reality engine subassembly 764 BSQT 77

Information WTF 2: The Candy Itself

I’ve already written one post in which I express skepticism about information-talk: Culture Memes Information WTF! I fear, alas, that it’s time for another. It’s not that I’m not aware of the concept of information, or that I haven’t made use of the concept, both the Shannon-Weaver technical concept and the more informal concept. But I think information talk is tricky.

As I said in that earlier post, a signal can be said to contain information only with respect to a system that can read and write that information. It’s one thing to talk about information when you have some technical understanding of the read-write mechanism, which seems to be the case in biology. But where such understanding is weak and vaporous, as it is in the case of human culture (the brain is more of a mystery than not) information talk is dangerous. Discussions of cultural evolution over the past quarter of a century or so have, for the most part, been blissfully uninformed by the cognitive and neurosciences and so, I fear, information talk has been a device for avoiding problems rather than solving them.

I get the sense that such talk is based on an underlying notion of pushing bits through a tube—cognitive linguists talk about the conduit metaphor. The nature of the tube doesn’t matter; its size, shape, substance, are all irrelevant. All that matters is the bits.

Language, for example, doesn’t work like that. All that goes through the tube is a physical signal—a sound wave, a visual signal (written or inscribed marks in the case of writing, gestures in the case of sign language), or a tactile signal (e.g. Braille). The meaning doesn’t go through the tube. The speaker’s intended meaning stays inside the speaker’s head. The listener constructs his or her own meaning according to his or her own perceptual and conceptual resources. In many cases these two meanings are congruent, especially in routine and relatively simple matters. There are, however, many cases where the two meanings ARE NOT congruent. In those cases congruence may be achieved through back-and-forth conversational negotiation. But not always, negotiations may fail.

The same, or course, is true of written communication. In that case the negotiations may span continents and decades or more—as is the case, for example, of negotiating terms of discussion appropriate for cultural evolution.

Let me offer a rather heavy-handed analogy (which is reworked from Three Evolutionary Pieces, p. 42). A speaker places meaning into a sound stream like placing candies in a tin floating in a stream. The fishes the tin out of the water the reader eats the candies, one delicious bonbon after another. The listener eats the candies supplied by the speaker. Yum!

But, as I’ve said, language is not like that. We have known that at least since the early 20th century when Saussure made it clear that the relationship between the sound (or image) of word and its meaning was arbitrary. Of course, society goes to great lengths to inculcate consistent linguistic practice among its members; it is not as though listeners get to construe the speaker’s meaning in an arbitrary fashion. Still, as we all know, misunderstanding is common. The reason is the meaning doesn’t go through the tube; all that goes through the tube is a signal that’s associated with meaning.

Let’s get back to that candy floating down river in a tin. Instead of candy, the sender places names of candies into the tin. The recipient who fishes the names out of the tin then goes and procures the candies that were named. But the candies he or she procures may or not be like the one’s the sender had in mind. And that’s very different from the first version, where the candies themselves go into the tin.

But you make my point, says the Info Head. It’s about information, and names ARE information, are they not?

No, they’re not, not in the terms I’ve established in this analogy. The point of information-talk is that it’s all about the information, not the particular form the information takes. It’s the information itself that’s the real deal, not the particular embodiment of the information.

Well, in that analogy, it’s the candy itself that’s the real deal. Not the candy wrapper, and certainly not the mere name of the candy. The candy itself is playing the role that “information” plays in most information talk. And where the information talk is about language, it’s not meaning that’s embodied in the speech stream (or the written text), but only pointers to (names) or meaningful ideotypes (to use the terminology I introduced in my previous recent post). The speaker’s pointers point into the speaker’s mind and the listener’s pointers point into the listener’s mind.

All too often information-talk is used as a way to forget or get around that inconvenient fact. THAT’s why I don’t like information-talk.

  • Tim Tyler

    I base memetics on information theory pretty explicitly in my book – and in various articles (e.g. see “Informational genetics”). The basis of the idea is that we need a science of heredity that unites genetics and memetics – and Shannon-Weaver information theory is the most attractive-looking common ground.

    For me, information theory seems pretty orthogonal to the issues raised in this post. It is true that sometimes people use words in different ways and that sometimes they misunderstand each other. For me, use of information theory doesn’t obscure such processes, it’s just a basic modeling tool.

  • http://new-savanna.blogspot.com/ Bill Benzon

    “For me, information theory seems pretty orthogonal to the issues raised in this post.”

    It is, which is why it has little or nothing to say about linguistic meaning. While it IS true, as I’ve said, that fairly large swaths of interaction have become so routinized that we can treat them AS IF communication weren’t so recondite as I’ve described. But that’s only AS IF. It’s provisional. And when the time comes to create new ideas, new meanings, that AS IF disappears. Those new meanings HAVE to be negotiated. Information theory can’t deal with that situation.

    As far as I know information theory has nothing to say about semantic information. And Dennett would seem to agree (see his remark at the end of question 4 here: http://ase.tufts.edu/cogstud/papers/computing.pdf).

  • Tim Tyler

    Dennett has hopes for a theory of useful (or adaptive) information. It’s possible to imagine such a theory – though it would be remote from Shannon and Weaver’s work.

    “Meaning” seems a bit different, though. IMO, evolutionary theory “black-boxes” the topic – in much the same way that it “black-boxes” developmental processes and selection pressures – and so is broadly compatible with a wide range of theories on these topics.

  • http://new-savanna.blogspot.com/ Bill Benzon

    And if the black box for meaning should turn out to contain a Transcendental Designer, what happens to your theory then? I always thought that a major virtue of Darwinian theory is that it doesn’t need such a Transcendental Designer to work.

  • Tim Tyler

    This is about how academic disciplines are divided. Traditionally, evolutionary theorists have made only a minimal study of developmental biology. It’s not because it isn’t relevant at all, it’s because it’s a quite different topic, requiring different specializations, and much useful work can be done without getting into it. If – hypothetically – an intelligent designer intervened in evolution during development, it would be the job of developmental biologists to find that out.

    Cultural evolution faces a broadly similar situation with psychology. Psychology is a large and complex topic, which affects how culture evolves. However it already has its own discipline and departments. Cultural evolution theorists should mostly try and interface to that existing body of work. That means models with pluggable “black box” components – but modular modeling is virtue, and the approach is often better than attempting to reinvent the wheel.

  • http://new-savanna.blogspot.com/ Bill Benzon

    Psychology is a large and complex topic, which affects how culture evolves. However it already has its own discipline and departments.

    Well, yes, the academic world is divided into disciplines and colleges and universities stuff those disciplines into department. But the lines between disciplines are not well-defined and can be quite porus. Furthermore, psychology has been ‘caught’ in two major interdisciplinary movements in the last 60 years or so. The first is cognitive science starting more or less in the 1960s, though the term was coined until 1973 (by Christopher Longuet-Higgins). Cognitive science cuts across and encompasses cognitive and pereptual psychology, linguistics, philosophy, computer science (AI), and neuroscience. More recently we have evolutionary psychology cutting across psychology, ethology (primate ethology in particular), evolutionary biology, and neuroscience.

    Cultural evolution theorists should mostly try and interface to that existing body of work.

    Yes. I’ve been studying culture, including cultural evolution, for over 30 years, and have published in literature and literary theory, cognition, brain theory, visual theory, and music. In 2001 I published Beethoven’s Anvil: Music in Mind and Culture. There I place music in the context of neuroscience, including a novel way of thinking about how individuals interact and communicate; cognitive and perceptual psychology; comparative anthropology and music history; evolutionary biology, paleo-anthropology and archaeology (music origins); and a bit of philosophy. In that book I discuss musical memes and phenotypes in pretty much the same terms I’ve been using in this series of posts. I chose those terms because they are mutually compatible with the various disciplines I’ve called on.

    I’ve also done quite a bit of work on the semantics of natural language, where I was trained by David Hays, one of the founders of the discipline of computational linguistics. I wrote a dissertation on “Cognitive Science and Literary Theory,” a great deal of which was technical work in knowledge representation for natural language.

    That means models with pluggable “black box” components – but modular modeling is virtue, and the approach is often better than attempting to reinvent the wheel.

    That implies interface standards to which one adheres. Unfortunately just about all the disciplines involved have various competing schools of thought which are not mutually compatible. Thus there is no widely-accepted standard to which one can design “pluggable” conceptual components.

    However, the ideas I present in this post have been widely accepted for over 100 years, though not necessarily in the terms that I use here. That language is patterned on two levels (known as duality of patterning), that of sound (or visual symbol or gesture) and that of meaning, is one aspect of the “interface standard” of language conceptualization. That standard also includes the idea that, in the case of words, the relationship between sound (or visual symbol) and meaning is an arbitrary one.

    And account of cultural evolution that ignores this or treats it as trivial thus runs the risk of not being plug compatible with contemporary thinking about language in the disciplines of linguistics, psycholinguistcs, cognitive psychology, and neurolinguistics, etc. You initial comment on this post seems to treat this point as trivial, as something you needed be concerned about. I conclude that your “information theory” account of memes is not “plug-compatible” the best contemporary thinking about language. Your model, and Dennett’s, can work ONLY if there is a Transcendental Designer to legislate word meanings from a point-of-view outside human society.

    I know you don’t believe in such a thing, nor does Dennett. It seems to me you have to change your theory in some way. Either supply it with mechanisms that are compatible with linguistic and psycholinguistic theory or start thinking about the design and operational specs of the Transcendental Designer.