Tuesday, October 11, 2011

Information

I'm not sure about the red fedora, but otherwise Raymond Tallis sounds like fun.


There's another review of his most recent book here. Part of it reads thus:
One of Tallis's central points is the discussion of "information". This word plays a central role in the Dawkins/Dennett world view, much more important and less obvious than the nonsense about "memes". Brains, computers, and even life itself, are all said to be processing information. DNA itself is pure digital information. But the word here needs scare quotes throughout, for it has two quite different and separate senses. The older usage of the term is inextricably bound up with meaning: information is something you know that carries a meaning. It is, in engineer's jargon, signal, rather than noise. Information, in this sense, is always information to someone or some system.
But there is a second sense of "information", arising from electrical engineering, and the beginnings of computer science, in which it is entirely measurable, and can be broken into discrete chunks. This has been an important and productive understanding – I couldn't be typing and you couldn't be reading without that kind of information science – but it came at the price of breaking "information" entirely away from meaning. Tallis quotes one of the pioneers in the field: "Information, in this theory is used in a special sense that must not be confused with its ordinary usage. In particular, information must not be confused with meaning. In fact, two messages, one of which is heavily loaded with meaning, and the other of which is pure nonsense, can be exactly equivalent, from the present viewpoint, as regards information."
This seems to be a good point well made.


4 comments:

  1. "This seems to be a good point well made."

    Not really. Several assertions in the review are questionable or flat out wrong. Eg, the distinction between the two interpretations of "information" has nothing to do with signal vs noise.

    The second kind of information - Shannon information - is merely the resolution of uncertainty. You and I agree that I will signal to you by raising either my left or right arm. Before I do either, you are uncertain which of the two actions I will execute. Assuming you have a clear view (ie, no "noise"), my raising an arm resolves that uncertainty, thereby conveying to you one bit (binary digit, in this case "L" or "R") of Shannon information.

    So far, there is no "meaning" attached to that resolution. But suppose we have agreed that raising my left arm ("L") means "by land", raising my right ("R") means "by sea". Then my action has also conveyed meaningful "information" of the first kind, ie, you now know what you should do.

    Returning to signal and noise, in my example noise may be introduced if the sun is directly behind me or it is getting dark, either of which can make it hard to detect my "signal". This can introduce errors, ie, affect whether or not the bit of Shannon-information I intended to convey is correctly received by you. But it doesn't affect the meaning attached to which ever bit you determine - correctly or not - that I sent. Thus, the reviewer seems to have the roles of the two uses of "information" backwards.

    Whether the reviewer is misrepresenting the author or faithfully summarizing the author's misunderstanding isn't clear - although the insertion of gratuitous pokes at Dawkins and Dennett leads me to question the reviewer's credibility in general.

    No doubt more than you really wanted to know about the topic. But relating it to philosophy of language ala LW, one can think of the four possible utterances in the primitive language used by the builder in PI Remark 2 as conveying two bits of Shannon-information to the helper (chance, or in conveniently assuming four stone types did LW anticipate information theory?). But only if the helper has been trained to respond in an agreed upon way does the language take on meaning, ie, convey information of the first kind.

    ReplyDelete
  2. Thanks! I'm always glad to learn, so it's not more than I wanted to know. I'm also pleased to find another Virginia-based Wittgensteinian, and one with a blog that's visited by people who visit here. A good find for me.

    Maybe the point was not so well made after all, given what you say about signal and noise. But I took the idea to be that we hear a lot about information or messages being sent from the brain to another part of the body, as if the various parts were intelligent agents, like those cartoons of the body as a machine full of rooms inhabited by little people who operate it. This kind of image is potentially useful, but also potentially misleading, and can cause people to accept a certain view of the mind, and to reject others, without really thinking about the issue at all. I take Wittgenstein to have been concerned with opposing this kind of thing, as Hacker and Bennett (I think) are, and as Tallis appears to be. Rightly, in my opinion.

    ReplyDelete
  3. That "information" is often misused in this context is indisputable, but IMO the problem has nothing to do with failing to distinguish those two kinds of information but instead with failing to distinguish between using a vocabulary, possibly appropriate for some purposes at one architectural level, at lower architectural levels where it is arguably inappropriate. For example, the vocabulary of agency or that of communication system engineering may be appropriate for describing behavior of humans at the social level but can be confusing and misleading (for the reasons you mention) at the brain architecture level where a neurological/physiological vocabulary seems more appropriate. As I recall, Crick and Koch (et al) have applied information theory concepts to brain function, but I don't know if that approach has gone anywhere.

    I'm afraid you grossly overestimate the extent to which my "blog" is a find - it's more a personal notebook, which I almost never advertise and rarely update. And I'm a "Wittgensteinian" only under the most generous possible definition possible, engaged as I am through a first reading of PI (although I've read lots of Rorty and some Davidson, thereby being indirectly exposed). Sorry to disappoint. On the other hand, I consider your blog a legitimate "find" - hat tip to Philip Cartwright.

    ReplyDelete
  4. That "information" is often misused in this context is indisputable, but IMO the problem has nothing to do with failing to distinguish those two kinds of information but instead with failing to distinguish between using a vocabulary, possibly appropriate for some purposes at one architectural level, at lower architectural levels where it is arguably inappropriate.

    Yes, that's probably a better way of putting what I think is meant to be the same point. Maybe I'm generously reading a good point into a bad mess, but I'd prefer to interpret Tallis charitably. At least until I actually read one of his books.

    And almost anyone who works on Wittgenstein is a Wittgensteinian in my book.

    ReplyDelete