chose to, we could honor Dawkins, by making "Methinks it is like a weasel" the initial phrase.

I will be happy to undertake this project if the funding is made available. Current estimates place the cost of the project at somewhat less than the Apollo space program.

Daydreamer

Bah, tell them they're making a Straw Man. The "Me thinks it like a weasel" example was intended to show how evolution is gradiated. There are really no good analogies for all aspects of evolutionary theory short of an full blown simulation or the real thing.

My personal opinion for an experiment would be to plug the computer with a spelling/grammar checker, and run the sim. The idea would be to create a coherent sentence from a coherent sentence that was radically different in structure and content.

Jerry Smith

But then I don't get the big funding $$s...

Good point though, simpler than my half-baked idea. Could do that with Microsoft Word for that matter, using the VBA scripting...

Automaton

Yes, I think Daydreamer had the more efficient idea. Say you start with a coherent sentence, something like "Creationists are really quite stupid." Every generation, create 100 copies of those with 3 unique "mutations" each. The grammer/spelling checker determines which is the most conherent, and if there is more than one, it's just a toss of the coin. After 100 generations, it would have 300 changes from the original, and hopefully convey -some- useful information (spell checker only sees how useful individual words are, it can't determine how coherent the sentence is as a whole).

My way I think would be better, because it would have a much better chance of producing a coherent sentence. First get a large "dictionary" file of english words (which can be obtained from any decent hacking site). Plug in a reasonably large paragraph that talks of some vague thing or another, and set it to go, with each "mutation" moving up or down the wordlist, or adding or removing a word. Then we have a human checker who tests for the most coherent sentence out of the population of mutants and this continues on. This technique has worked for developing pretty images through Darwinian means, with the subject testing for aesthetic quality. The problem is creationists might claim that the subject would (unwittingly or otherwise) tend to lead the generated sentences in a particular direction, thus making it not Darwinian. This is a problem, but can be solved if you use a panel of say, five subjects. You could even let a creationist do it so that they know the evil scientists weren't using their biases (there's no guarantee however that the creationist won't try to sabotage it by selecting the least coherent sentences. )