lpetrich

Creationists often claim that "information" cannot be created by natural processes, though they generally do not bother to describe what they mean by "information".

Maybe it's something like "created kinds", something that they cannot really define without making their case look weak.

RufusAtticus

Creationists won't define information or a way to quantify it, because once they do they won't have anymore wiggle room to say that it applies to biology but not to evolution. In reality, most creationists argue aginst increacising the capicity to hold information rather than information itself.

Ernst Mayr in What Evolution Is

I've been debating about information in biology over at Christian Forums, and I think I might write something about. (Maybe get a dissretation out of it if I can find something original.)

~~RvFvS~~

Quetzal

It's actually kind of a silly debate - unless you've got some creationist with at least a smattering of genetics, they'll never even understand why they're wrong.

You’re right about one thing though: the creationist arguments about information theory are totally bogus, just like their immutable “kinds” arguments. Most of the ones I’ve seen argue this are merely regurgitating creationist websites and authors, and have absolutely no clue what they’re talking about. Getting them to define information is impossible – any more than it’s possible to get them to define kind. Bottom line: information theory OF COURSE holds no problem for evolution, since all “information” is is a metaphor. Naturally, along with every other aspect of science, creationists have no compunction about misrepresenting and/or misusing science to gull the ignorant. That being the case, you need to present a counter – the actual science – to gain any ground. I seem to have a special place in my heart (or some would say a soft spot in my brain ) for information arguments.

My counter goes something like this:

Creationists LOVE using words like “code”, etc, as though there were some intrinsic meaning or existence to the terminology. I guess since a lot of people have computers, and most people know that computers are “programmed” using “code”, it’s easily assimilated. The fallacy is trying to claim this terminology has some fundamental reality. “Coding” is a convenient way for humans to visualize the deterministic chemistry involved in DNA replication and transcription, because there are certain semantic similarities between computer coding, for instance, and genetics. There is no more “code”, “information”, or “instruction” involved than there is in 2H2 + O2 - 2H2O yields water, or any other chemical reaction. All of the terms are abstractions. Creationists try to state that an abstraction has an intrinsic physical reality.

"Code," and "information" are metaphors, terms used by science and technology and given special content for special purposes. Saying that DNA contains information about amino acid sequences is simply to say that a knowledge of the DNA sequence is sufficient to provide knowledge of the amino acid sequence. DNA is not self-replicating any more than a letter put into a photocopier is self-replicating. DNA sequence does not specify protein, but only the amino acid sequence. The protein is one of a number of minimum free-energy foldings of the same amino acid chain, and the mechanisms and chemistry of the cell together with the translation process influences which of these foldings occurs –(see, for example, spliceosomes). Finally, organisms are not determined by their DNA but by an interaction of genes and the environment.

A nucleotide is not a “code” (in fact, if you want to use an abstract analogy, a nucleotide would be a “datum” at best). A nucleotide consists of a sugar and either a purine or pyrimidine base. Nothing more. It is not information. DNA does not “decide” anything. Adenine will ALWAYS pair with thymine. Cytosine will ALWAYS pair with guanine. It is a fundamental property of the chemical composition of the molecules and the way molecules combine – there is literally no way they couldn’t. The sequence of nucleotides in a functional gene simply dictates which amino acid it synthesizes – again based solely upon the chemical composition and sequence of the nucleotide triplets, again solely because of the rules governing how molecules combine. The entire biochemistry of the cell is wholly dependent on chemical triggers either within the cell or from the rest of the interlocking tissues of the given organism.

Once we get that bit out of the way, since creationists won’t define information as they’re using it, I generally argue “information” from the standpoint of the molecular biologists that use information theory in their work, specifically Shannon-Weaver Information. Here’s one example of how Shannon Theory can be an extremely useful tool in genetics: <a href="http://www.lecb.ncifcrf.gov/~toms/paper/schneider1986/latex/paper.html" target="_blank">The Information Content of Binding Sites on Nucleotide Sequences</a>. Here’s another paper from Tom Schneider: <a href="http://www.lecb.ncifcrf.gov/~toms/paper/nano2/latex/index.html" target="_blank">Sequence Logos, Machine/Channel Capacity, Maxwell’s Demon, and Molecular Computers</a> which provides a good overview of information theory as it relates to molecular biology. These are legitimate uses of information theory.

For example, in order to measure the Shannon information content of the process of transcription/translation, you cannot simply take the DNA strand or the readable portion of the strand out of context. The DNA exon doesn’t even constitute the entire information content of the process. If it did, information would axiomatically be “lost” during the process through Shannon entropy. Since DNA does get replicated – usually pretty well, in fact - obviously something else is going on. In this case, the other elements in the cell (from various types of RNA to enzymes to RNA polymerase to the various transcription factors) constitute additional information needed for transcription/translation. If you want, you can add all this “information” content to the system before replication, transcription/translation starts – in reality the maximum information potential of the system doesn’t change. However, doing so has significant (negative) implications for any argument that mutations don’t add information. Under Shannon, ANY change to the system by definition “adds” information. In addition, this leaves open the possibility of adding information by increasing the string length – through gene duplication, insertion of new bases, cross-over, etc.

In addition, since there are sequences that designate what is an exon in the gene that cellular machinery must recognize, mutation can cause the machinery to "read" the DNA in different ways or alter the mRNA transcript. For example, mutation might cause the machinery to read an intron (non-coding DNA) as an exon. The intron, when incorporated into the final protein, can alter the protein structure and either hinder, help, or even create a new function for that protein. Further, the machinery might skip an exon that was previously read before the mutation, or alter the transcriptional rates of exon and intron reading - all these would alter the final product of the protein in a major way – and hence change the information content!

Dawkins, as you can imagine, has a good article, <a href="http://www.skeptics.com.au/journal/dawkins1.htm" target="_blank">The Information Challenge</a>. For your creationist friends who insist that information can’t arise spontaneously – besides getting them to define what they mean by information – refer them to these papers: <a href="http://www.mdpi.org/entropy/papers/e3040273.pdf" target="_blank">Self-Organization of Template-Replicating Polymers and the Spontaneous Rise of Genetic Information</a>, then read <a href="http://www.pnas.org/cgi/content/full/97/9/4463" target="_blank">Evolution of Biological Complexity</a>, and finally <a href="http://www.lecb.ncifcrf.gov/~toms/paper/ev/latex/node2.html" target="_blank">Evolution of Biological Information</a>.

Hope this helps.

[Edited to fix URL.]

[ July 01, 2002: Message edited by: Morpho ]</p>

lpetrich

I have had a fair amount of experience with data-compression algorithms, and that has given me some experience with the technical meaning of "information", as opposed to the vernacular meaning of "information" as simply description.

Imagine a system that can exist in any of several states i, each one having probability p[i] However, these states can be distinguished by some appropriate sequence of yes-no questions; how many such questions need to be asked, on average, to distinguish different states of the system. The more yes-no questions needed, the more information necessary, in the technical sense of the term. The answers to these questions may be mapped onto the numbers 1 and 0, the two digits of a binary (base-2) representation of a number, thus the common information-content measure "bits".

In 1948, Claude Shannon derived a famous formula for finding the minimum number of such questions for some probability distribution:

I = - (Sum over i of p[i]*log(2,p[i]))

An algorithm that gives the arrangement of yes-no questions with the fewest questions on average for some probability distribution was invented in 1952 by David Huffman. It reaches the limit for power-of-2 probabilities; for non-power-of-2 probabilities, it will do somewhat worse than the theoretical limit, but usually much less than 1 bit worse.

Now for some examples. Imagine that you and your friends like various pets, which I'll reduce to 4 kinds for simplicity: dogs, cats, rats, and hamsters. Now let's see how many questions you have to ask your friends to determine what is their favorite kind of pet, on average.

If they like all four equally, then the theoretical limit is 2 questions, and there are three possible arrangements of questions necessary to distinguish these pets. One arrangement is:

Does it like meat?
-- Yes, does it bark?
----- Yes, it's a dog
----- No, it's a cat
-- No, does it have a tail?
----- Yes, it's a rat
----- No, it's a hamster

This arrangement will give 2 questions per pet no matter what your friends' tastes in pets are. But if you discover that your friends much prefer dogs, and have a lesser preference for cats, would it be possible to ask fewer questions?

Yes! And if 50% of your friends like dogs, 30% like cats, and 10% each like rats and hamsters, the Huffman algorithm tells us that the optimum arrangement of questions is

Does it bark?
-- Yes, it's a dog
-- No, does it meow?
---- Yes, it's a cat
---- No, does it have a tail?
------ Yes, it's a rat
------ No, it's a hamster

On average, you have to ask 1.7 questions, with only one needed for dogs, but three needed for the rodents. The Shannon limit is approx. 1.6855 questions.

But if you try this arrangement on some friends who like hamsters, and have a lesser liking for rats, but who have a serious dislike of dogs, you will have to ask more questions on average. For 5% of your friends liking dogs, 10% liking cats, 25% liking rats and 60% liking hamsters, we find an average of 2.8 questions.

The Huffman tree for this case is

Does it have a tail?
-- Yes, does it like meat?
----- Yes, does it bark?
------- Yes, it's a dog
------- No, it's a cat
---- No, it's a rat
-- No, it's a hamster

And not surprisingly, it has the inverse sort of performance, 1.6 vs. 2.85 questions, with the Shannon limit being approx. 1.5332 questions.

One can get beyond the Huffman limit for a single state by coding more than one state together; a two-at-a-time example using dogs and cats for those who prefer dogs to cats is:

Do they both bark?
-- Yes, they are both dogs
-- No, does the first one bark?
---- Yes, the first one is a dog and the second one a cat
---- No, does the second one bark?
------ Yes, the first one is a cat and the second one is a dog
------ No, they are both cats

But one can never get past the Shannon limit; it would be like 2 + 2 = 5.

Thus, the information content of a system is very context-dependent, which makes creationists' comments about "information" very childish.

[ July 01, 2002: Message edited by: lpetrich ]</p>