The methods of modern research involves dissecting and focusing in on finer and finer details. We would be forever blind to these finer details if it weren’t for instruments such as the microscope and the telescope. These tools allow specialists to focus in on the parts and gain a tremendous amount of knowledge in narrow fields.
But if researchers don’t look beyond these isolated islands of existence they will settle for a fragmented view of reality. And this causes problems for building theories about development and evolution of life. Researchers begin by looking at the parts to try to understand how they “build” bodies. Viewing things from this perspective it was expected that humans would have many more genes than turned out to be the case.. This is the type of error produced by this way of thinking Initially they did not understand the way in which the organism used its genes because they approached it from the wrong direction. Genes are in reality never isolated from the context of networks, cells and organisms.
Jaap van der Wal argues that we have become accustomed to thinking the human organism is made by a process of cells multiplication. But there is another more realistic way of thinking about it. From conception to adulthood a human being has always been a complete organism with a form and function suited to its environment. A machine is assembled from parts and it can only function as intended when all the parts are in place. Organisms are not like this. Where the organism is concerned the cell or cells of which it is composed serve the whole organism throughout its existence. It is not gradually built from parts. Machines are always built from the parts to the whole but organisms are never anything but complete wholes.
It is time to start paying more attention to how the whole determines the parts within it and luckily this view is becoming more prevalent.
Why?
I see that as almost certainly wrong. I see it as the conclusion one might reach by starting with a representational account of perception, and assuming computationalism. But direct perception is supposed to give a different picture.
Imagine a leopard, chasing down an antelope. As the leopard runs in hot pursuit, the visual information is jumping around all over the place. So you want the leopard to build a structured panoramic picture of the environment, and from that to compute down to the antelope. If the leopard does that, then it will go hungry. It will fail to capture the antelope.
What the leopard needs, instead, is to concentrate on the antelope. It can mostly ignore the larger environment, except for avoiding running into an obstacle. It does not need a structured view of the environment. It just needs to see and chase the antelope.
According to Marr’s theory of vision, we build that structured panoramic view of the environment. And then we compute down to identify objects. That’s probably what the Uber self-driving vehicle was doing, when it struck and killed a pedestrian.
Why Uber’s self-driving car killed a pedestrian
If, instead, we go by Gibson’s theory of perception, then we start by identifying objects in our visual field. To the extent that we can have a panoramic view of the environment, we build that out of the objects that we have separately identified. If the Uber car had been using Gibson’s approach, maybe that accident could have been averted.
In terms of the philosophy (or Sellars philosophy), Marr’s theory of vision tells us how to build the scientific image. And Gibson’s theory tells us how to build the manifest image.
Our current robots can implement Marr’s theory. But, as far as I know, they are unable to implement Gibson’s theory.
By contrast, a biological organism such as an animal (cat, leopard, human) can presumably implement something like Gibson’s theory. But it is not capable of implementing Marr’s theory.
And, by the way, this is all related to why I favor direct perception and why I am opposed to computationalism.
Neil Rickert,
I disagree with two major parts of that comment: (1) that Gibson’s approach to perception is part of the manifest image, whereas Marr’s approach to perception is part of the scientific image and (2) that Gibson and Marr are incompatible, or at least if you have one you don’t need the other.
On the first point: a central claim of ecological psychology is that animals directly perceive affordances in their environments, but that doesn’t means that affordances aren’t posits. The ecological psychologist posits the existence of relations — affordances — as what animals are directly perceptually aware of. But humans are rarely (if ever) able to report the affordances that they perceive or describe them. In this sense affordances are epistemologically unobservable — an important point made by Chemero.
Although there’s a lot of literature that treats Merleau-Ponty and Gibson as basically talking about the same thing, I think that this is a serious mistake. Merleau-Ponty is doing phenomenology: he is describing the structure of perceptual consciousness for a self-conscious subject. Gibson isn’t doing that; he’s explaining what an animal has to be able to do in order to use optical information to navigate its environment. In other words Gibson is a theorist of the scientific image, not the manifest image (if we’re going to use Sellarsian terms).
On the second point, I admit that I’m underwhelmed by the contrast between Gibsonian ecological psychological arguments for direct realism and Marr-style computationalism. I know that this contrast matters a great deal to lots of really smart people in philosophy of cognitive science — Evan Thompson, Daniel Hutto, Shaun Gallagher, Tony Chemero, Mark Rowlands — and maybe I’m just kind of dumb, but I don’t see it.
Let me see if I can articulate why.
The ecological psychologist is asking “what kind of information do animals need to be able to detect in order to navigate their environments?” On this view, the focus is on information out there in the world — the structure of the optical array, for example — and the animal doesn’t need to do anything to the information. It’s just there — as ecological information — and the animal just needs to know how to use it.
The computationalist is asking, “what are brains doing with information in order to systematically relate sensory inputs with motor outputs?” On this view, the focus is on the information inside the cognitive mechanisms (whether neurobiological or “artificial”), and the task is to specify the kind of computational architecture that might plausibly relate inputs to outputs sufficient to generate observed behaviors.
It is this difference in level — that of the whole animal embedded in its environment vs that of the brain — that makes me think that the debate between ecological psychology and computationalism relies on a category mistake: the mistake of failing to distinguish between the personal and the subpersonal.
Kantian Naturalist,
As so often happens, we are talking past one another.
I see you as disagreeing with things that I never actually said. And I see you as not even mentioning the important things that I did say.
I never actually said that. But perhaps I was not sufficiently clear.
My point was that if you manage to build a perceptual system based on Gibson, that will provide something like the manifest image. And if you manage to build a percepual system based on Marr’s methods, that will provide something more like the scientific image.
I agree that Gibson was not addressing the manifest image, and Marr was not addressing the scientific image. I was trying to make a point about how the two approaches differ.
I didn’t say that either, though perhaps they are incompatible.
To a first approximation, Marr is giving a mechanistic account of perception while Gibson is giving a teleological account. Robots are mechanistic devices, so Marr’s mechanistic approach fits rather well. Biological organisms (such as animals) are more teleological, so Gibson’s account is a better fit.
This is unsurprising, so I don’t see it as important. Affordances are going to be very subjective things, which is why we cannot talk about them. But they do give us the ability to perceive objects, which we can talk about.
I don’t agree with that. But I won’t further comment on that point for now.
This is where I disagree. It’s where I even disagree with Gibson.
There isn’t any information out there.
Well, of course, there are radio signals carrying information, and similar. But if we look at only what nature provides, there isn’t any information out there. That is to say, information is not the name of a natural kind.
This is a very common mistake that almost everybody is making. It is most serious for the computationalists. For the computationalist, the term “information” should only be applied to something that we can compute with. And nature is not providing anything that would count.
I’m less critical of Gibson, because he was not looking at information in terms of computation.
A perceptual system is a creator of information. We see information everywhere, because we are using our perceptual systems to do the seeing. But the information that we see is not out there in the world. It is in here, as constructed by our perceptual systems.
By no means am I saying he is wrong.
Only that that I suspect that book is a popularization of his ideas, and popularizations spend time on anecdotes and time on explaining basic ideas in the field. I don’t want to spend time wading through those parts of the book.
I’d pay for a service that summarized the key ideas in these technical popularizations. But the summarization services out there focus on self help and business theories.
Math information theory is a tool used to build models to describe postulated correlations between neural structures and the aspects of the environment an organism can sense. The scientific process then evaluates the success of those models. As far as science goes, claiming information is found in those neural structures or in the environment is just a useful short way for describing the process of building and testing those models.
Whether information is real is a philosophical question about scientific realism. Here Scott A claims that if we treat energy as real, information deserves to be so treated as well. I’ve seen Sean C make a similar claim. However, it is important to say both these claims are based on information as used in the theories of fundamental physics.
https://www.scottaaronson.com/blog/?p=3327
Modern computationalism goes well beyond the GOFAI approaches involving abstract symbols and traditional GOFAI processing using them.
It will take me quite a long time to study all the arguments here, to follow the links, and to try to clarify in my mind what people are saying. But its the only way I can try to figure out where my thinking is going wrong and where I am correct in what I say.
So I might respond quickly here and there as the arguments progress but I am going to take my time to reply to some of the comments that will need a lot more time and effort on my part to do them justice.
One area where I do think I am correct: Perceived entities are received by us as disconnected, reduced, chaotic inputs, and by means of gaining concepts we make the connections which piece by piece brings the wholeness of reality back for our understanding. What we perceive is individual and personal but our concepts are shared.
A lecturer might draw a figure on the board and ask the students to write down what figure they see. If they all write, “triangle” it is because they all share the concept triangle.
No. That would be because they all share the word “triangle”. We share words. We do not share concepts.
A correlation is not information. And that’s part of my point.
Information is something that you can just pick up and use.
If all you have is a correlation, then you have some work to do before you have information. Doing that work is what I am describing as “creating information.”
If you ignore the importance of doing that work, then you are not going to be able to understand consciousness.
I see it more as a question about the meaning of “information”.
I was using information in the Shannon sense, esp mutual information. But admittedly that’s based on only very limited view of scientific literature.
As best I can tell, you are using the term in some sense of your own creation. Which would not be surprising to me.
Which relates to the carrying capacity of telephone lines. Best to qualify with an adjective. Examples: useful, new, active, functional, complex specified!
Shannon information is information in a communication channel. It is not correlations found in the real world.
But, of course, there appear to be multiple meanings of “Shannon information”. There’s the one used by Shannon in his theory of communication. And then there are the meanings based on taking a remark made by Shannon, and using that to construe “Shannon information” as something quite different from what Shannon intended.
That’s one approach. OTOH, there is science
https://www.eneuro.org/content/5/3/ENEURO.0052-18.2018
BruceS,
It appears there’s a difference between information capacity (Shannon) and content (useful in some situations).
None of those adjectives have much to do with science or math. Quantum, Shannon, KC have the math behind them that science needs.
See my link in post to Neil for the further info (colloquial sense!).
The math still has to model reality to be useful. Ignoring content is not going to give us much insight into how brains work.
BruceS,
I followed the link and read through the parts referring to definitions. Is there an extract you think is pertinent?
Certainly not complex specified information! 🙂
The scope of the math models presented and the range of applications in the list of scientific papers cited. That is what matters. Not blog posts.
ETA: Also, to really engage with the science, you have to learn the math and the details of the models it is used in. Not argue by dictionary or rudimentary logic.
Congratulations on finishing the book.
Maybe if you did an OP on it on TSZ, some lurking publisher would read it and contact you?
Bit of a long shot, I admit.
Upthread, I mentioned a list of sciences for studying concepts in various guises. Looks like I may have forgotten one: machine behavior
“This Review frames and surveys the emerging interdisciplinary field of machine behaviour: the scientific study of behaviour exhibited by intelligent machines”
https://www.nature.com/articles/s41586-019-1138-y
Also
https://www.quantamagazine.org/iyad-rahwan-is-the-anthropologist-of-artificial-intelligence-20190826
Have at ’em, Neil.
Which model? Maybe I can adapt my rudimentary understanding to follow well enough. Though a holistic approach would be best but maybe start with a simple model. Pick something between run&tumble and human imagination. (Though Fox’s supposition rules out our ability to understand ourselves.)
At one time, I believed that a semantic notion of information was required to understand human cognition. I eventually came to recognize that I was wrong about that.
The semantics is not in the information. The semantics is in the procedures that you follow to construct the information.
Can you tell me what about the concept triangle you believe to be personal to you?
No, I can’t. And that’s already an indicator that we don’t share concepts. It is not at all clear what we even mean by “concept”.
I have a word “triangle”. And I use it in various ways, sometimes privately and sometimes in discussion with others.
CharlieM, to Neil:
Neil:
No, it’s an indication that your concept of triangle doesn’t differ, in any way that you can name, from everyone else’s.
You attach a meaning to that word — a notion of what a triangle is. It’s your concept of triangle. I’m willing to bet that it includes things like “has three sides”, and that it enables you to recognize triangles when they are drawn on a whiteboard.
You wouldn’t be able to do those things if the concept weren’t attached to the word.
So the words have the same function in different languages, they signify or point to something specific. If I asked you what you think is meant by the concept “triangle” what would your answer be?
For any of us to gain knowledge we must gather concepts and link them into a consistent, meaningful whole. A young child who has just been taught the basics of geometry will have the concept of a triangle being a three-edged two dimensional figure on a flat plane. They may then be taught about isosceles and equilateral triangles, the concepts of which fit within the more inclusive concept triangle. The teacher will be happy when the class all get and share the same concept of the relevant figures. Any triangle that is presented to the class is not separate from the concept. The concept and the individual item belong together and form a unity. It was never any different. The only thing that has changed is that the children now have knowledge of this unity, they understand the concept. They see a triangle and recognise it from the concept they hold.
By gaining the concept of non-Euclidean geometry you have increased in knowledge from the stage reached by the group of children mentioned above. You have added the concepts curvature and projection to the concept you held of the triangle. Basically you have introduced the concept of extra dimensions and how they relate to the two-dimensional space of the triangle.
If you did not have the original concept of the triangle you would not be able to understand how it changes by projecting it onto a curved surface.
Would it also be contradictory to talk about the representation of a triangle?
Well maybe, but none of that addresses the need to distinguish concepts from correct concepts and the consequent need to understand the nature of correctness. That is what I was trying to emphasize.
In any event, I’ll wait for further reply until you look at these three posts: how my concepts differ from yours, the long post which gives details on the 4 issues in your posts as I understand them, and the third, following post which tries to capture how I would word the point I think you are trying to make.
BTW, I agree with Keith that concepts are different from the words used to name them. That is an issue I neglected to raise in my long post.
Concepts are never isolated. When I see or hear the word “Pegasus” it invokes certain concepts in my conscious mind – horse. wings, flight, mythology, white, Ancient Greece, mythological beast, constellation, square, freight company; to name but a few. Some of these concepts relate to visible objects some don’t.
I know that the concept “Pegasus” would have meant something different in the consciousness of an ancient Greek to what it means to me with my modern Western educated consciousness. The consciousness of those ancient people were far more poetic and pictorial than our modern analytical minds.
The concept triangle has an ideal form that everyone should be able to agree on unlike that of Pegasus which is much more subtle and open to interpretation. There is a simple set of rules which defines the triangle, it is not so for Pegasus.
The concept “non-Euclidean triangle” is understood by the way it is related to and departs from the concept triangle which the majority of people have in mind.
I’m having a think about the concepts of abstract triangles versus abstract triangles themselves and I’ll get back to you when I’ve had more time to mull it over.
Bring in “understanding” adds more complications which I will avoid. You’ve also used “grasping” in other posts which to me connotes something similar and I also think should be put aside for now.
If one considers concepts as mental representations, then the idea is that the representation differs from what is represented. There are many maps of say the city of Toronto, but they are all differ from the content of the map, which is city of Toronto itself (according to how I was using ‘content’ in long post).
I take the maps as corresponding to concepts, that is how every person can have their own.
I’ve avoided correctness which complicates the above (since someone could have content which is incorrect). But that is enough for now.
Another complication I avoided is a type versus token distinction for concepts. I have assumed concepts are tokens — particular instances of mental representations. I think that is consistent with most of the posts in the thread.
And here is where our main difference lies. I do not consider concepts as mental representations. I consider the concept of a triangle to be the lawful aspect of any triangle that is perceived and as such it belongs with the perceived triangle regardless of my thoughts about it. Any mental representations we have are perceptible entities, we see them with the mind’s eye as opposed to observing with our physical eyes. The concept does not represent the triangle it completes it. Take the summer triangle of the Northern hemisphere which is made up of three stars. For our sight this is a very incomplete triangle but by applying the laws of the triangle to these three points allows us to understand their triangular relationship.
With the excerpt which follows Steiner asks us to think about what we are doing when we bring the concepts “thunder and lightning” into our minds:
Here we see that thunder and lightning are not separate entities although we perceive them as such. The concepts give us a fuller idea. They are both aspects of the same event. The sound of thunder is how the event is transmitted to our consciousness through hearing and the flash of lightning is how the event is transmitted to our consciousness through sight.
An area of land is complete in itself no matter how many maps are drawn. But our perceptions are incomplete until we have added the concepts to them.
It is we ourselves who, through our perceptions, tear apart what is in reality a whole, and it is through our concepts that we get the wholeness back.
OK, thanks. I am not interesting in discussing Steiner, so a la prochaine.
Neither am I. But I am interested in what he writes about epistemology in relation to thinking, observation and concepts in his books, The Philosophy of Freedom and Truth and Knowledge.
Okay. À bientôt
However, careful I tried to be about concepts and language, I am definitely careless about accents.
I think that this is not quite the standard use of “content” among philosophers. You are using “content” to mean referent — what the representation “points to” in the world as what is represented. Under that use of “content,” my thought of Pegaus is contentless, since there is no such animal. But then what I am thinking about?
One can see here why it is so difficult to get a firm handle on intentionality or aboutness, since clearly we can we have thoughts about non-existent objects. And that makes intentionality a very peculiar sort of relationship: one of the relata does not need to exist, in some sense of exist, in order for the relation to exist. This is where all the puzzles of Meinong and Husserl kick in with a terrible vengeance.
In most contexts the term ‘content’ goes together with the idea of a vehicle for that content. The vehicle for my representation of Pegasus is going to be different from yours because our brains are wired in somewhat different ways. (For one thing, my mental image of Pegasus has been decisively shaped by the picture on the thermos bottle that went with my Clash of the Titans lunch-box that I had a small child in the 1980s; see here.)
So while we differ in vehicle (at least at a neuronal level), we nevertheless take ourselves to have the some content in our thoughts at a degree of generality
To some extent, sure. I mean, whether cognitive maps are individual or social depends on all sorts of factors, and my guess is that the degree of individuality or sociality will depend on where the map lies within the overarching cortical hierarchy. The maps that are much more closely tied to my individual perception and action will be individualized to me — e.g. my map of my route to work, or my map of the layout of my apartment. But this doesn’t preclude us from having shared maps. I tend to think that the main function of language is to facilitate the construction of shared cognitive maps.
That’s a good way of looking at it. More briefly, language is thought sharing.
Thanks for the clarifying thoughts on content. I did not think content had to exist, eg my examples of Pegasus upthread, but I agree I was not careful in my description of it, eg in equating it to referring.
On vehicles: I had a footnote in an draft of an upthread post referring to that omission and simplification in my discussion, but decided to omit ithe footnote and the point because the post was more than enough already.
I agree that language is a tool which can align the content of concepts in the language community. That’s is what I was trying to get at by noting up thread the enforcement of norms for correctness being a social activity.
In terms of convincing CharlieM, my posts have been in vain.
He is is starting from a different worldview, one that sees a role for Steiner in issues I leave to the cognitive sciences and analytic philosophy.
But it was helpful for me to try to formulate my thoughts. Getting feedback from someone knowledgeable is a nice bonus.
If only sophistical was still around to point out the errors in my QM posts to J-Mac in another thread. J-Mac has tried to do that, but for some reason I do not find him convincing.
It may be to some extent the building blocks of cognition: https://en.wikipedia.org/wiki/Linguistic_relativity
Well I cheated. I just copied and pasted and trusted that the person who wrote the original phrase got it right 🙂
Language and thinking are intimately connected and Owen Barfield spent a lot of time and effort studying how language developed from before the rise of ancient Greece up to our present age. By studying language he traced the evolution of human consciousness.
From Wikipedia, Owen Barfield:
We have developed from a time where our concepts were more unified to our current position in which we see things in finer and finer detail. IMO the time has come to reunite our concepts only now with a more self conscious understanding of reality. Barfield used the terms “original participation” and “final participation” for these past and future states of how we act in the world. At the present time we have what he described as an “onlooker consciousness”.
Barfield shows us what the concept of Pegasus means to him with a 2720 line poem, Riders on Pegasus”, two of the verses I have copied below:
Language is not just a series of tokens for entities perceived by the senses. It is a medium which the composer can use to create emotive and inspiring scenes in the minds of its listeners or readers.
The reason that I have not engaged your arguments in detail here is that I wished to concentrate on the concept triangle which is a very straightforward, simple subject compared to dealing with how concepts and ideas sit within a cultural framework. I was trying to get an understanding of how the participants here as individuals see the relationship between their perception of triangles with the concept triangle.
Regarding the “onlooker consciousness” I mentioned above I’d like to refer to Paul VanderKlay.who refers to Revd. Malcolm Guite who refers to Owen Barfield 🙂
VanderKlay has published a long (about 2 hours) video,
Owen Barfield’s Meaning Crisis and the Evolution of Consciousness. It contains lots of links to other sources. One of these sources is a talk by Malcolm Guite , Owen Barfield: Knowledge, Poetry & Consciousness. At about i hour in the Vanderklay video and 20 minutes in Guite’s video, referring to the onlooker consciousness of the modern scientist, Guite says:
We must become participants again instead of being mere onlookers.
And one becomes a participant again by readng sappy poems about fantasy horses on the internet, instead of actually getting your hands dirty. Did I get that right?
Sorry I made a mess of the second link. Here’s the link
‘fraid not! When it came to the natural world Goethe certainly got his hands dirty.
You imply that reading poetry and studying nature are mutually exclusive. Why would this be the case?
That sounds like what you would get when science is criticized by somebody who has no understanding of science or of scientists.
More Than a Scaffold: Language is a Neuroenhancement
On this view, language doesn’t just give us an ability to make our thoughts public but it re-shapes the kinds of thoughts that one can have. That seems right to me.
I did no such thing. You approvingly cited Malcolm Guite scolding “the modern scientist” for “losing grasp of the meaning of nature and himself” while praising him and Owen Barfield, neither of whom have been active in the natural sciences. Do you really expect them to have the authority to lecture others on what meaning can be found in nature?
It was me who used the term “modern scientist” whereas Guite quoted Barfield as using the term “scientific man”. I used the term expecting to get some flak for seeming to criticise scientists and generalising them in this way. I was curious to see how many responses I would get for this and what would be the level of outrage at my words. I believe Barfield used the term “scientific man” to generalise an outlook which is typical of the vast majority of us in the Western world. He was not criticising scientists in particular although scientists are among the most knowledgeable among us. Guite says in the video that in the present time we are swamped by information, but to organise this information knowledge is needed and to deal with this knowledge wisdom is needed. This is where the participation comes in.
Science doesn’t make meaning go away. If anything, it enriches meaning.
If you look at science as just providing mechanical technological toys, perhaps that could result in a loss of meaning. But there are many other ways of becoming detached from reality, and most of those don’t involve science.