Prototype theory of concept formation

Rand's theory of concept-formation is basically an iteration of the classical theory, a process of separating like objects apart from unlike objects based by reference to some definitive feature(s). In the 70s, Eleanor Rosch introduced an alternative theory dubbed the prototype theory, a process of comparing how similar objects are to to some definitive "core" object(s). Both seem plausible to me. Both hold different epistemic implications. I'm not sure what to make of this. Thoughts?

Jordan

(Edited by Jordan on 9/24, 12:00pm)

Post 1

Wednesday, September 23, 2009 - 6:19pm

Just on a quick read it seems partially plausible. I can imagine people utilizing a prototype process after concept formation, but I didn't see any theory of where the "prototype" comes from. Read kind of like a platonic ideal. The nativism paragraph seems to suggest that as well, without actually saying it.

Spoiler alert: I'm not a professional philosopher, nor do I play one on TV.

Post 2

Wednesday, September 23, 2009 - 6:59pm

nor do I play one on TV.
.............

Heh, heh - good one, Ryan... ;-)

Post 3

Wednesday, September 23, 2009 - 11:17pm

First, it's not new. It's a reworking of Greek skepticism and of Mediaeval nominalism. The Greeks skeptics said our ides were based on mere appearances. The Nominalists said there is no essence, just us naming similar things. Prototype theory is nothing other than an attempt to codify the perceptual bound mentality.

Prototype theory usually says something like this: "Think of a bird. I'll bet you thought of something that looks like a dove or some songbird, not an ostrich or a peacock or a penguin. That's because when we think of concepts, we think of a typical example of that concept, not an odd one. That's what a concept is. A concept is our notion of a typical example of a concept and the things that resemble it, more or less." Bull shit.

First, this is plausible only if we restrict ourselves to perceptual level concepts. Bird, tree, ball, face, table, shoe. How does it apply to the concept "yesterday"? Or "selfish"? Or "absolute"? Can you show me a typical "fraud"? This is an Orwellian attempt to define abstractions out of existence. It leads to skepticism in higher thought and the look-say method of teaching in elementary school.

Second, it steals the concept. It tries to define a concept as being the idea of a typical example of a concept, and all the things that more or less resemble it. See the circularity there? It assumes you already know what a concept is, what sort of things it subsumes. If you didn't already know what a bird was, how could you know what a typical bird is?

This sort of nonsense may describe the way a lot of people do actually think. Up to the third or fourth grade they do fine with perceptual level concepts. But they never learn to abstract. Doing multiplication is about the limit of their mental ability, since they can imagine a rectangular box of apples four rows and five columns deep, they can count them in their imagination (or on a piece of paper) and come up with twenty. Asking people like this whether some abstract notion follows, like, say, whether someone's advocating anarchism entails their advocating civil war, and they will get all defensive. Their brains hurt. Unless they've heard someone repeat a formula before, and learned which buzzwords go with which. Listen to a "typical" Democrat explain himself or a fratboy try to argue.

This desire to set up the sloppy way some people do happen to do things as an enshrinement for how they ought to do things is a widespread trend in everything from academics and morality to consumer product design. We have to make everything "user friendly." Every button in a car has to have some little picture. You don't have to speak and read English to operate a piece of machinery, you just have to know what the international symbol is. Expecting people to read a word on a button is just too much. There is no reason for rational egoist to countenance, tolerate or willingly accept this.

(Edited by Ted Keer on 9/24, 8:32am)

Post 4

Thursday, September 24, 2009 - 6:37am

Re:

Can you show me a typical "fraud"?

How about a recent Presidential campaign?

;-)

Re "the look-see method of education" -- Ted, I knew about the "look-say method" of teaching reading (which produces folks who can't read but who believe they can), but not about "look-see." Please explain what "look-see" involves, how it resembles (and/or differs from) "look-say," and what problems it creates.

Post 5

Thursday, September 24, 2009 - 12:11pm

Hi Ted,

What you're attacking is not protoype theory (PT). Under PT a concept is not a conceptualization or a "typical example of a concept and the things that resembles it." PT still countenances concepts as mental categories. It just asserts that those categories form around a core existent, rather than a defining characteristic. Or put otherwise, existents belong in a category based on their likeness to a particular existent, rather than based on their instantiation of a particular trait.

The category is built from a core existent; the core existent is not built from the category; there is no concept stealing or circularity here. This process of comparison to a core existent would apply to abstract concepts as well; it is not a percept-bound theory. And PT bears little to no resemblance to Greek phenomenology or nominalism although it does bear some resemblance to Plato's archetypes, which Ryan notes, except that PT is simply epistemic. Idealism is at base metaphysical.

Jordan

Post 6

Thursday, September 24, 2009 - 12:44pm

How would the "core existent" be identified for an individual? Would it just be the first existent meeting those criteria, which is then implanted in to the psyche? Kind of like a duck and its mother? How would corrections be made to a flawed impression of the core existent?

How can any theory be PURELY epistemic? How can a theory of how we learn and categorize be completely divorced from what we learn and categorize?

(Edited by Ryan Keith Roper on 9/24, 12:45pm)

Post 7

Thursday, September 24, 2009 - 1:35pm

Jordan, you are stealing the concept "core." Core is a relational concept. Either the "core" of a concept is determined in relation to the concept itself, which is prior, or it is determined in relation to what? Chance encounters with Ryan's ducks? Why would you light upon any particular existent as the core of a concept in this theory other than chance or whim?

And of course you still haven't told me how one picks the core of a concept which is not at the perceptual level. What is the core example of "yesterday"? This theory ignores that concepts are hierarchical, and can't deal with abstractions from abstractions.

I thought that the problems with this notion were obvious. I assume you are familiar with and can argue Rand's theory of concept formation. She doesn't waste time with dubious notions like core existents. She has a very simple and obvious two-part rule of thumb for when one does and does not form a concept. She has no problem dealing with concepts at different levels of the abstraction hierarchy. If you want to argue this, then I suggest that rather than tell me what I am and am not criticizing (and Ryan, the non-professional philosopher here has no problem seeing the holes I have pointed out) that instead you should present a criticism of Rand's theory from the standpoint of prototypicalism. My interest in answering your question was not to argue for Rand's theory, which I know is correct, but to explain the Randian criticism of this theory. To repeat, Rand would argue it fails because it begs the question (steals the concept) of what the core is, and it fails because there is no plausible way to apply it to abstractions like "yesterday" that are beyond the perceptual level.

Post 8

Thursday, September 24, 2009 - 1:54pm

Hi Ryan,

Good questions. From what I gather, an initial core existent is an existent that isn't easily sorted into preexisting categories because it's so different from all other existents. It's a novel stimuli; it's interference pattern in the brain is unique. So it gets mentally filed away in a place of its own. I suppose this is like the duck-to-mother imprinting, except other existents eventually get filed with that initial existent based on their likeness to it and their difference from others.

What serves as the core existent can change over time so as to allow more or fewer existents in its category (to correct flaws, as you put it) -- much like how definitions grow or shrink depending on what existents belong in a concept.

By calling PT simply epistemic, I didn't mean to suggest that it's divorced from exteral existence, only that it's silent as to metaphysics, like whether everything is mere appearance or whether abstracts exist externally.

Jordan

Post 9

Thursday, September 24, 2009 - 3:30pm

Good questions. From what I gather, an initial core existent is an existent that isn't easily sorted into preexisting categories because it's so different from all other existents. It's a novel stimuli; it's interference pattern in the brain is unique. So it gets mentally filed away in a place of its own. I suppose this is like the duck-to-mother imprinting, except other existents eventually get filed with that initial existent based on their likeness to it and their difference from others.

But the first stimuli a person encounters would be the most novel example of whatever it is, by default. So we have the duck issue again. The theorized filing away of other existents based on that initial impression would be catastrophic. The duck is in terrible enough danger as is, I can't imagine the failure cascade that would insue if the duck saw all creatures similar to the first one it sees as "mother". If any creature actually formed all of its concepts this way I can't see it ending in anything but a massive failure cascade and death.

What serves as the core existent can change over time so as to allow more or fewer existents in its category (to correct flaws, as you put it) -- much like how definitions grow or shrink depending on what existents belong in a concept.

I agree an effective theory of concept formation must allow for definitions to grow or shrink. I don't see how this theory does so. If duckman comes in contact with an existent that is a better exemplar of X concept than the first existent of X he made contact with, how would he know? How could he possibly determine that THIS thing is more X than the original exemplar he assigned to X? Prior knowledge? Prior knowledge of what? It seems to me that any existents encountered past the initial would have to be intepreted as FLAWED versions of the exemplar. That would be a best case, at worst the duckman would just conceptualize this new thing as different. I can't even imagine conceptualizing second or third order concepts by this method. Rand's theory (which I subscribe to) explains how a definition might grow or shrink, I can't get my head around how this theory could possibly do the same.

By calling PT simply epistemic, I didn't mean to suggest that it's divorced from exteral existence, only that it's silent as to metaphysics, like whether everything is mere appearance or whether abstracts exist externally.

I know that's what you're doing. I'm just saying that this theory does seem to hint at metaphysics a bit, in the inferred connections to possible prior knowledge. To me, this theory utilizes a sort of platonic ideal, but the originators correctly identified that there is no magic ideal realm. Instead it reads to me like they decided that the first X a person meets would set a personal platonic ideal in their mind. I realize the theorists probably didn't start at metaphysics, but since epistemology pretty much touches everything I can't see how they could help but end up there.

Post 10

Thursday, September 24, 2009 - 6:33pm

I run into this all the time at my job. As web designer/implementer for a huge website covering hundreds of products and presenting them from all kinds of angles to the potential buyer, as well as one of the major players in generating the tech manuals, the photos, the catalogs, etc., I have to try to maintain the ability to find and retrieve some photo or manual from 15+ years ago on a system with half a million files on it. I have to be able to switch jobs five or fifteen times in a morning and return to the earlier jobs in order of priority without losing focus.

It doesn't help that I've had several major systems crashes, with incomplete backups, or that management in its infinite paranoia has deliberately moved jobs to remote sites so that whatever version of a product manual or photo I have is probably not the most current. Or that Windose XP simply cannot handle that many files and frequently returns complete garbage on a simple search that Win2K had no problem with. I cope, not always happily. Invectives and cursing helps.

The point is that the world is chaotic and often perverse as well, and people cope with complexity by rule of thumb approximations which they attempt to scale to the current need on the fly. My job is a more extreme case than most people get to handle. But much of thought and attention is not of the form of creating a conceptual abstraction that will stand up to the ages, but rather getting just enough information to finish the current job, deal with the immediate brush fire.

That doesn't mean that precise, philosophically sound, explicit conceptualization is not important, and there are lots of jobs that require it. Just that most people, even those who deal with a universe of information all day long as our primary work, are flying by the seat of our pants. What Rand primarily focussed upon in her ITOE was the case of original concept formation in an ideal environment, where you did have the time to precisely tease out the proper defining characteristics of a set.

Other methods also work, without invalidating her approach at all. Forming a concept of all blood red objects might not seem very sound epistemologically, but there are jobs in graphic design that might require it. You might have a set graphic background for an ad that management has chosen in their infinite wisdom that is primarily shades of green. Now they want some textured red objects to offset and call out. So, you have your red pile of images on your hard drive and links on the web. You assign a database keyword or set aside a new directory called T-Reds. Effectively, you have formed a new concept of textured Red images.

Or, another example, in beginning algebra one of the first things you learn is that you have to have as many independent equations as you have variables if you expect to arrive at a solution space of a single or finite set of points. A similar criteria applies to general problem solving. In examining a specific issue, whether to construct a dam, a power company has to bring to bear a number of different dimensions of analysis, such as engineering, ecological, financial, legal, and economic.

The process of making a decision about whether to procede and how to procede will go through multiple interations from all these different perspectives until a clearly defined set of solution spaces has been elucidated. The dam might be perfectly feasible from the engineering dimension but cost too much, or it might ruin the watershed below the dam, or it might disrupt an entire ecosystem (e.g., Salmon runs) that impacted fishermen and ultimately the cost of food, thus posing the liklihood of lawsuits. There is generally no simple, clear, linear way to deal with such problems, other than to bring as many independent perspectives to bear as possible.

Post 11

Thursday, September 24, 2009 - 7:16pm

Deleted

Thursday, September 24, 2009 - 7:22pm

I'm not sure where you're going with that, Phil, but Rand's theory accounts for the ability to refine a concept and to use them for the sort of approximation you're describing. As far as I can see, the prototype theory doesn't. It's a theory of concept FORMATION as far as I can tell, not of concept UTILIZATION. Taking a concept, measurements omitted, and intentionally defining the working parameters broadly for ease of work is not analogous to taking a random concrete as an exemplar and somehow shoehorning similar concretes into that mold.

The red example you gave is simply combining three concepts into a larger concept of what you need for a job. If your goal is to choose the pics that work for your purpose you will refine that original concept of "what I need for this project" by tightening down on the acceptable measurements. If you "invented" a concept via prototype theory by combining those three concepts into a new fuzzy concept, you would be paralyzed. The blurry new concept would be your exemplar. At least that's how it seems to me.

Post 13

Thursday, September 24, 2009 - 7:27pm

Phil,

You mistake a proper (Randian) concept formation with a kind of fully-applied Scholasticism and it's inevitably-entailed intellectual asceticism (and uselessness) in actual thinking and actual living. You think that thinking so "properly" is a hindrance in the "real world" or in "real time." Your examples of color categories and all of those dam contingencies aren't good examples at all.

Your outlined philosophy is essentially a pragmatism, where the standard of what's true is what works (in limited time-frames, and limited contexts, and limited etc.'s ...).

Do you agree that that's accurate?

Ed

[edit: this cross-posted with Ryan's response above, and I agree with Ryan's evaluation of your color category example]

(Edited by Ed Thompson on 9/24, 7:30pm)

Post 14

Thursday, September 24, 2009 - 8:31pm

(Yes, Phil, both Ed and Ryan's comments are quite relevant and valid, and reflect the core exemplar of my own reaction to your comments, so that I would call them Keerian.)

Phil, you seem to think that there is a Randian argument against perceptual level concepts like "blood red objects." That is not the case. The concepts you form are based on your cognitive needs within the context of the realities you are dealing with, not on some criterion of scientific purity.

For example, there is a concept common among the world's languages that we might best in English call "critter." It includes small cold-blooded animals like insects, frogs and snakes. This is horribly wrong so far as biological taxonomy is concerned, but nevertheless as a lexical concept it is extremely common throughout the languages of the world. The cognitive basis for this concept is that such animals are small annoying animals that are much more likely to be pests than valuable food sources. (The concept also pretty much coincides with unclean animals in Jewish law. Consider the plagues called down upon Egypt by Moses. He did not summon lions.) And while small, cold-blooded creatures may not form a valid taxonomical group, ecologically they do exist on a similar trophic level in the food chain. Whatever is the case scientifically, such a concept is valid because it unites all the units under the contextually essential criteria.

Now, of course, this concept "critter" happens to wreak havoc with the prototype theory. Who in his right mind is going to argue that such critters as a lizard or a spider resemble each other more than a lizard resembles a crocodile or a spider resembles a lobster? Animals that eat or are eatne by man are excluded from the category. And again, we see the failure of the prototypical theory. What is the "typical" critter which all other critters resemble? A schmoo?
(Edited by Ted Keer on 9/25, 12:01am)

Post 15

Thursday, September 24, 2009 - 11:14pm

Re:

... what is the "typical" critter which all other critters resemble?

That question makes me wonder if "prototype theory" is just the latest synonym for Plato's theory (that all critters -- for example -- more or less resemble an "ideal" critter not locatable in reality).

Post 16

Thursday, September 24, 2009 - 11:44pm

Platonism and nominalism are the exact same epistemological theory. They differ in metaphysics. Plato posited a supernatural realm within which he placed the "prototypes" he "intuited" of his concepts. Nominalists do away with the other realm and admit that their intuitions are purely arbitrary.

Post 17

Friday, September 25, 2009 - 5:17am

There is work in this area in Section II of Merlin Jetton’s Pursuing Similarity (1998).

Outline

I. Philosophy: Realism, Nominalism, and Conceptualism
• Circa 450–300 B.C. (pp. 43–47)
• Circa A.D. 1050–1350 (47–51)
• 1690–1748 (52–59)
• 1929–51 (59–63)
• 1966–84 (63–73)

II. Cognitive Psychology
• Classical View of Concepts (pp.73–82)
• Prototype Theory of Concepts (83–99) (Rosch: 83–90)
“Rosch deserves credit, in my view, for her principles of categorization and for recognizing the roles of concept economy, cue validity, and differences within categories in concept formation. However, she seems to have given insufficient attention to the role of differences across categories, . . . . / Osherson and Smith . . . . distinguish between a concepts core and its identification procedure. The core is concerned with those aspects of a concept that explicates its relation to other concepts, and to thoughts, while the identification procedure specifies the kind of information used to make rapid decision about membership. / Smith and, separately, Paul Thagard have also noted the prototype’s inadequacy regarding the role of causality in concepts.”

III. Developmental Psychology
• Jean Piaget (pp. 100–112)
• Lev Vygotsky (112–14)
• Frank Keil (114–22)
• Criticisms of Stage Theories (122–24)
• Kinds of Similarity (124–26)

~~~~~~~~~~~~~~~~

Two essays by David Kelley deal with Rand’s theory of concepts, including its relation to prototype theory:

Kelley, D. 1984. A Theory of Abstraction. Cognition and Brain Theory 7:329–57.
“Typicality or representativeness must be distinguished from ‘instancehood’ or membership in a kind.”

Kelley, D., and J. Krueger 1984. The Psychology of Abstraction. Journal for the Theory of Social Behaviour 14(1):43–67.

Post 18

Friday, September 25, 2009 - 10:58am

Rand's theory of concept formation makes sense to me; the prototype theory doesn't -- unless the difference between Randian concept-formation and prototype concept-formation is, possibly, the difference between an objective thinker and a second-hander?

For example:
Let's imagine two small children, whom I'll call Howard and Peter.
Both Howard and Peter, at age two or thereabouts, are adding the concept (and the word) "bird" to their conceptual and linguistic repertoires.

Howard and Peter have each noticed that some animals have feathers and some do not -- Howard and Peter have each heard other people apply the word "bird" to chickens, canaries, seagulls, robins, etc.
/1/
Howard notices that the animals called "birds" all have feathers, and that animals (or other things) without feathers don't get called "birds" -- Howard thereby forms a concept of "bird" as "an animal that grows feathers."
When Howard goes to the zoo and sees an ostrich and a parrot, he therefore unhesitatingly (and correctly) recognizes them as birds, even though the ostrich looks very different (and the parrot also looks rather different) from any bird he has seen before.

BUT ...

/2/
Peter, unlike Howard, doesn't rely on identification of like and unlike features: instead of forming a concept of "bird" based on some feature that all birds share (and that no non-bird shares), two-year-old Peter forms his concept of "bird" by mentally listing (and not analyzing) all the things he's heard other people call 'birds' (seagulls, robins, canaries, chickens, pigeons, etc.)
Having accepted this unanalyzed mental agglomeration, Peter creates as his concept of "bird" whatever features appear in most or all of the birds in this mental database.
Young Peter (building this mental database of "what people call 'bird') uses this database to create a sort of mental "average" (or, better, a sort of "greatest common factor") of the different things he's seen the label "bird" applied to.
In other words:
Peter's mental prototype of "bird" (that he uses for a concept) will probably blur together, on equal or roughly equal terms, an assortment of a wide variety of criteria such as "feathers" and "beak" and "uttering high-pitched sounds" and "flying" and "less than three feet high" because all those things hold true of all the birds he's seen so far (seagulls, robins, canaries, chickens, pigeons)
When he later sees his first ostrich and his first parrot, (sometime after having already formed this mental prototype), he may have trouble regarding it as a bird
(even after he learns to call it a bird, somewhere "deep down inside" (where he built his mental prototype) he'll continue to regard ostriches as "not real[ly] birds" -- not part of the prototype -- because ostriches contradict his prototype by reason of their size and their inability to fly. (The parrot won't give him a problem, because its size, flying, and other features don't contradict the prototype.)

When Howard and Peter grow up (and both men accept architectural commissions which may include the design of aviaries), Peter may have some conceptual obstacles to overcome in efficiently correctly designing that part of the aviary that houses ostriches, penguins, and other birds that don't fit his mental prototype-blur of "what a bird really is." (E.g., if Peter subconsciously doesn't regard ostriches, emus, and penguins as really birds, he may inadvertently put less effective thought and effort into designing their sections of the aviary, as compared with designing the sections intended for parrots, pheasants, hummingbirds and other birds that -- though he didn't build his prototype around them -- don't happen to violate his prototype.)

Post 19

Friday, September 25, 2009 - 11:42am