What do you think?

[coberst]

What do you think?

Mammals evolved on this planet about 200 million years ago. One type of mammal, the hominid, began using audible signals to convey meaning about 4 million years ago. Language, as we comprehend that word, began much less than 4 million years ago.

What is thought? The dictionary gives us various definitions of thought; I would guess that it is accurate to say that the actions of neural networks that control our sensorimotor actions can be regarded as thought. In other words, such things as memory, control of movements, and processing of sense inputs are all a process of thinking. Thinking produces thoughts. Thinking goes on all the time even while we sleep.

I guess that we will agree that all mammals had to have the ability to think. This leads to the conclusion that thinking was been happening on this planet at least 200 million years before human language existed on this planet.

Those individuals who accept the science of evolution must then conclude that humans may think in linguistic forms some small percentage of the time but that most thought is not in linguistic form.

“It is the rule of thumb among cognitive scientists that unconscious thought is 95 percent of all thought—and that may be a serious underestimates.”

What does all this mean to you? It means that most of the things that you think are true about thinking are pure non-sense. This also applies to many of the things we all believe that are based upon the philosophical attitudes that fills our life are like wise pure non-sense.

Quotes from “Philosophy in the Flesh”—Lakoff and Johnson

[Starglider]

Lakoff and Johnson fell into the common cogsci trap of doing good work in one relatively limited field ('Women, Fire and Dangerous Things' is pretty good if a bit anthropological, 'Metaphors We Live By' somewhat less so, it's just a big mostly-unstructured list of metaphors), proclaiming that they had the answer to all cognition, getting sidetracked into philosophy, realising they did not in fact have the answer to all cognition, then writing a tortuous semi-philosophical tract trying to package this as a great insight. Could've been worse I suppose, could've turned out like Jerry Fodor.

The cognitive hardware for human language use (of which there is quite a bit of dedicated wetware) evolved to codify, compress, transmit, decompress, contextualise and to a certain extent analyse models of the world. It added a lot of capabilities (as did other-entity social modelling, which also has dedicated wetware and evolved in parallel) that later got repurposed to produce major enhancements to general cognition. Human language use and conscious thought in general purchases generality and flexibility at a massive cost in terms of loss of parallelism, and to a certain extent serial speed, compared to unconscious (which is to say, not reflectively accessible even to the limited capabilities of human reflection) thought. One of the many cool things about artificial substrates is that they don't have to make this tradeoff, or at least that it is nowhere near as severe, essentially due to greater plasticity and much easier storage.

If you undertake an intensive course in cognitive science you will probably find yourself noticing lots of bizarre and unsettling things about your own cognition - I did and most of my friends with strong cogsci backgrounds have mentioned it at some point. Even that's just scraping away the surface layer of the simplifications and papering-over-cracks that the human self-model does; our internal/reflective relevance filtering, desensitisation and normalcy adjustment is at least as ruthless as the corresponding processes acting on external perception.

[coberst]

Starglider

I have not read "Women, Fire...." but have spent a good deal of time with "Philosophy in the Flesh". I think that you might modify your judgment of their work if you read this work.

[Starglider]

I have not read "Women, Fire...." but have spent a good deal of time with "Philosophy in the Flesh". I think that you might modify your judgment of their work if you read this work.

Read it, of course, otherwise what basis would I have for this comment? It's sitting right next to me on the dedicated cogsci/AI bookcase.

[Darth Wong]

This isn't a precursor to some bizarre attempt to prove that human logical constructs are illogical because humans are irrational by nature, is it?

[Ar-Adunakhor]

What is the question here?

[Patrick Degan]

What is the question here?

"Just how far up their asses do these two guys have their heads?"

[Hawkwings]

How do they get "most thought is non-linguistic" from "mammals think"? Can you even compare the two kinds of "thought"? I would say that language is a prerequisite for complex thoughts that go beyond instincts, which animals have.

I don't know about the rest of you, but I'm fairly sure that most of my internal commentary is in language. Sure, it's a mishmash of two languages and shorthand (short-think?), but it's definitely language.

And what thoughts can you have that are not in language? Do emotions and instincts count as thought?

[Darth Wong]

I would love to know how they differentiate between linguistic thought and non-linguistic thought, or conscious thought and unconscious thought. I don't see how the brain allows for such clear delineations.

[Setesh]

I would love to know how they differentiate between linguistic thought and non-linguistic thought, or conscious thought and unconscious thought. I don't see how the brain allows for such clear delineations.

It doesn't really, since we can't ask animals that don't have linguistic cognition as a test case its entirely subjective. Can you hold an image thought in you head without words to go with it? Would you even notice a differance or are our brains built like a computer where data is data.

What these guys are doing is setup for one side of a philosophical question: Which comes first the idea or the words to describe it? This sounds a great deal like the 'Idea' side, who tend towards the thought that language evolved almost entirely to give voice to abstract ideas. The other side says that you have to have language in order to have abstract ideas.

[coberst]

Categorization, the first level of abstraction from “Reality” is our first level of conceptualization and thus of knowing. Seeing is a process that includes categorization, we see something as an interaction between the seer and what is seen. “Seeing typically involves categorization.”

Our categories are what we consider to be real in the world: tree, rock, animal…Our concepts are what we use to structure our reasoning about these categories. Concepts are neural structures that are the fundamental means by which we reason about categories.

Human categories, the stuff of experience, are reasoned about in many different ways. These differing ways of reasoning, these different conceptualizations, are called prototypes and represent the second level of conceptualization

Typical-case prototype conceptualization modes are “used in drawing inferences about category members in the absence of any special contextual information. Ideal-case prototypes allow us to evaluate category members relative to some conceptual standard…Social stereotypes are used to make snap judgments…Salient exemplars (well-known examples) are used for making probability judgments…Reasoning with prototypes is, indeed, so common that it is inconceivable that we could function for long without them.”

When we conceptualize categories in this fashion we often envision them using spatial metaphors. Spatial relation metaphors form the heart of our ability to perceive, conceive, and to move about in space. We unconsciously form spatial relation contexts for entities: ‘in’, ‘on’, ‘about’, ‘across from’ some other entity are common relationships that make it possible for us to function in our normal manner.

When we perceive a black cat and do not wish to cross its path our imagination conceives container shapes such that we do not penetrate the container space occupied by the cat at some time in its journey. We function in space and the container schema is a normal means we have for reasoning about action in space. Such imaginings are not conscious but most of our perception and conception is an automatic unconscious force for functioning in the world.

Our manner of using language to explain experience provides us with an insight into our cognitive structuring process. Perceptual cues are mapped onto cognitive spaces wherein a representation of the experience is structured onto our spatial-relation contour. There is no direct connection between perception and language.

The claim of cognitive science is “that the very properties of concepts are created as a result of the way the brain and the body are structured and the way they function in interpersonal relations and in the physical world.”

Quotes from “Philosophy in the Flesh”—Lakoff and Johnson

[Starglider]

I would love to know how they differentiate between linguistic thought and non-linguistic thought, or conscious thought and unconscious thought. I don't see how the brain allows for such clear delineations.

'Linguistic thought' isn't as blurry as you might think. The brain is relatively modularised; when you think in terms of words, various bits of the left lobe (Broca's and Wernicke's areas in particular) light up. You can see this when you give someone a language puzzle when they're sitting in a functional MRI scanner. Conscious vs unconscious thought is more arbitrary. A huge amount of psychological work is based on just asking the subject 'what are you thinking' or 'how did you do that'. Supposedly, if something shows up consistently in these accounts it's conscious/reflectively accessible, otherwise it is unconscious. Unfortunately humans and particularly untrained humans suck at providing accurate accounts of our internal mental experiences; read through a few hundred of these (which most psych postgrads will) and you'll see people confabulating random crap left right and centre. This is why the physicalist movement in cogsci tries to ignore subjective accounts and concentrate on objective instruments (currently fMRI is producing a rich vein of results) as much as possible. Unfortunately we don't have a good physical definition of 'reflectively accessible' yes; it'll be something to do with which areas have reverse projections back to other areas, and what level of signal filtering goes on in various neural layers, but (AFAIK) neural mapping just hasn't progressed far enough to generate strong hypotheses about this yet. In abstract models and AI systems the distinction is much neater of course.

As for the OP, it sounds like Coberst read PITP and had lots of 'wow, huge insight, must share with everyone' moments. Lots of people get this when they read their first competent cogsci overview book, be it Pinker's 'How the Mind Works', Deacon's 'The Symbolic Species', Calvin's 'The Cerebral Symphony', Cotterill's 'Enchanted Looms', Bergland's 'The Fabric of Mind', Dennet's 'Consciousness Explained', even Wilson's 'Conscilience' or Hofstadter's 'Metamagical Themas' or 'GEB'. That's just rattling off some of the more famous ones I've read personally; there are thousands of these sort of 'cogsci overview/author's personal grand theory of mind' books, many of them by complete hacks. When you read the first one, if it's a good one, the novelty of the concepts (and maybe the field) can make it seem awesome regardless of how it stacks up compared to the others. By the time you get to the 20th or so book of this type (which I was at by the time I got to PITP) you've seen it all before, the basic stuff is obvious, the speculative stuff looks less impressive when it's as wildly speculative as the previous 20 books which didn't agree with each other or provide (currently) testable predictions, and you have to really dig to find any novel usable insights.

So coberst if you think PITP is the greatest cogsci book ever, be aware that you can't make a rational judgement of that unless you've read a ton of similar books. Copy-pasting sections of it here isn't likely to impress many people, because they're either not interested in the field, somewhat interested but not impressed without the full context, or experts in the field who find packaging a combination of basic facts and wild speculation as great insights annoying.

[coberst]

Starglider

You make some very valid points.

My first contact with CS was in the early sixties when I was a new engineer going to my first computer conference in San Francisco. At that time AI was the great news that had everyone jazzed up. Since then I have glanced through several books but became discouraged because there seemed to be no clear paradigm. It was like Kuhn commented about a science without a paradigm.

I must admit I am thundestruck by the conceptual metaphor. For the first time I see some info about human conception that makes sense to me.

As you say few people will display any knowledge or interest about what I say about this matter. However, they will have begun to see the words and the phrases that represent the first consciousness of a new and revolutionary idea. Every journey begins with that first step.

[Starglider]

became discouraged because there seemed to be no clear paradigm. It was like Kuhn commented about a science without a paradigm.

Lakoff and Johnson's stuff is nowhere near concrete enough to be a 'paradigm'. You can't build an AI to their plan and it doesn't serve as much of a guide for functional neurology either. The 'philosophy' part of the title is not their for fun. Generally, from the ultramaterialist (i.e. right) point of view, 'philosophy' is worthless (computational philosophy, a tiny subfield, may be an exception) while 'cognitive science' is a vital pursuit, providing a material bias to replace the last bastions of fuzzy dualist thinking. Books like this which straddle the more-concrete end of philosophy and the most abstract end of cognitive science require a bit more subtly to judge. IMHO (though it's been a couple of years since I read this) PITF is a combination of common insights you'll get from any decent cogsci overview book since the mid 90s, some moderately novel insights that were covered better in their earlier work anyway (where they stuck more to findings and less to overarching speculation) and a fair slice of pointless philosobabble. Overarching speculation into a pet model of the mind (not rigorously functionally verified or grounded) and attempts to suck up to philosophers and be an 'intellectually groundbreaking' book is pretty common for late-career cogsci people trying to write a magnum opus.

I must admit I am thundestruck by the conceptual metaphor. For the first time I see some info about human conception that makes sense to me.

Well, good, but that's only one insight out of many that various people have detailed, and only a tiny fraction of what is required to actually understand the mind (human or in general - still very much a work in progress).

However, they will have begun to see the words and the phrases that represent the first consciousness of a new and revolutionary idea. Every journey begins with that first step.

Maybe. I hear this argument a lot from cranks who turn up on various AI and cogsci forums (e.g. Mentifex, the archetypal offender) - actually you see this on maths and physics forums a lot too. Generally these short phrases are only imbued with massive insight in your own subjective context (this is a huge problem for cogsci writers too). They tend to bounce off uninterested people and may end up doing more harm than good (see the person above who wrote off Lakoff and Johnson as 'up their own ass' despite them actually being renowned and widely cited researchers with decades of useful work in the field).

[Darth Wong]

I would love to know how they differentiate between linguistic thought and non-linguistic thought, or conscious thought and unconscious thought. I don't see how the brain allows for such clear delineations.

'Linguistic thought' isn't as blurry as you might think. The brain is relatively modularised; when you think in terms of words, various bits of the left lobe (Broca's and Wernicke's areas in particular) light up. You can see this when you give someone a language puzzle when they're sitting in a functional MRI scanner. Conscious vs unconscious thought is more arbitrary. A huge amount of psychological work is based on just asking the subject 'what are you thinking' or 'how did you do that'. Supposedly, if something shows up consistently in these accounts it's conscious/reflectively accessible, otherwise it is unconscious.

Some of those definitions seem rather circular. For example, we decide that word puzzles are linguistic, and we monitor the kind of brain activities associated with those thoughts. But how many of our patterns of thought have been laid down due to our lifetime use of language, and what effect do those mental habits and patterns have on other kinds of thoughts? Do we account for that kind of linguistic connection to thought? If so, how? And how do we measure its influence?

[Molyneux]

Granted that I'm not exactly well-educated in cognitive sciences...but from the OP, it seemed to me like they were redefining the word 'thought', and then calling attention to that new definition as if it were some groundshaking truth.

[Starglider]

But how many of our patterns of thought have been laid down due to our lifetime use of language, and what effect do those mental habits and patterns have on other kinds of thoughts? Do we account for that kind of linguistic connection to thought? If so, how? And how do we measure its influence?

I'm not sure precisely what you mean by 'patterns of thought', but it's a difficult area to be precise in. The words themselves are just handles for concepts; the phonetic details have a real but pretty limited effect. But the extreme chunking that language use forces (speech has very little bandwidth compared to the various senses or the brain's internal connections) has a huge effect. Much has been written about how limiting, expanding or influencing people's conceptual palette changes how they think, and word meanings are the 'user interface' to that. Unfortunately a key aspect of this is what happens when two templates are combined to get a composite model. For something like 'pink teapot' there are decent models (e.g. the one presented in Barsalou's 'Perceptual Symbol Systems' and later elaborations) based on back-projection on the visual processing chain (following on from the verbal recognition process where you identify the words) with massively parallel constraint satisfaction creating the composite model. It's not quite rigorous yet, but it's close to linking up with the neurology. Unfortunately for abstract stuff like 'compassionate conservative' that goes on in the frontal cortex, there are no real answers for how composite concepts get formed other than 'magic happens'. We know it's some sort of joint activation followed by massively parallel constraint satisfaction, with some kind of focused attention and level of detail control, but all we have on the details are a series of highly speculative (and incompatible models), plus some experimental AI designs that may or may not resemble how the brain does it. Until we understand how complex concepts combine (particularly under linguistic invocation) there won't be definitive answers to this.

That said most of the metaphors Lakoff and Johnson go on about aren't really verbal ones at core (in that they don't need 'verbal thought' to work, and many of them are language and culture independent - though their anthropological approach seems to result in a focus on the minority that do vary by culture and language). There is good evidence that much abstract logic application (formal and informal) is based on hardware developed for processing grammar rules, with some contributions from social modelling wetware when it comes to things like interpreting laws and 'fair' division of spoils. Kahneman and Tversky wrote numerous excellent papers (later collected into a series of books) that detail how people use grammar and empathy wetware to reason about abstract problems, largely in the sense of how this results in pervasive human reasoning flaws, and how these can be avoided by recasting questions in a way that fits human intuitive reasoning.

The field is full of open questions and competing models (many badly defined, in both cases) from a wild assortment of academics. Unlike say physics there are no nice neat one-paragraph answers, concise formal models or textbooks you can point to and say 'the universe operates so, enlighten yourself'. At least not yet. We're working on it.

[Darth Wong]

I'm not sure precisely what you mean by 'patterns of thought'

I'm sure you're familiar with the fact that your brain physically rewires itself depending on what you tend to think about.

The field is full of open questions and competing models (many badly defined, in both cases) from a wild assortment of academics. Unlike say physics there are no nice neat one-paragraph answers, concise formal models or textbooks you can point to and say 'the universe operates so, enlighten yourself'. At least not yet. We're working on it.

Until then, why do you expect anyone to take the speculative work seriously? Do you know how much bullshit can be found from the more speculative developing period of any branch of science?

[Starglider]

I'm sure you're familiar with the fact that your brain physically rewires itself depending on what you tend to think about.

Sure, but the gulf between knowing that neurons form and break synaptic connections depending on the patterns of activation their local network gets and ultra-abstract stuff like 'word meanings influence how people think' is a huge one. Right now only speculative models bridge that gap and even they do so with several intermediate layers of organisation (ultra-simple stuff along the lines of 'one neuron per word' were discredited in the 1960s)

Until then, why do you expect anyone to take the speculative work seriously?

I don't, but I didn't start the thread.

Do you know how much bullshit can be found from the more speculative developing period of any branch of science?

It runs into exactly the same problem you get with 100 different people screaming 'believe my religion or burn in hell' - at best only one right and really there's no reason to assign any of them a nontrivial probability of being entirely correct. Nowhere near as badly of course, though sadly a fair sized minority of cogsci academics do proclaim 'I have the answer!' and expect everyone to believe them, then write a big popsci book grubbing for popular appeal when no one does. From personal experience it seems to be considered rude to point this out - 'I've noticed some weaknesses in your model...' is ok but 'None of you agree. 9 out of 10 of you are spouting nonsense and I'm not sure I should be that generous' gets a torrent of complaint and if you're unlucky kicked off the forum.