Prototype theory of concept formation

Stephen,

Thanks for the link to Merlin's work. Twas very good.

Jordan

Post 21

Friday, September 25, 2009 - 12:16pm

I realize that imprinting is an imperfect analogy, but that's about the best one I can come up with for PT. Complex imprinting of concepts. The first example of a concept is what imprints the prototype in their mind, right. What method would be used to add or subtract detail to a concept after the initial (for lack of a better word) imprinting? Rand's theory accounts for this, I'm not seeing how PT does. How would you alter a prototype based on later observed existents if your prototype was fixed by the initial existent you encountered?

Post 22

Friday, September 25, 2009 - 2:11pm

Thanks, Jordan!

Friday, September 25, 2009 - 2:46pm

Oh. Ryan, prototypes aren't fixed. What typifies a category can change over time based on the relationship among the category's constituent members.

Jordan

Post 24

Friday, September 25, 2009 - 3:35pm

Ok, what criteria are being observed to motivate such a change? Once the person in question has this large database of existence that roughly correspond to the original prototype, how does the person apply "updates" to the prototype? This sounds like the person is synthesizing their own platonic ideal for the concept through a process of pattern recognition. Its a little more plausible, but still not quite there. The only process I can think of that would allow for refining that ideal image of a concept would be Rand's method, determining qualities of the concept without specific measurements. At that point PT seems more of a theory of concept utilization to me, not formation. There is still the problem with higher order concepts.
(Edited by Ryan Keith Roper on 9/25, 3:36pm)

Post 25

Friday, September 25, 2009 - 4:38pm

Hi Ryan,

Given your last question, I think I somewhat misrepresented PT. PT apparently doesn't entail updating the prototype of a category. Rather it entails updating which existents serve as prototypes in a category. From what I gather, it says we do this by correlating similarities among the existents. For example, say we are initially exposed to a modest number of storks, flamingos, herons, egrets, cranes, emus, and ostriches, but no other birds. We might find that storks are the initial exemplar here. They correlate to all those other birds better than any one of those birds correlates to all the others. But then say we are further exposed to myriad robins, ducks, crows, gulls, sparrows, parrots, swallows, pigeons. At this point, perhaps storks become less of an exemplar, and robins more of an exemplar, because of robins' superior correlation to all the others. The core switches with new correlation.

I highly recommend the prototype theory excerpt from Merlin's book that Stephen linked us to.

Jordan

Post 26

Friday, September 25, 2009 - 4:48pm

And yet when I think of "bird" I don't think of any specific species of bird. I doubt most people do. I'm just not buying it. Even determining the differences and similarities between all of those species of birds appears to require an method of concept formation. Again, this has the feel of a utilization system. What would your exemplar for "assault" be? Or really any other social or second + order concept? A mental snapshot of the event?

Post 27

Friday, September 25, 2009 - 5:04pm

"The core existent is just the existent in a category against which other existents would be measured." You are still stealing the concept (i.e., begging the question) - now you are calling it a "category" in stead of a concept. Where did this category come from?

The theory is hopelessly circular. It says we form a concept (which presumably doesn't yet exist) by identifying a typical example (typical example of what? of the concept that doesn't exist yet?) and then categorizing others by their similarity to that typical example.

Given your silence, I assume you cannot give an answer to the request for a typical exemplar of the concept "yesterday" based on prototype theory. I am curious if these TPers ever acknowledge that our conceptual faculty is hierarchical.

As for the fact that some scientists accept this theory, that acceptance is based on a faulty notion of the scientific method. As with math and logic, the correct way of determining the proper method of concept foundation is not to conduct a survey of untrained laymen. Some psychologists accept the prototype theory because when they run "experiments" showing people pictures of birds and asking them to identify whether or not they are birds the people answer more quickly when the image is of a robin than of a nightjar or a bird of paradise. This proves nothing. Or, I should say, it proves just as much about how to calculate the value of pi by surveying people.

Post 28

Friday, September 25, 2009 - 8:15pm

Are these birds?

At the top is a displaying male superb bird of paradise. Second is a buff nightjar (a.k.a. "goatsucker," known in Spanish as the chupacabra):

The obvious question is, if resemblance to a "typical bird" is what determines whether something is a bird or not, then how is resemblance to be judged? Frankly, I think the nightjar looks like a leafless Ent and the splendid bird of paradise looks like a dancing smiley face icon or one of Steven King's langoliers.

Post 29

Friday, September 25, 2009 - 8:40pm

That's hilarious! What a fine dancing bird that is :-)

I danced right up and sanctioned you.

Post 30

Friday, September 25, 2009 - 8:51pm

Oh, Steve, if you've never watched Sir David Attenborough's "The Life of Birds" documentary, run out and rent it immediately! Wonderful!

Post 31

Friday, September 25, 2009 - 10:06pm

Thanks, Teresa. I put it on my Netflix queue.

Post 32

Saturday, September 26, 2009 - 5:36am

Re:

the correct way of determining the proper method of concept foundation is not to conduct a survey of untrained laymen.

What made you suppose that the psychologists even *wanted* to determine "the proper method of concept formation"? Their method (which wouldn't work for determining how one should, ideally, form a concept) works admirably for finding out how people *do* in fact form concepts.
Psychologists (like practitioners of the other, more advanced sciences) have far more interest in discovering what does happen than in deciding what ought to have happened instead.

Post 33

Saturday, September 26, 2009 - 9:36am

Okay, Kate, I will grant that those who designed this "experiment" had no interest in finding out how one properly forms a concept.

But neither does it show how they do form a concept. Asking people to use a criterion no more admirably shows how people do form criteria than asking people to use banana cream pies shows how they do in fact form banana cream pies.

This experiment has nothing to do with the nature of concepts as such. The experiment cannot be replicated with such concepts as "yesterday" or "fraud" or any higher level concept. It has nothing to say about definitions by essence, since the question isn't even formulated. The fact that people can identify songbirds as birds more quickly than they can identify a hoatzin chick (which has claws on its wings) as a bird may not contradict the prototype theory of concepts, but it in now way demonstrates that theory, and neither does it contradict a "definition by essential properties" theory of concepts.

We are nowhere near having the proper technology to watch the brain form the structures that underly the possession of a new concept. We are currently limited to the method of introspection and counterexample. We can form a guess as to what concepts are and how they are formed using introspection. The theory developed by that method can be tested against counterexamples. Rand's theory is useful. It shows how stolen concepts are invalid. Its hierarchical structure mirrors both the developmental history of the child who first forms perceptual level concepts and then forms higher level abstractions from abstractions. This congruence with the timeline of a child's development of concepts is paralleled by the history of the development of concepts in human cultures, with words like "certain" and "discrete" developing from earlier simple roots like *sker- with the meaning "to cut." Prototype theory cannot deal with abstractions from abstractions. It is easily shown to be false with such counterexamples as the concept "yesterday" while Rand's theory has no such difficulties.

The essence of Objectivism is Rand's theory of concepts. The hierarchical nature of concepts demonstrates the invalidity of fallacies such as skepticism or moral relativism which try to use higher level abstractions to deny the validity of the lower-level abstractions upon which they are based. If a person does not understand Rand's theory of concepts, he might parrot her dogma but he would not be capable of showing how to derive her results from her premises. Such a person is not (yet) an Objectivist.

(Edited by Ted Keer on 9/26, 8:18pm)

Post 34

Saturday, September 26, 2009 - 10:07am

Re:

If a person does not understand Rand's theory of concepts, he might parrot her dogma but he would not be capable of showing how to derive her results from her premises. Such a person is not (yet) an Objectivist.

Granted!
This brings to mind the likelihood that such persons will arise even in completely Objectivist societies -- at least some people who have the intelligence to become Objectivists, and who also decide to have children, may have at least some children who unfortunately don't have the intelligence to become Objectivists.

Just as the d'Anconia family (for very good and sufficient reasons) did not regard any of their offspring as d'Anconias until the child had attained a certain standard of achievement,
might an Objectivist government (on any larger scale than the very small town of Galt's Gulch) find reason to decide that those who don't understand Objectivism do not (yet) qualify for citizenship?
Such a government might defend such a view on the grounds that someone who does not understand the basic premises and philosophy of the civilization into which he or she was born should not take part in the government of that civilization (e.g., by voting or holding political office). Given modern technology, one might imagine a computerized voting booth that won't display the slate of candidates until the voter has correctly answered a randomly chosen question whose answer requires correctly applying one or more Randian premises. (Answer the philosophy question correctly -- you get to vote. Fail to answer correctly within five minutes -- the screen goes blank, the controls lock, and the curtain opens to admit the next voter.)

Comments? What facts might urge, might make permissible, or might forbid a test of this sort in a free society?
(Edited by Kate Gladstone on 9/26, 10:09am)

Post 35

Saturday, September 26, 2009 - 10:43am

The purpose of voting is to provide an alternative to civil war in choosing who will be the de facto arbiters of force. It would be nice if rationality were the criterion, but in the end it is the threat of armed rebellion by those with the resources to wage war to protect their own concerns that really matters. Formal criteria such as residence, citizenship, and the age of consent all utlimately come down to who has the means and the motivated interest to contest the sovereign, should it come down to a fight.

Post 36

Saturday, September 26, 2009 - 11:28am

Hi Ryan,

When you ask people to think of the concept birds, they do report thinking of particular species. They are also able to more quickly sort some bird species into the concept birds, the idea being that more typical bird species are easier to sort. And I would guess an exemplar for "assault" would be some mental snapshot (or video). But I don't know. I'm not sure what PT has done or can do when it comes to studying acquisition and structure of higher order concepts.

Ted,

You continue to make this error, which might well be my fault for misdescribing PT. A new concept in PT starts with identifying a novel existent. Initially, that novel existent typifies nothing (except perhaps itself?) because there's nothing else in the category for it to typify. Later, other existents get added to the category based on their likeness to that first existent. Still nothing circular here. Please consider checking out the PT section in Merlin's book if you insist that the circularity remains.

I used the abstract concept furniture instead of yesterday because PT used it in their tests, but also because yesterday is an odd abstract concept to use to describe concept-formation in either PT or CT. Unlike a great many abstract concepts, it can contain only one member at any given time: there is always and only one day immediately before today. (Now, someone might object to this and say there are many yesterdays since many days have been and will be designated as "today," but then we start off on this tangent about whether those days are actually members of the concept; is something that will be or once was a member of a concept still a member of the concept? And so on. Big distraction from just trying to describe how PT treats abstract concepts.)

That said, you mentioned "fault" and Ryan mentioned "assault," both of which are abstract concepts, and we can easily agree that each has more than one constituent. As I said, I suspect Ryan is correct that PT would propose that those abstract concepts' exemplars are like mental snapshots. But again -- I don't know. And there has been criticism against PT on this very point, as I mentioned.

Interestingly, PT readily acknowledges that knowledge is hierarchical, and criticizes CT for ignoring the hierarchy. PT complains that CT lumps all members into a category as equals, ignoring the PT notion that some members are more important for the category than others. PT also complains that CT doesn't have a way to grow or shrink concepts in light of knew experience. Rand had this problem with CT, too. In traditional CT, essence is immutable: Either something always belongs under a concept, or it never does. Rand's theory escapes this criticism by rendering essences as epistemic, thus subject to adjustment, which is effectively the same solution nominalists came up with for this particular problem (though of course they greatly differ from Rand in numerous other ways).

Jordan

Post 37

Saturday, September 26, 2009 - 5:36pm

Which of These Things is Not Like the Other?

1) You are always forced back on definitions. "Novel" existent is just another way of stealing the concept. How do you identify an existent as novel? Every thing you encounter is just a thing. Why would any new thing be identified as new, when it could just be subsumed under the concept thing? It is because the new thing does not fit the existing definitions. Definitions can divide up the world. Definitions are either/or. The rule is that one forms a new concept of type when one encounters two or more objects which are similar to each other but which don't fit the definition of any prior concept. You are always forced back upon the use of definitions whether you want to or not. What would the PT experimenter asking people to identify birds do if one of the experimental subjects kept insisting that Elizabeth Hurley was "quite a bird"? You can't even explain the PT at all unless you understand definitions and take them as valid. We accept ostriches, penguins and the like as birds because they do indeed fit the definition. If concepts truly were determined by "resemblance" then there would be people who insist that penguins and ostriches are not birds because they don't resemble a typical bird. The phrase "typical X" inherently depends on the epistemological priority of "X" and "x" is always some defined concept.

2) As for your complaint, Jordan, that "yesterday" is a hard case, that is what a counterexample is! Are you going to complain that every good counter example I give is a bad one precisely because it is a good one?

Do you deny that "yesterday" is a counterexample?

3) As for this mental snapshot idea for more abstract concepts, that is special pleading. An editorial cartoon of a man with his pockets turned out is not an example of bankruptcy. Reality is not full of novel editorial cartoons. There is no image, no "core example" of bankruptcy.

4) People can and do agree on common definitions, and if they disagree they can either stipulate two concepts, one for each definition, or they can provide counter examples. When two people disagree as to whether a chicken or a sparrow is the actual core example, how do they decide?

5) Let's forget about the supposedly "special" case of yesterday, let's just use words in your last post. I can give an actual definition for each of the following concepts. Can you provider an actual typical image (i.e. an immediately recognizable perceptual form, not an editorial cartoon at which one can guess conceptually) of any one of them?

error
fault
misdescribe
concept
identifying
typify
except
perhaps
because
category
later

6) Some perceptual level concepts do have more and less typical members. Rand talks about borderline cases. But concepts have all sorts of weaker and stronger mental associations and connotations. The word bird in our mind might indeed be linked with the image of a robin, but it is also linked to sexy female and to rhymes with "word" and a whole lot of other things. None of this has anything to do with the price of tea in China.

7) As for PT criticisms of CT, do any of these actually apply to Rand's explanation of the Objectivist theory? If so, please do tell.

8) Finally, which of the following does not belong? Wounded cop, retired policeman, motorcycle cop, military policeman, fake cop, English Bobby, undercover policeman, or policeman killed in the line of duty?

Conventional theory has no problem with this. The fake cop is not a cop, all the others are. But the police impersonator presumably fits the core example closer than any other. The undercover cop may look like an criminal. He may undertake criminal activities to hide his identity. The retired cop might look like Frasier Crane's father. The bobby might look like a fireman with his tall helmet. The fake cop will be dressed as a cop. He might make traffic stops, and handcuff people, and go to crime scenes and attend police conventions. According to P theory the police impersonator is in every way more a cop than is the undercover cop. The only difference is a mark on a personnel file and a perhaps a paycheck. The difference is the definition.

Is anyone here buying PT?

(Edited by Ted Keer on 9/26, 8:13pm)

Post 38

Saturday, September 26, 2009 - 8:05pm

Ted,

This subject was presented in a very dry way, but you've managed to make it extremely entertaining and interesting.

You should have been a teacher.

Post 39

Saturday, September 26, 2009 - 9:48pm