OK, so by now a number of you are either quite puzzled or are up in arms about this notion of mine that genes aren't information. First I'll recap and then make some general philosophical and historical points.
I argue that something can only be said to be information bearing if it is isomorphic to one formalisation of what counts as an Information Processing System (IPS). Anything that doesn't have the requisite mapping between these formal descriptions and physical systems doesn't count. I further argue that genes do not show the right properties of these formal systems, and hence should not be called IPSs.
But what's the harm? Surely, as many have noted, information talk is actually heuristically useful, as it suggests ways to think that we might not have otherwise done. And really, how much damage does it cause to talk about information in biology when it's not "really" there?
So I need to give some reasons. From least important to most important, they are:
1. Anti-evolutionists use "no new information" and a host of spurious mathematical arguments to "prove" that new information cannot be caused by ordinary physical processes like natural selection. By buying into the information talk (the way, for example, Dawkins does), one grants them illicit premises.
2. Information talk privileges genes over other causal actors in biology (which is the point of the Griffiths and Gray "Parity thesis" counter-argument). But genes are only genes in the context of complex developmental systems - in a test tube, DNA sits there and slowly denatures. So by using information talk we allow causally significant parts of the system to fade into the background.
3. Information talk biases some functions over others in a way that obscures the multiple complicity of genes in various processes. For example, the Gene Ontology project, which lists genes with their functions, privileges the first function identified thereafter, because that it its "information". This tends to restrict further investigation (where, for example, knockout studies that treat genes just as factors that are causally downstream from phenotypes, doesn't).
4. Information has become a metaphysically magical property, which confuses us no end. It is this that I want to discuss today.
Information is not a physical cause of anything. Take my computer - it "reads" information from keystrokes, right? Well, only in an abstracted manner. What physically happens is that I complete a circuit when I hit a key, which causes electrons to be transported to processing circuitry, which converts that signal into a particular sequence of high and low voltages, which... and so on. The abstraction "information" does nothing here. The physical processes that we abstract out as information do.
This is a rather rigid and unforgiving physicalism, I know. I am taking Quine's Desert Ontology to the extreme (but then, I'm not a realist, but, as my friend Stefan keeps reminding me incredulously, I'm an instrumentalist). The only ontological categories I feel the need to accept here are those of hard physics (I'll do a post sometime soon on the Pessimistic Induction). So in the light of this Spartan, or is it Calvinist, metaphysics, what role does information play? I argue, none at all.
Information, as I argue, is an abstraction of patterns. In the older metaphysics still influential on Leibniz and through him Russell, explanatory categories included form and substance, a view taken of course from Aristotle. Substance is the undifferentiated mass on which form is imposed by something. Substance has no particular causal powers. Informed substance does. In Aristotelian metaphysics, form was causally efficacious; in fact the so-called "four causes" (aitia, or accounts) included material (about substance), efficient (about force), final (about ends or purposes) and formal. And I would argue that in the Aristotelian tradition, if not for Ari himself, formal causes were the most significant.
This view has a name - hylomorphism, from the Greek for substance (hulé) and form (morphé). Think of, as Aristotle did, substance as clay, and form as the shape the clay is molded to. Hylomorphism was the foundation for western thinking and science right up until the 17th century. What happened then? Well, an old opponent of Aristotle's came back to life: atomism.
Under atomism, the properties of objects came out of the nature and arrangement of atoms (a-tomos, not-divided). Form was at best the arrangement of these particle. Democritus and Epicurus had argued for these views, but because they seemed unable to deal with the physical phenomena as well as the hylomorphic view, and because atomism was thought to be irreligious (a slander reinforced by theologians of the Christian era), it languished until Dalton.
We can skip the next three centuries to now - in these days of particle and quantum physics, do we need the notion of form+substance any more? Certainly not for molecular biology, at any rate - let's leave those fundamental physics guys to do their thing for a century or so. Morphology no longer plays a causal role in biology except as the arrangement of particles and ensembles of particles. Want to explain why protein A cleaves to nucleotide X? Do so in terms of the shape if A and X, but do not stop there - form is the effect of the properties of the parts; the strong and weak bonds, the medium, the ambient energy level, and so on. We might be able to infer similar properties from similar forms, but the full explanation, in biology at any rate, has to include the makeup of the molecules and their degrees of freedom in folding and bonding properties.
So I think that if we are to take modern science seriously in our ontology of biology, we have to treat information as an abstraction made for our benefit, rather than as an inherent property of the objects themselves. This means of course, that my claim to treat as IPSs only those system that are isomorphic to a formal IPS resolves down to what we intend to take as satisfying those conditions. But as an instrumentalist, I am happy with that. Human brains process information. Genes don't. Somewhere in the middle is a border case in which I cannot say for sure, but my instrumentalist rule of thumb might be - can we use it as an IPS, even just theoretically? If so, it's an IPS.
Mathematical Platonists will of course disagree. I am applying a metaphysics they would reject outright. That's fine - it pays to know, however, what we are committed to when we adopt particular metaphysical presumptions. I think we can do without information the same way we do without teleology. That is, we can drop both the final and the formal causes in modern science, and modify the material to become the physical, and the efficient to become the energetic. Numbers, information, purposes, and other abstractions don't do any work. Except in our thinking, which is itself an abstraction.
- Log in to post comments
Cool article. Way over my head in parts, but I think I got the gist.
The platonic forms seem beloved by many. I guess because maths gives some objective truth (2+2=4) that it must be so tempting to say that it is real. There must a 2 and a 4 out there with a perfect triangle having a form party. I like to think of it all as abstractions in our mind. I think saying mathematics is real in the sense of substance is reification of some kind. If we say 2 pigs + 2 pigs = 4 pigs, then there is information about piggies, which is real. But 2 + 2 = 4 is abstract, not an entity.
Anyhoo. I'm not a mathematician nor a philosopher.
1 seems like a silly reason. Cranks "use" calculus to "disprove" special relativity all the time, but that's no reason to abandon Calculus. It's a reason to make sure you correctly apply the idea.
I like 2, because I think it points out that information has a lot to do with time scales. Information in physical systems is always associated with a timescale of degredation. Even the information in your digital computer.
3 seems like a miopic definition of information. Information can be talked about in a multi-broadcast or multi-reciever (or both) (Sepian-Wolf settings for example.)
4 is way over my head, but I certainly agree that information isn't a "thing." I do believe, however, that information carriers are physical things and thus the laws of physics have a lot to say about information. For example, I might argue that physics of interacting systems is really the right arena to define what is and is not a system which carries information.
When I said that the reasons were listed from least important, I forgot to mention how very low was number 1 :-)
As to 3, the problem is not the term or notion "information" but the systematic way it has misled some important aspects of biology. Multicasting, ghost channels, and the like are irrelevant to that objection.
As to your comment about 4, while all information bearers must (on my ontology) be physical, not all physical things must be information bearers. So at best we can, for instance, talk about a particular system in terms of the thermodynamic properties of a bit, we cannot talk about all bits in terms of thermodynamics. At best we can say that to acquire or lose a bit of information there is an energetic effect, but not what they must be per bit.
"Mathematical Platonists will of course disagree." A bit hasty there in that conclusion I feel. I consider my philosophy to be a form of Mathematical non-dual neo-platonic idealism (somewhere between Rudy Rucker and Ken Wilber) and I am in full agreement with you. At best talk of information and the use of computer science paradigms in biological processes like genetics or consciousness has seemed naive at best to me. Admittedly I know more about CS, math and philosophy than I know about biochemistry and genetics, so my opinion is just a hunch. lol
Why does "hypostatization" run through my cerebral IPS? Or do I mean anti-hypostatization? Hm. :)
There is really only one thing that matters in applications of information theoretic concepts to biology: do they yield robust predictive models for some biological phenomenon. If they don't, they can be safely discarded. If they do, then IMO it's scientifically dangerous to look for some justification for dismissing them a priori.
There is a lot of handwringing about information being a "projection" or merely a concept that doesn't exist unless applied. Perhaps this is true under some Berkelyan idealist conception, but the idea implies that the last 50 years of research in computer science hasn't taken place. The main accomplishment of the field is that it has shown that computation has objective and invariant properties that can be rigorously studied. This happens in a theoretical framework, of course, but so do all well developed sciences.
Mathematical constructivists and even semi-constructivists, like myself, would however wholeheartedly agree.
I completely agree with Tyler here, and I think all of the stated reasons for rejecting the idea of genes conveying information are poor.
#1: As Dave pointed out, misuse of a concept or method by the malicious or ignorant is not a valid reason for discarding it.
#2: You could say exactly the same thing about a hard drive. All this "information" talk with regard to what's stored on a hard drive means nothing without the context of the entire computational system, including the external environment. And hard drives that just sit in a desk drawer will tend to break down over time. Does this mean that it's useless to talk about what's stored on a hard drive as "information"?
#3: You seem to be saying that because the effects of genes are often either indirect or distributed that this is somehow an argument against them carrying information. Again, the analogy breaks down if you're talking about your average desktop PC. The way information is stored and retrieved is often either causally downstream or distributed and non-localized.
#4: This has echoes of #1. Because it is a complex abstract concept that is often poorly defined, misused, and misunderstood, it shouldn't be applied to genes.
Why stop with genes? As far as I can tell, all the arguments you list could just as easily be applied to computers, which you are considering formal implementations of IPSs, right?
I can't speak for John, but what I would say is that it's a heuristic metaphor in this context just as it is in biology. John explicitly disclaimed an intent, and I certainly have no such intent, to say that such metaphorical usage can't be useful- of course it can. But in some contexts it's important to remember that that's what it is, on pain of sometines serious confusion. No doubt computer science is one of the contexts in which the potential for confusion is the lowest (maybe zero), since computer systems were deliberately designed by us to have operations that we can interpret as information processing. Such is not the case with biological systems.
I've often wondered about the "information" thing too, that it's been granted ontological status it doesn't deserve.
If I write "Fox" on a blackboard, does it contain information? It seems that a necessary condition for the word to contain information is someone or something that can percieve it as having information. Otherwise it is nothing more than chalk on slate.
Or perhaps I'm confusing meaning with information. Can there be information without meaning?
"Or perhaps I'm confusing meaning with information. Can there be information without meaning?"
Yes. Information in the technical sense is different than the colloquial sense. Shannon theory explicitly ignores meaning, the only requirement is that the quantity communicated is the same for both sender and recipient.
Numbers, information, purposes, and other abstractions don't do any work. Except in our thinking, which is itself an abstraction.
The problem is that what you perceive as "information" is "meaningless" independent of some "context" (for example, software without hardware to run it). The context itself can also be information, perceived in yet another context, and so on, until the ultimate context is reached: your mind. That's where "meaning" is assigned (and I've toyed previously with the idea that "meaning" is a subset of qualia.)
You argue well. I wonder though, how these views on information can be reconciled with the functionalist view of mind that you also (seem to?) have - i.e, if mind truly understood itself, then information would contain it's own context. Unless perhaps you take the odd view that functionalism is true, even though mind can never be completely understood.
This blog entry somehow misses an important point in my opinion. The known universe (at least so far) is entirely indistingusable from a simulation of itself so that in fact it is valid to view everything as interacting information. Meaning is attached to a viewpoint ie. an observer. So if a self perpetuating system such as life and its dna functions quite well without anyone to observe it, meaning is irrelevent function is all. Information is an abstraction of function and is described mathematically. Protean coding is encoded in genes as function and the abstraction yields information, there is no way around it. Disrupt the encoded sequence and functionality is disrupted exactly the way your hard disk reacts to the impact of a sledge hammer. Evolution is the slow or not so slow accumulation of functionality due to interference and errors in the encoding process. Only those encoding processes which are able to reproduce themselves survive and also encode the functional history of previous generations. Now a rock also is encoded information but more limited. It holds radio isotopic information, crystallization temperature, deposition and may even contain an imprint of a trilobite. Re-encoding the rock is a destructive process but no less informative.
@10 (murffy)
Re: I'm confusing meaning with information. Can there be information without meaning?
I'd also 2nd Tyler's yes. A good example would be sections of DNA known as junk DNA and/or noncoding regions.
At one point, these regions were believed to contain no "information" in the sense that researchers did not understand what particular function such DNA served. But recent research is now suggesting that junk DNA appears not to be junk since they appear to serve some function.
If you go with the colloquial definition of information, that would suggest that junk DNA wasn't information and then suddenly became information once the researchers figured out the function of junk DNA.
Regardless of the increase in the "meaning" of junk DNA, the information was their and clearly existed independent from our understanding it.
Re: junk DNA example
Wouldn't it be similar to "fox" on a blackboard? Without something to perceive it as information (in this case cellular machinery), it remains just a strings of guanine, cytozine, etc.? It seems to me it can only be considered information if there's something that interprets it as such, or potentially interprets it as such.
It's possible you could have some random string of DNA components floating around out there. It finds its way into a cell and the cell's machinery uses it to fold a protein in a different way. Did the random string of DNA components contain information prior to its being absorbed by the cell?
I've thought about this stuf some, but not very successfully. Gregory Bateson comes to mind. I'm thinking there's got to be some kind of feedback loop in play before you can have information.
"I'm confusing meaning with information. Can there be information without meaning?"
Yes, although it's been sometime since I read through the talk.origins archive on the subject... Information, in terms of information theory, concerns only the amount of bits it takes to describe an entity, not the actual content that the information may or may not code for.
In that sense, a stretch of repetitive non-coding DNA (eg Alu) has information as it takes x number of bits to describe it, but that sequence of bits doesn't on it own say all that much about the content of that information. Comparisons and fun with statistics is required to actually begin to look at what the information actually encodes for here.
It's at that above point which YEC/ID fails at, using "folk" definitions of information and treating it as though it's a truly abstract property, divorced from what ever medium it's recorded on, or describes...
eg: http://www.gnmagazine.org/issues/gn58/tinycode.htm
I think the all-to-tempting slide from Shannon information- an abstract mathematical quantity with no relation to any semantic content- to the ordinary-language meanings of "information" is the heart of the problem that our esteemed bloghost is trying to address here. The trouble is, that even biologists who talk about information have a lot of difficulty sticking rigorously to the former- when we talk about genes as repositories of information, I submit that we're usually wanting to talk about something more than just their Shannon entropy. And that's where the word starts to lose precision and become a source of confusion.
Shannon information is more in terms of understanding human information processing. English is less information-rich than a code of 26 letters might seem, since they don't all occur with equal probability. For example.
But all Shannon information is - log 2 (probability), and the only way to get more information is to drop the probability further, so all the whining about "you can't add information" is just another probability whine, and we know how to deal with those.
Specifically, point to the lottery numbers in the daily paper. What are the odds on those? And all you need to do to "add information" is include yesterday's paper.
@15
The junk DNA example was used as a suggestion that the existence of particular forms of information are independent of our imparting meaning to such information. So in this instance, junk DNA was doing whatever junk DNA does before our understanding of what it does.
This applies to the fox example as well because the information 'fox' (i.e., a string of letters) can still exist independently from our imparting meaning to this letter string. Consider for example, a child that has not yet learned how to read vs. an adult that can. Is it possible to say that 'fox' does not represent information for a child while it does for an adult? I would say no, because the letter string exists independent from any kind of interpretation.
The question of whether a random string of DNA contained information prior to cell absorption, I believe, is an argument about information from a contextual perspective. Really it is the same as the question of whether the word 'fox' is not-information for a child while it is information for an adult. Further, consider reversing the stated context--does the random DNA string loose information when it leaves the cell and has no apparent function outside of it?
I think this entire discussion would be made clearer if a distinction was made between the word information and the concept of inscription devices--the indirect observational methods that scientists use to communicate and think about actual phenomena. The clearest example would be the use of graphs to visualize data from phenomena.
Using the fox example, the letter string fox and a picture of a fox would be inscription devices. An actual fox would not be. Also important to point out, one inscription device is closer to a real fox (i.e., picture of a fox) than the other (i.e., the word fox). This last example is important because I think this is really what Dr. Wilkins is arguing about.
The question really is not about whether something is or is not information, but rather, about whether there are inscription devices so abstract, that they do not represent any actual phenomenon.
This is all over my head. I have understood that transfer of information requires energy, and thus one can think about information in thermodynamic terms. (I say one and do not mean me.) We have a DNA sequence which specifies a particular polypeptide, and, with energy input, the polypeptide molecule is constructed. Is this any evidence for presence of information in the DNA sequence?
Jim,
It is evidence that physical causation is going on. Why would it be evidence that information transfer is going on, especially? Causal templating occurs is many chemical reactions, particularly catalytic :-) so why call it "information"?
How about if physical causation is information transfer?
Here's a paper by philosopher John Collier proposing this idea:
http://www.nu.ac.za/undphil/collier/papers/causinf.pdf
Is the text in a book considered information? The score for a symphony? The blueprints of a building? If yes, then it seems that genes would also constitute information by analogy. If not... I must confess ignorance in formal definitions of information, but surely there is correlation between the formalism and intuition?
Information has to be independent of meaning, observer and even the observer's existence. Because if it's not, then it's subjective and as such cannot be fully described. Now, if it's independent from whether we exist or not, it has to be described in abstract terms, surely not in terms of "physical carriers".
Thinking out loud ...
1. Information is distinct from meaning.
2. The information content of a object is equal to the number of bits that it would take to describe it.
Applying 1 &2 to the random DNA string example, we can say that the string contains information equal to what it would take to describe it. This information content is independent of a perceiver. Okay, so far so good.
The DNA string is absorbed by a cell. The cell's internal machinery uses it to fold a protein in a certain way. Is the information that the cell gained from the DNA string a subset of the total information needed to describe the string?
Now the DNA string is ejected by the cell and is absorbed by a different kind of cell. This new cell uses the string to fold a protein but in a different way than the previous cell. Is this another subset of the total set needed to describe the string? It's looking to me like maybe it isn't. There's potentially an indefinite number of uses the DNA string could have for different cells depending on the nature of the a cell's internal machinery. Therefore the string possesses an indefinite amount of information.
It's looking to me like you can't divorce information from context. Or you have to say an object contains essentially infinite amount of information.
2. The information content of a object is equal to the number of bits that it would take to describe it.
Assuming you can know all the "bits" that it would take to describe an object, any "description" of the object is still not equivalent to the object itself. "Something" is lost. What is that "something"?
Derek I know John Collier's stuff well - he was the first guy to teach me about information theory. But if you grant his conclusion, then information is nothing special in genes - we can say there is information in everything, so the informational role of genes is deprecated to what we previously just called "causation". Which is my claim.
John also had a lot to do with Brooks and Wiley's claims. Again here the notion of information is roughly equivalent to "complexity", but B&W (and John) did not deal with how that complexity is arrived at. If I'm right, they are simply saying that (i) thermodynamic entropy is decreased and this means (somehow) that informational entropy is increased, and (ii) the things we are interested in get more complex to describe. Not many folk are convinced by Brooks and Wiley on this. Including me.
Mikhail, elegantly stated, and what I think too.
Information has to be independent of meaning, observer and even the observer's existence.
Can information ever be separated from meaning? Can you even speak of information without "meaning" being present? For example, even a random sequence of one's and zero's has meaning: "random" means something.
It's analogous to speaking about something that is an observer or something that is observed independently, without considering the other. You can't, since the existence of one implies the existence of the other.
Information has to be independent of meaning, observer and even the observer's existence.
...and even the word "information" has meaning. The two cannot be separated. Without meaning, you can't think or talk about anything at all! (unless perhaps you're a politician, a newborn babe, a postmodernist, a professional idiot, or a creationist.)