Preface

I expect that many people on LessWrong have heard of Korzybski, of General Semantics (K's name for his system), and of some of the phrases he coined, such as "the map is not the territory". abramdemski has recently posted on an aspect of his system that I want to communicate more fully here. I wrote a comment on this rather a long time ago. This is the long version.

Korzybski invented the term "consciousness of abstraction", which has to do with being aware of the "ladder of abstraction" (I think another of his coinages). It seems to me that this skill covers a lot of other things that have been spoken of on LessWrong and elsewhere, such as "Looking", the leakiness of abstractions, the nature of definitions, and abramdemski's article. It relates to a lot of the series "A user's guide to words". It is a fundamental aspect of rationality, but it has not yet been the primary focus of an article.

Korzybski is the source where I encountered it, but I can only recommend reading the original work ("Science and Sanity", 4th or 5th edition^[1]) to those who specifically want to see what the original said, and are willing to deal with its great length, turgid style, and outdated science. Instead I would recommend S. I. Hayakawa's "Language in Thought and Action", which gives the ideas in briefer and more readable form.

Consciousness of abstraction

There is a world out there. (Those inclined to dispute this are invited to do so elsewhere, after putting this to the test by leaping in front of a train.) Our senses give us the naive impression that we have some sort of direct knowledge of that world, but a minimal familiarity with the mechanisms of perception show that that is an illusion. We have physical sense organs that interact with the outside world. These generate electrochemical signals in nerve fibres, and somehow (I shall not attempt to unpack the "somehow") we experience perceptions. Put an external obstacle in the way and the perception goes away. (The sun is always shining, even at night.) Interrupt the nerve signals and the perception goes away. (Your mouth is still there, before the dentist's anaesthetic wears off.) The thing out there and its effect on the sense organs are physically different things. The sense organs and the signals they send are physically different things. The signals and what we make of them are physically different things. And the latter continues within us. We experience sensations, organise them into perceptions of things (what K. calls the "object" level), give names to the things, make inferences from the things, have emotional responses to the inferences, and so on. This "ladder of abstraction" can be continued indefinitely.

Our sense organs only interact with the outside world in limited ways. We perceive a single octave of visible radiation. Some other creatures see further into the infrared or ultraviolet than we do. Some have senses that we lack entirely, such as sensitivity to the Earth's magnetic field. But whatever the sense organs, they leave out almost all of the possible things one could sense about the things out there.

The nerve signals leave out most of what the sense organs might supply. Our experienced perceptions are less detailed than the nerve signals.

Then, presumably in our brains, the incoming nerve signals are subject to further levels of processing. Consider the great number of photosensitive cells in the retina. We cannot perceive their individual firings. Most of that information is discarded in the construction of more abstract entities: neurons on the optical cortex detect edges, moving lines, color constrasts, and so on. We are not aware of those processes either.

After further layers of abstraction, we are aware of "objects". A computer, a cat, a houseplant, and so on. In recognising and naming the thing, we identify it with other similar things and leave out its particularity.

Consciousness of abstraction is being continually aware of that process. A realisation at a basic level that one's thoughts about the thing are not the thing, the thing is not the sensation, and none of that is the part of reality out there that gave rise to them.

Korzybski developed a teaching aid for explaining the ladder of abstraction, which he called the Structural Differential. Here is a picture of it.

The Structural Differential

The ladder of abstraction ascends downwards. At the top is the lowest level, the thing in itself, whose parabolic shape is intended to suggest that it goes on forever. The thing in itself is inexhaustible. There is always more that we can discover about the thing. Below that is the first level of abstraction, say, the interaction of our sense organs with the thing in itself. The strings connecting them suggest how our senses select certain properties of the thing. They are insensitive to others, indicated by the unconnected strings dangling from the level of the thing in itself. The word "deification" in the image I chose should be "reification", the upwards arrow indicating the fact that our abstractions are themselves part of reality, the parabola at the top. The "Etcetera" refers to something different, the fact that the chain of abstractions can be continued indefinitely below the bottom of the picture.

The discs and rectangles representing the senses and the further levels of abstraction are finite, representing the fact that our abstractions are finite, limited.

The chain of abstraction can, in humans, be continued indefinitely. On every level of abstraction we can build a new one. In this, we differ from other creatures. The others only have a bounded capacity for abstraction. This is indicated by the disc off to the left, which has no further discs below it. It does not matter whether a dog is capable of two or three or five levels: what matters is that it reaches a ceiling. Dogs have only a limited capacity to contemplate themselves, and lower creatures even less. A rock has none. This is why the device is called the Structural Differential: it exhibits the difference between humans and other creatures.

It is not necessary to enumerate a precisely defined series of levels of abstraction. One might divide them more or less finely. A precise set would have to await a precise understanding of the mechanisms of the brain. The principle, that there is such a ladder of abstraction, is what matters. There is no such thing as "Korzybski Level III" or whatever.

Korzybski urges his readers to make a physical Structural Differential and handle it, to give a more physical understanding that this level is not that level, than would obtained merely by reading and thinking. I must admit that I have never done this, although I have drawn out the diagram a few times. But I believe that I have learned its lesson.

The map is not the territory. The sensation is not the thing. The perception is not the sensation. The object is not the perception. The reaction to the object is not the object. Emotions, thoughts, etc. about the reaction are not the reaction. The name is not the thing named. And so on.

Whitehead said (and Korzybski quotes him) "The negative judgement is the peak of mentality." I do not know the context in which Whitehead said it, but Korzybski means by it the fundamental realisation, the grokking^[2], that this is never that. No two things are identical. This apple is not that apple. This apple today is not this apple tomorrow. You right now are not you of a day ago, or an hour ago, or a second ago. You cannot step into the same river twice, because both you and the river have changed. The differences may be important or unimportant in a given context, but they are always there. This is the principle of non-identity. This is what Korzybski is talking about when he decries "Aristotelian" thought, based on the law of identity: "A is A". "This apple", the object in your consciousness, is not "this apple", the portion of reality out there. Neither is the same "apple" as a moment ago or a moment hence.

If all of the different levels are so different, if they leave out so much, how are we able to use these maps? What makes a map useful, when it leaves out so much? Korzybski answers this: "A map is not the territory it represents, but, if correct, it has a similar structure to the territory, which accounts for its usefulness." Category theory had only just begun in K's time and I doubt he knew of it, but if you know category theory, you will instantly see what he is talking about.

Korzybski and the Sequences

I do not know if Eliezer has ever read Korzybski, but I find it difficult to imagine that he has not. He does say that he read Hayakawa's book as a boy. In the General Semantics community, the question has been asked, "Where do we go from here?", i.e. since Korzybski's death in 1950. I think the best answer to that is "The Sequences", which I would characterise as "Korzybski done right". Vastly better written, more up to date science, and more wide ranging. This post is only intended to turn a spotlight on a single concept that has been touched on many times in the Sequences, and say, this is an Important Thing that you should have a handle for.

Some applications

Hayakawa contrasts two different ways one might respond to the question, "what is red?" We could go, "Red is a colour." "What is a colour?" "A perception." "What is a perception?" "A sensation." And so on, up the ladder of abstraction. Or we can go down the ladder of abstraction and point to examples of red things, saying, "these are red." Philosophers, and Korzybski, call these two approaches "intensional" (with an "s") and "extensional" respectively. The extensional orientation is more useful. Compare Eliezer on "Making beliefs pay rent", and "Extensions and Intentions". Eliezer gives this same example but is uncertain of whether his attribution to Hayakawa is correct; it is. Both his and my versions are paraphrases. I don't have the book in front of me.

To deal with the fact that this is never that, Korzybski recommends the use of subscripts. "This apple₂₀₂₀₋₁₂₋₂₁" is clearly a very different thing from "this apple₂₀₂₁₋₀₁₋₂₁".

To maintain awareness that elements are left out in every step up the ladder of abstraction, he recommends the frequent use of "etc.", and used the contraction ",." to mean ", etc."

To avoid leaping to judgement and unconscious identification of thought with thing, K recommends pausing, to allow the slower, higher-level systems time to catch up with and take control of the knee-jerk reactions. He called this the "cortico-thalamic pause", based on what might be outdated neuroscience. The more current jargon, if not necessarily more empirically founded, is System 1 vs. System 2.

Something to protect

Korzybski's other important work, "Manhood of Humanity", is his vision of how mankind can and must develop beyond its "childhood" and become truly sane, which he saw as the most urgent problem of his time. That is the motivation for his life's work. This is where he coined the term "time-binding": the ability that man has and other creatures do not, of building upon the past, lifetime after lifetime. Here is a quotation from early in the book:

"The conclusion of the World War is the closing of the period of the childhood of humanity. This childhood, as any childhood, can be characterized as devoid of any real understanding of values, as is that of a child who uses a priceless chronometer to crack nuts.

"This childhood has been unduly long, but happily we are near to the end of it, for humanity, shaken by this war, is coming to its senses and must soon enter its manhood, a period of great achievements and rewards in the new and real sense of values dawning upon us."

The world, alas, has not gone that way. If you are wondering which World War he refers to, the book was published in 1921. Later, we had to start numbering them.

Books referred to

These are all quite old, but you can find them in the usual places. Dates are of original publication.

Korzybski's magnum opus: "Science and Sanity: An Introduction to Non-Aristotelian Systems and General Semantics", 5th edition 1994.

Korzybski's authorised abridgement: "Selections from Science and Sanity", 2nd edition 2010.

Korzybski's other important work: "Manhood of Humanity" (1921). You can get this on Gutenberg.

S. I. Hayakawa "Language in Thought and Action" (1949)

The 5th edition is the latest, but it was posthumous, and the 4th is the latest that Korzybski produced. The 4th contains significant additions by K over the 3rd. The 5th contains all of the 4th, but I am not sure how substantial its extra material is. ↩︎
"Grok" is Heinlein's word. Heinlein references General Semantics a few times in his fiction, although I don't recall specific examples. I have the impression from his fiction in general that he was quite familiar with it, and not just name-checking a meme of the times. In fact, it was through his stories and van Vogt's that GS became widely known of in the science fiction community. Van Vogt makes it a central theme of his Null-A^[3] trilogy, although in an over the top way that might put serious enquirers off rather than interest them. ↩︎
Null-A, or Ā as we can now write in Unicode^[4], is K's abbreviation for "non-Aristotelian". He also coined Ē ("non-Euclidean") and N̄ ("non-Newtonian") for reasons not important enough to go into here. ↩︎
Unicode spoofing is an example of the non-identity principle. "This URL" (what I take it to be when I see it) is not "that URL" (the actual string of character codes, and hence the web site that it references). ↩︎