Cryptotheories and cognition

One of the things I most enjoy doing is spotting holes in linguistic maps – places where people habitually circumlocute their way around a word – or, more properly, an important bundle of concepts tagged by a word – that they don’t know they’re missing.

Sometimes, filling one of these holes can shake up everyone’s view of the linguistic map near it in a way that changes their thinking. One of my favorite recent examples is Martin Fowler’s invention of the term “refactoring” in software engineering, and what that did to how software engineers think about their work.

About a year ago I invented a hole-filler that I think is useful for getting to grips with a large class of slippery problems in the philosophy of mind, knowledge, and perception. I’ve meant ever since to develop it further.

So, welcome to three new words: “cryptotheory”, “acrotheory”, and “mesotheory”. Of these, the most important (and the motivator for the other two) is “cryptotheory”.

A theory is a prediction-generating machine. It takes as input some set of observables and generates as output predictions about the not yet observed. One of the classic examples is Newton’s Three Laws of Physics. Using these, expressed in the formalism of calculus, we can put in observations of force and mass and motion and get out predictions of future force and motion.

(If you are studying something that people call a “theory”, and you can’t identify what inputs it takes and what predictions it generates, the “theory” is almost certainly bogus. I say “almost” because it’s possible you don’t yet understand the theory well enough to do that identification. On the other hand, there are a lot of bogus pseudo-theories floating around out there; caveat thinker.)

We’re used to applying the term “theory” to prediction generators which we arrived at by conscious reasoning, apply with conscious reasoning, and modify with conscious reasoning. But this is not the only kind of theory there is; today’s hole-fillers contrast it with two other kinds.

A “cryptotheory” is a prediction generator of which we are normally unaware because it is not learned but rather wired into our brains and nervous systems.

For example, your eye-brain system has a cryptotheory that if we attend to two extremely similar-seeming objects on a relatively blank field and they are of different sizes, that means the smaller one is further away. There are well-known optical illusions that exploit this. Some other illusion confusions, especially those having to do with edge recognition, seem to arise from image processing in your retina, before visual stimuli even reach the optic nerve (let alone the brain).

What reaches your attention when you see is not ‘reality’ but a mix of light measurements with cryptotheories that were useful for making snap judgments in the environment of ancestral adaptation.

Animals can have cryptotheories. In fact, every response we call “instinctive” is precisely a cryptotheory about how and when an animal has survival reasons for a snap judgment. When your cat jumps at your feet moving under bedclothes he is running a cryptotheory that recognizes certain kinds of motion as predicting the presence of tasty small prey animals.

Optical illusions, stage magic, and trompe-l’oeil art confuse us by using our embedded cryptotheories to mislead us about what is in front of us. They work when our cryptotheories yield false predictions.

By contrast, an “acrotheory” is a learned, conscious theory used in conscious reasoning – like Newton’s Three Laws. It is also useful to be able to speak of “mesotheory” – this is a learned theory applied unconsciously, as when you instantly identify a shiny moving object with wheels under it as a vehicle and instantly generate a bundle of predictions about where it is likely to move next and what will happen if you don’t get out of its way.

(The derivation of “cryptotheory” should be obvious. The terms “acrotheory” and “mesotheory” were analogized from the way the terms “acrolect” and “mesolect” are used in sociolinguistics.)

Some philosophers are fond of claiming that “All observation is theory-laden.” This claim causes more confusion than it should because neither those philosophers nor their audiences are very clear about the distinction between crypto-, acro-, and mesotheory.

It would be difficult for even a philosopher to claim with a straight face that all observation is laden with acrotheory. On the other hand, it is reasonable to claim that all observation is cryptotheory-laden and much of it is mesotheory-laden.

Before anything in your phenomenal field gets to your attention, it will have been heavily filtered and classified by a stack of cryptotheories. Some of these are embedded in the way your sense organs process their input, as we’ve already seen with the visual system; others are, for example, your brain’s habit of chunking observations into representations of persistent objects. This doesn’t only apply to vision; if we smell eau de skunk over a period of five seconds, we automatically bin that as “at least one skunk continously emitting thiols”, not as “five skunks and their thiols popping in and out of existence once a second”.

When you see a car heading towards you, several cryptotheories in your eye and brain assemble those observations into “persistent object”, then a mesotheory classifies it as “car”. You will probably jump out of the way before you have time to apply any acrotheory.

These terms are useful because they give us a way to talk about how raw stimuli arriving at our sense organs becomes knowledge of what is going on around us. Awareness of how heavily processed those stimuli are before they reach our attention can also help dispel some common philosophical mistakes. I’ll have more to say about that in a future blog post.

49 comments

  1. Depth perception, edge detection, motion detection and so on are all still useful in the modern environment too. Unless you’re looking at an optical illusion picture.

  2. Nice post, sensei. You blinded me with science. ;-) And with a bit of Kantian philosophy, I think.

    You mentioned stage magic in relation to how our brains work. Have you read Sleights of Mind? If so, is it sound?

    And the usual nitpicking (I can cut it if you want):

    > places where people habitually circomlocute

    Did you mean “circumlocute”?

    > Of these, the most important (and the motivator for the other two) is “cryptotheory.

    You omitted the closing quotation mark.

    > (The derivation of “cryptotheory” should be obvious. The terms “acrotheory” and “mesotheory” were analogized from the way the terms “acrolect” and “mesolect” are used in sociolinguistics)

    Shouldn’t there be a period right before the closing parenthesis?

    1. >And with a bit of Kantian philosophy, I think.

      It isn’t Kantian without some added premises about the relationship between sense perception and “reality”. I was deliberately avoiding those in this post, trying to stick to describing mechanism that nobody can really argue the existence of at this late date.

      >Have you read Sleights of Mind?

      I have not.

      >And the usual nitpicking

      Useful, thanks.

  3. Present-day human neurology is the end product of millions of years of evolution, and what is in existence is what has “worked” in the sense that it has survived the gauntlet of trials that nature has imposed upon our predecessors. Complex lifeforms have the ability to receive sensory input, process data, and then execute an output function. When this chain of events leads to an improvement in survivability and reproduction, then the organism has an opportunity to persist, and this persistence will typically occur when the pattern becomes genetically encoded.

    For higher life forms, genetically encoded behavior is just the beginning (think computer boot process). Immediately after birth (and concurrent with initial brain growth and development), enhancement of these genetic behaviors begins to occur. This process incorporates a greatly expanded sensory input stream, somewhat randomized auto-reactions, environmental feedback, and data logging (memory acquisition). Positive feedback (generally) results in repetition of behavior patterns and this, in turn, causes physiological effects on neuron growth (e.g. hardwiring).

    Evolution has further enhanced this modality by providing a mechanism for later stage reprogramming, which occurs via what might be described as learning at the school of hard knocks. It is at this higher level of brain function (complex neuro-processing) that abstraction and prediction may occur, and this process has it’s roots in both genetically encoded base functions and developmental hardwiring.

    For most of our species’ history, this higher level reprogramming was subconscious because we lacked complex language skill and hence a means to contextualize self awareness.

    Somewhat off topic, but sentient AI will most likely occur when we can recreate this type of evolutionary environment in abstraction space and allow software programs to compete, survive, and evolve.

    1. >Oh, go all the way! “Caveat cogitor” is charming and comprehensible. =)

      This is a subtle point, but will help you understand my writing style better. Yes, I could have written “caveat cogitor” and sounded erudite. I preferred to deliberately mess with the reader’s expectations, sounding playful and unafraid to float the rules of formal-register English.

  4. I’m not sure I see what’s non-obvious here — but then I’ve been reading machine learning textbooks recently, and they tend not to take common sense for granted. Of course when you update on an observation and end up with less than maximal entropy for P(something | that thing you saw) you’ve got a theory about the distribution of one given the other. How could it be otherwise?

    On the other hand, this sort of thing might be why I have trouble talking with ordinary, non-math-aware philosophers….

    1. >On the other hand, this sort of thing might be why I have trouble talking with ordinary, non-math-aware philosophers….

      And one of the purposes of this post is to make it terminologically easier for philosophers who don’t grok Bayesian inference not to play stupid. See, my version doesn’t require them to grapple with math

  5. The problem with “theory” (be predictive) part in “cryptotheory” is that many of examples use *heuristics* and not strict rules; in many cases how fast you can get the result in the presence of unknown is more important than how precise it is and how large the error of the prediction is.

  6. The problem with “theory” (be predictive) part in “cryptotheory” is that many of examples use *heuristics* and not strict rules

    No problem there. Theories can be more or less accurate; Newtonian mechanics is largely hardwired into our brains, and though it’s not the end-all of physics, it’s suitably useful for everyday life.

    1. >Theories can be more or less accurate;

      Furthermore, accurate theories may output probability distributions computed from fuzzy input data. It’s only a language prejudice that makes us think “real” theories have to yield binary truth values with respect to Aristotelian categories.

  7. Just for comparison’s sake, I also had a machine learning word at hand for these sorts of ideas. But I think the relevant word is “bias”, in the way the machine learning community defines it. A system is “biased” because of what it is capable of representing; if you have a system whose input is “roundness” and “yellowness” and you train it on oranges and bananas, it will be biased to be unable to usefully distinguish between an orange and a red apple, because it literally can not represent a useful difference between them.

    Of course, as already observed, this gets a bit mathematical for common usage.

    However, I have found it intriguing at times to muse on how at first it seems like the machine learning definition of “bias” and the common English one seem quite different, but, the more you think about it, the less true that seems. In practice, modeling someone who is, for instance, very “biased” towards $POLITICAL_THEORY as being literally unable to correctly represent the argument of their opponents in their brains seems to work quite often. To the point that it becomes a genuine surprise when that model breaks down in the wild…..

  8. Oh, and to be clear, I do not mean that ML “bias” completely subsumes those ideas. It obviously does not break the space into the three useful categories you delineated. It just wasn’t as much a linguistic hole for me as it might be for others, and I find it interesting to compare existing coverage of the space in an academic sense.

  9. I could have written “caveat cogitor” and sounded erudite. I preferred to deliberately mess with the reader’s expectations, sounding playful and unafraid to float the rules of formal-register English.

    You could also have written “caveat putator” and gotten the bonus that anyone who knows modern Romance languages but not Latin would think you were calling them a whoremonger.

    P.S. Nitpick: I suspect you meant unafraid to flout the rules

  10. I am not sure whether you do not mean “model” where you write “theory”. Theory has many connotations of “explaining” and “understanding” where the ability to predict is not very relevant. We can use the word “theory” for historical events or literary works where input/output relations would be strained.

    I would agree immediately that the human retina and visual cortex have a model of how to relate photons hitting sensor cells to objects in the real world. However, calling that a “theory” suggests that this relation can be made explicit. But current ideas about neural networks would rather say that these relations are implicit and cannot be separated from the neural connections.

    I think that the word “model” is more vague about whether or not the input/output relations can be made explicit.

    On the other hand, your proposal does parallel the ideas behind the “theory of mind”.

    @esr
    “Some philosophers are fond of claiming that “All observation is theory-laden.””

    I understand this idea differently. What is meant is that an “observation” isolates an object (a “fact”) from the total collection of perceived stimuli (light hitting the eyes, sounds hitting the ears). But this can only be done if you have a “theory” that isolates some part from all that is perceived (all light hitting the eyes, all sounds hitting the ears).

    A historical example is the relation between tides and the phase and position of the moon. This simple “observation” only became a “fact” after there was a theory about gravity.

    1. >I am not sure whether you do not mean “model” where you write “theory”.

      I could have used “model”, and was aware of that possibility when I coined “cryptotheory”, but I was launching of Hilary Putnam’s notion of theory-ladenness.

      >I understand this idea differently.

      You are describing Putnam’s notion of theory-ladenness exactly. In my terms, “fact” does not emerge from “stimulus” until it becomes the input of a theory and generates predictions that can be tested.

  11. In physics, all theories must have a mathematical framework (think equations) that describe the relationship between cause and effect. Experimentation is then employed to test the theory by collecting data and comparing inputs and outputs. There is always error in both input and output measurement (hence the fuzziness) and so statistical analysis was born. Knowledge accrues when the correlation attains 3+ sigma. Science of this sort is a very recent development in the history of planet Earth, but the technique has limitless potential for understanding the Universe.

    Evolution of living things is a crude analog of the above. When a behavior pattern achieves a high sigma correlation efficiency, the lifeform tends to persist. When it doesn’t, extinction is frequently the failing grade. It this way, life itself is a natural mechanism for the enhancement of knowledge.

  12. @TomA
    “In physics, all theories must have a mathematical framework (think equations) that describe the relationship between cause and effect.”

    But in the other sciences, the mathematics are generally much less exact. For instance, synthetic chemistry and biology do have models, but these have much less rigorous mathematical foundations. On the other hand, there is a lot in mathematics that does not involve calculating exact quantities.

  13. A while back I realized that English needs a “Dictionary of Internal Events” so that people wanting to understand and change aspects of their cognitive processes would have a conceptual vocabulary for describing them. Such a tool, obviously, makes it easier to spot the places where you’re going off the rails and to try to correct your mistakes.

    When I heard about John Koenig’s intriguing “The Dictionary of Obscure Sorrows” project I wrote a blog post discussing it’s implications for rationality, and proposed a few neologisms of my own.

    Later, as part of The STEMpunk Project I began performing a technique I call a “Failure Postmortem” whenever I found myself trapped in an error cascade. One of the useful insights that came out of this was recognizing how inadequate the language used to describe mental failure modes is.

    In the linked post, for example, I describe trying to troubleshoot an air conditioner setup. During this process I abstracted out phrases like “increasing causal opacity” and “implicasia” to describe mistakes and then designed interventions to prevent the same failures happening again.

    I did this for a few different error cascades, and eventually implicasia became “alpha implicasia” (implication blindness resulting from impoverished system models) and “beta implicasia” (implication blindness resulting from “emotional clouding”, a “sticky gestalt”, or some other failure to invoke rationalist protocols). This allowed even more specific interventions, as well as tantalizing hints as to how powerful a mature version of this technique could be.

    Just think about it: we have ways of describing how golf swings, free throws, jokes, sex, swordplay, and a billion other things are suboptimal, as well as a library of techniques for optimizing them. Yet we have almost zero analogs when it comes to optimizing how we think.

    1. >and proposed a few neologisms of my own.

      Those are good. “Agnosis” and “implicasia” seem especially useful.

      I think your “soulshatter” woud be best as “soulshard”, or perhaps. riffing off Celtic folklore, “soulfetch”.

  14. Y’know, between this and the DSL stuff it occurs to me that you might be able to write a primer on adopting the correct mental stance required to spot and name orthogonal ontological primitives and gaps in linguistic mappings.

    I sometimes jokingly think of myself as an “information topology rotation specialist” or a “conceptual framework maintenance engineer”, but in truth I have a non-trivial talent for this kind of thing and would benefit from such a document.

    At a minimum it’d need to discuss the best ways of drilling down to the ontology of a problem space with some thoughts on sane language design.

    But if you did nothing more than compile examples of successful attempts, with some historical context, and maybe high-level principles that people use to design clean programming language architectures, I bet I could intuit the rest.

    It’d help with The Dictionary of Internal Events, some of my thinking on introspective heterogeneity, The STEMpunk project, as well at least two notetaking project I might tackle in the future, one for logistics and one for getting third-parties to process information for inclusion in Anki.

  15. @ Winter – “But in the other sciences, the mathematics are generally much less exact.”

    See Chaos Theory for a description of how predictive models can become ineffective in the presence of input approximation and dynamical complexity (for example, the inaccuracy of long term weather forecasting).

    That which can be predicted with accuracy, will eventually yield to a theoretical framework. But some phenomena will always remain ambiguous in the maelstrom of chaos. The neurology of human thought processing may fall into the latter category. If so, then it is likely to be a feature and not a bug, since evolution has been a stunning success for our species.

  16. Arcotheories are made of beliefs. Cryptotheories are made of aliefs. This is a spectrum, not a quantization, so the fuzzy middle is populated by mesotheories.

  17. Not to be too rude, but there is a vast bibliography on the meaning of “theory” in science, very little of which I can tie to this post. For starts, theories aren’t individual prediction algorithms, but the gathering of a few such into one more general structure. A general truth, not a specific rule or heuristic. Why throw away decades of other peoples’ earnest (and often very insightful) work – almost the entire field of philosophy of science (much of it based on the history of science); without even reading it? Jejune, at best. Here’s a place to start to gain the basics of the shared vocabulary before attempting to expand it:
    http://plato.stanford.edu/entries/hempel/

    1. >For starts, theories aren’t individual prediction algorithms, but the gathering of a few such into one more general structure. A general truth, not a specific rule or heuristic

      I don’t think I excluded that possibility. In fact, the principal example of an acrotheory I gave – Newton’s Three Laws – has the kind of structure you describe.

      >Why throw away decades of other peoples’ earnest (and often very insightful) work – almost the entire field of philosophy of science (much of it based on the history of science); without even reading it?

      Your assumption that I have not read academic philosophy is neither justified nor true. I minored in it at an Ivy League school.

  18. This is misleading. For one thing, the circuit ?representing a cryptotheory? can be significantly simpler than the cryptotheory it ?represents”. Furthermore, the circuit doesn?t have to predict anything except what will happen if you carry out the action the circuit recommends.

    For example, when you attempt to catch a ball, you follow a very simple heuristic, namely: move so as to keep the angle of the ball with respect to your eye constant. Note, that it makes no predictions about where the ball will land. While it is possible to explain this in terms of cryptotheories, note that your explanation will be significantly more complicated than the heuristic your explaining.

    1. >This is misleading. For one thing, the circuit ?representing a cryptotheory? can be significantly simpler than the cryptotheory it ?represents”.

      Yes. So what? There’s no reason a cryptotheory has to “represent” anything, just be a successful adaptation.

  19. >This is misleading. For one thing, the circuit ?representing a cryptotheory? can be significantly >simpler than the cryptotheory it ?represents”. Furthermore, the circuit doesn?t have to predict >anything except what will happen if you carry out the action the circuit recommends.

    (mumble mumble poor entity-encoding mumble mumble)

  20. “mesotheory” – this is a learned theory applied unconsciously,

    This is good.

    I have been thinking about this – what I call “kung fu” – learning a mesotheory – learning to react to what you see almost as quickly as hard-wired cryptotheories.

    Driving a car in traffic is great kung fu.

    How ’bout – Some parts of martial arts and driving a car connect what you see to the parts of your brain that make it possible to run through a forest?

  21. “mesotheory” sounds too much like something from those ads on cable TV promising to sue your [or that of a now-deceased relative to whose estate you are an heir] former employer over exposing you to asbestos.

    “Call the Law Offices of Boyd, Dewey, Cheatham, and Howe at 1-999-SUE-MESO for a free consultation.”

  22. “By contrast, an “acrotheory” is a learned, conscious theory used in conscious reasoning – like Newton’s Three Laws. It is also useful to be able to speak of “mesotheory” – this is a learned theory applied unconsciously.”

    By extension, we also should have ‘necrotheory’ or ‘idiotheory’ to describe learned conscious response syndromes, or belief systems used in conscious reasoning, which responses or belief systems are described as predictive theories but which are untrue or have no valid predictive quality.

    The explication of examples is left to teh student, as an exercise…..

  23. @ Brian Marshall: “Driving a car in traffic is great kung fu.”

    In fact, it practises a necessary kung fu (wing chun, actually) skill, namely, weaving around obstacles in the environment. In wing chun, you do this without a car, but the mental skills you need to do it with a car in traffic are the same, even though the physical ones are not.

  24. For many years, I have used the phrase “kung fu” to describe a deliberately learned mesotheory, often learned so that a person can do something that wouldn’t be possible at the slower speed of arcotheory application.

    Martial arts, of course. Driving a car without dying or killing people. Maybe most things about which people say “It’s just like riding a bike”. I can “touch-type” pretty fast (I type words that I am thinking, mostly without being conscious of individual letters).

    Once learned, a mesotheory is very much like an arcotheory – a hard-wired response – and operates at close to the same speed.

    To generalize a point in my previous comment, a mesotheory can be particularly effective if connects (highly processed) input to a high-level set of arcotheories, such as those that make it possible to run across boulders (making ~instant decisions about which one to jump to next), run through a forest, balance on a log, maybe use a hand to deflect something about to strike you in the face.

  25. @ esr

    > It isn’t Kantian without some added premises … I was deliberately avoiding those in this post

    That seems to imply you are a Kantian of some sort (albeit not an idealist). :o)

    Is any instinctive or impulsive behavior – human or animal, normal or abnormal – rooted in a cryptotheory? For example: do “cleptocats” act on a theory that attributes some usefulness to the things they steal? Or maybe they act that way precisely because they lack some cryptotheory present in normal brains?

    (Ultimately, the explanation of such phenomena in terms of cryptotheories must be accepted or rejected according to its predictive value; but right now, I just want to know your model in a little more depth.)

    1. >Is any instinctive or impulsive behavior – human or animal, normal or abnormal – rooted in a cryptotheory?

      I hesitate to make a universal claim here. I’d rather discuss examples.

  26. > “caveat cogitor”

    You mean “caveat cogitator” (cogito/cogitare is first declension).

  27. @ esr

    > I hesitate to make a universal claim here. I’d rather discuss examoles [sic].

    Challenge accepted. :-) I’ll try to come up with more examples; for now, what do you make of “cleptocats” and the case reported in this story? (I know you’re against “cute-cat stories”; but I think this case is genuinely interesting, and the article is ridiculously short to boot.) What, if any, are the underlying cryptotheories?

  28. @ESR

    On crypto- and mesotheories: recently I was wondering if a third epistemological category, or a third way of reasoning and arguing needs to be introduced. I came to believe most human desires come from a status drive now. But I can neither bring rigorous empirical evidence nor flawless logic to support it, I have a third kind of argumentation: that this is something that has always been in front of our eyes and requires only as much evidential support as saying the sky is blue, however a psychological mechanism tends to suppress it. Therefore, to prove such statements, all you have to do is to point them out and people will more or less instantly recognize their own experiences in it. It requires neither empirical nor logical evidence, just pointing out which more or less automatically defeats the psychological suppression mechanism and generates the insight.

    I named this third way to argue, beyond rigorous facts and logic, the Naked Emperor type argument: you just point it out and people realize they were duping themselves. But is this actually a valid way to argue?

    Of course it must be limited to things people have a lot of experience with. Human body, or human psychology. I use the Naked Emperor for only four insights so far: that humans like to belong to groups, that humans like to have high status both individually inside their group and for their own group compared to other groups, that morality is mostly a status-driven emotional reaction, such as an internalizing shame or praise, and that these three are so important they explain basically about half of everything you want to know e.g. about the dynamics of history, for example it perfectly explains something like religious wars.

    It is a very special tool not something to be used lightly, but only as an exception. In fact I don’t even find an example where I could use it to state something about the human body. Only about the human mind and emotions. It seems our minds mostly tend to suppress truths about themselves, so it is connecting some dots specifically in the field of psychology is where plenty of Naked Emperors abound.

    1. >But is this actually a valid way to argue?

      Oh, hell no. Though it’s good way to direct peoples’ attention to sound arguments, which is almost as important,

  29. @ TheDividualist – “But is this actually a valid way to argue?”

    Eye of the beholder. Lack of evidence, facts, or root logic is a non-trivial deficiency. In essence, you are making an argument asserting that the human species has a common recognition of the cited attributes, which then imbues them with some universal truth. By this standard of measure, I suspect that belief in a deity is more commonly recognized that any of the things you cite. Most people don’t cogitate about group affinity, prestige, morality as emotion, or historical analysis.

  30. As promised, I’ve been looking for more examples of phenomena that may involve cryptotheories, or at least mesotheories. These came to mind:
    – the Sapir-Whorf effect;
    – neurotypicals’ ability to pick up social and emotional cues;
    – our visceral responses to music (without music-appreciation training, which gives you an acrotheory);
    – and the medium’s effect on our processing of the message, provided that McLuhan was right and that I’m not misconstruing his thesis.

  31. @TomA

    > I suspect that belief in a deity is more commonly recognized that any of the things you cite.

    Yes, and that is an about medium strength evidence – not for its existence but for the bio-psychological need/utility/mechanism for that kind of belief. This is roughly why this kind of thinking is mostly useful in psychological matters.

  32. > and the medium’s effect on our processing of the message, provided that McLuhan was right and that I’m not misconstruing his thesis.

    It’s a good thing he’s dead, or he’d tell you you know nothing of his work.

  33. @ EMF

    > It’s a good thing [McLuhan]’s dead, or he’d tell you you know nothing of his work.

    You’re probably right, for I haven’t read any of his books. But that’s because you only gave me books with the word “death” in the titles. :-P

  34. An adaptation is only as homomorphic to a theory as it is homomorphic to a function. A theory is a prescient presumption; whereas, adaptation is a non-terminating process of species fitness.

Leave a comment

Your email address will not be published. Required fields are marked *