I would agree that this was long, but still a good read. As I read through the second chapter, I was waiting for ideals to come up in the discussion of the prototype and exemplar theories, and was glad to see it finally touched on in the discussion of knowledge theory. I think this reinforces the fact that none of these approaches on their own can sufficiently explain how we categorize.
One thing the article doesn't address very clearly (although it is hinted at in the knowledge approach) is the role of context. It explains all these theories as if we observe the surface features of these items without taking into account contextual information. For example, a knife might be more readily classified as a kitchen utensil or a weapon depending on the context (e.g. if it's lying on a countertop next to a fork or being wielded by a guy in a hockey mask). The article seemed mostly concerned with how we use surface features to form concepts, but I think context is also an important factor.
It is a well-representative introduction to how researchers strive to define categories (or more specifically, meanings) using logic(the classic view)/semantic categories/features and weights/statistical(frequency of occurrence) in theoretical terms. Throughout these two chapters, I agreed heartily with author’s statement saying that the nature of meaning itself is like a continuum, a continuous stream of concepts! Therefore, the fuzziness among categories is something inevitable. For example, like categories used in infant vocal productions--vowel-like sounds and consonant-like sounds. How much is one vowel-like sound produced by a baby similar to a vowel produced by an adult? Are they 50% alike, or 52% alike? In what dimensions they are similar, articulatory, acoustic, both, or others? And how each dimension be evaluated in terms of likeness? Is it fair if we attempt to categorize infants’ sounds in terms of adults’ mature categories? If not, why not? If yes, why so?
Nice, a reading that I actually understand! I hope this is true for forthcoming readings as well. I have a decent background in Barsalou’s recent works and I typically find his ideas pretty convincing. Especially convincing is an article from 1998 in which he and Goldstone stress a connection between “lowly” perception and the archetypical example of “pure cognition”, conception. I wish this grounded theory would have been discussed because I feel it coheres to an organic conception of the mind. The closest equivalent mentioned in the article is the knowledge approach which is more ecological in my opinion. Concepts do not need to be derived from overly complicated logical inferences or abstract equations if they are simply grounded in perceptual experience. In the world in which we evolved we needed to quickly categorize and conceptualize in order to survive. We could quickly filter out heuristic essences based upon the context in order to aid in achieving rapidly changing goals. Concepts are therefore a first principle in that they are built into the very structure of experience rather than being something that we need to apply afterward experience. That’s my opinion at least
While reading this I found myself recalling tenets of connectionism. (Are the ideas expressed here similar to those founded there? Is it a stretch to view them as similar?) I am fascinated with the idea of patterns. Ben's statement "an organic conception of the mind" really hit home. Borrowing from Physics, if form follows function, could it be that we seek categorization and patterns because our brains, on a molecular level, process information as such? Like circuits, represented as patterns, that can be overlain onto new concepts? As well, once a circuit is formed, can shortcuts be administered when new information is presented? If negative information is processed slower than positive information, could it be the mind is searching for a pattern that isn't there? Positive information is being processed faster because the pattern is pre-established?
I really enjoyed this reading! I’ve read about a lot of the concepts before, so it was a nice refresher and it also gave more detail and discussed things I hadn’t learned about before. I found it very interesting when it was noted that those who try to salvage the classical view are more often philosophers than psychologists. I am in accord with the author in wondering why this may be, and I also wonder if there is anyone in the class who thinks the classic view should/can be salvaged. I’m also curious to see how the class feels about the prototype view (either original or a modified one to include feature lists for example) versus the exemplar view. I’m really just exited to hear everyone’s opinions on the topic. Also, the last part of the text, the General Context Model, kind of threw me. I thought the material before that section was more clearly written. Perhaps that’s just because I am somewhat familiar with the theories mentioned earlier in the text rather than this last one. Although I am usually interested in practical applications, when it comes to this topic, I am much more content to ponder it from a theoretical viewpoint, rather than mathematically calculating the distance between similarities and such.
I agree that this was a great read (I lol'd a couple of times!). I also agree with previous statements that context is extremely important. David's comment about kitchen knives summed that up pretty well. I think that how how we directly interact with the item/category is a main factor of categoriness (Barsalou would probably dig that). We (Wisniewski, Clancy, & Tillman, 2005, to properly give credit) looked at categorization of plural nouns (multiple entities), and found 4 distinct categories based on grammatical "behavior." For example, you can't have "a soapsud" or "a Rockie," a star out of the constellation Sagittarius is not in itself a constellation-nor is the rest of it with that star removed, and so on. So, while I see the wileyness of putting our finger directly on what is/not a category, a lot can come from indirect evidence like "seeing" the wind move leaves. We were able to discern conceptualization (role-governed, feature-based, etc.) of a category by its grammatical context.
It seems to me that while all of these theories deal with how people make the decision that an object belongs to a specific concept, the more interesting decision making is for the borderline objects. Specifically, the borderline objects that oscillate between two or more concepts for the same person. What are the factors that alter a person's classification of the same object? Does this classification change based on new experiences (an alteration to the prototype or exemplar)? Or are there some individual differences that cause people to alter their classification? For me, this type of decision making is more interesting than how people go about deciding that an object definitely belongs to one concept.
***This post is from Haiying** Typicality and the Classical View of Category
This article is much easier for me to understand than the topic model one. However, the figure 2.1 seemed a little bit confusing. For example, what do “words” of “oo”, and “ror” mean? At first, I thought it represented the “concept” of “氵”and “歹”, but it seems not. And what do the letter series above “I”, “II”, “III” etc. represent? They did not represent the meanings or concepts of the Chinese words below. Therefore, I could not connect this figure with the topic that the author elaborated.
Was this example fabricated to illustrate Hull’s argument?
Besides “necessary” and “sufficient”, the experiments that reseachers conducted to test learning normal concepts in the perspective of classical view actually depend on an essential characteristics of individual such as the perception and language expression accuracy etc. How were these factors excluded to merely test concepts learning.
I extremely like the theoretical section of “In-principle Arguments”, for it approves that the world is not absolute entity, but has the relativity.
This ability of categorization, especially of the fuzziness without the explicit boundary, relies on partly the human cognition. I remember a story about the different capabilities of color identification that male and female have. Female can name more colors; Does that indicate that the color difference in male and female lies in sex choromosome or just the perception and attention difference?
I really liked the read (any article with hermaphrodite slug as examples has to be good right ... right ..). I had come across the different theories before and this was a great expansion in particular showing the more recent developments. My only request in terms of content would be to expand on schemata theory. He seemed to treat it as a subsection of prototype theory and covered it briefly.
I like Benjamin's point about the importance of categorization. It for this reason I am not surprised at the complexity of the field and that various theories seem to all have explanatory power. Each theory seems to have identified individual effects of categories and I would not be surprised if these were separate processes. Which brings me to the question, is there any neuroscience data on this subject? Have they identified localized areas for categorization?
My last couple question are. The examples and experiments concentrate on physical categories; how are abstract concepts handled? Have these been explored by neural network theorists. The probabilistic nature seemed to include certain threshold in putting something in a category, and this reminded me of activation energy in a neural network. Might be a stretch but thought I'd ask?
As I was reading, I think that some of the similar themes that have been touched on were coming up in my head. In terms of how we go about classifying a particular "object" into a concept, I am wondering if in any of these proposed models, if the class may see emotions (more specifically, affective network theories) as possibly having any impact? For example, does the emotional experience that a person has with the exemplar or prototype effect how that is remembered and how new objects are compared to the existing schema/exemplar/prototype? And what about the emotional charge of an old object vs the emotional charge of a new topic...what happens if they are different?
I would be interested to hear what factors the class sees as the most important or if anything is missing..context related or not.
I believe my question(s) are somewhat related to Blair's line of thinking about objects that change categories. As I was reading, I was thinking about humans categorizing things instinctively. We see early categorization in children as they come to understand the world - all large things with legs get called cows for awhile (or in my case, hippos), etc. I think we can see the features children use to categorize things in their drawings. Children's drawings are largely symbolic; they attempt to represent the features of things without making the drawing actually resemble the object (e.g. 2 stick figure legs). Then their drawings seem to change as they get older. I wondered, is this change a symptom of the changes taking place in their concepts? Are the important "features", "resemblances", "prototypes", etc. changing as they acquire more information about the world? If we were to track a few concepts over time in the same people, would what differences would we see over time and what would trigger those differences? Does our knowledge evolution also change our concept/categorization evolution?
If I were to summarize my understanding of the text in one phrase it would be: These chapters attempt to describe why the Classical View of operating with knowledge categories was largely abandoned in favor of approaches that include category fuzziness in their core and then describes what these alternative Views of knowledge representation are.
In a shorter phrase: Why and how to get rid of the "law of the excluded middle".
My first comment is this was an interesting reading. My second comment is that I am looking forward to the next views that Science comes up with, this time with the major contributions from Neuroscientists and AI researchers. I am looking forward to views that try to integrate actual neuronal activity, as observed in human/animal brains and confirmed by experiments on subjects, including Artificial Neural Networks (which when I was 10 years younger I was certain would happen by now). How does the awesome pattern recognition machine that our brain is construct the notion and the instances of a "concept"? A sheep can recognize her offspring's call in a flock of thousands of lambs, all born just half an hour before. How does that impressive machine gradually change its nature and then steps on the shaky grounds of representing and operating with mathematical notions (as abstract as "38" and even more)?
I am convinced, however, that these views that I read about are a necessary stepping stone in order to understand the next ones, the "brain" views. That's why I consider this reading (and hopefully the subsequent presentation) are extremely important knowledge, although I do understand perfectly why they might seem "a bit too philosophical" to my computationally-oriented Computer Science classmates (if any). A top-down approach is, therefore, crucial for the future understanding and integrating of the huge amounts of bottom-up data and hypotheses that Neuroscientists are continuously offering.
I have a lot of minor questions and comments, but they are all about non-critical components of the main theses of the paper. I will provide just a few examples. -on page 23, "if eagles became sick" example is overlooking people's common representation of bird interaction -on page 24, "4 is a more typical even number than 38", i believe they wrongly leave out memory, which is something else then processing. I believe the "frequency" explanation on page 37 does not work here. My impression, backed by no scientific reference that I know of, is that the number 4 is generated mentally, represented and stored in a different manner than number 38. I also believe that the number 38 IS generated similarly to number 42, but not to number 4. (That is unless you are in some special categories of autistic savants, for which 391 is hugely different from 396, since the first is 17x23 while the second has 18 factors) -on page 62, with Barsalou's discovery that "ideals" are important to determining typicality I would have enjoyed to be reminded of Plato's "eidos" philosophy
In the paper it addressed the issue of concepts, in which it says that we are very bad at addressing, for example, the definition of what a "dog" is, even though we can recognize a dog easily. I think this is rather fallacious, because it did not take into the consideration of knowledge affects, in which people tend to have a harder and harder time to define something actively using reductionist techniques. So, if someone doesn't know much of anything about dogs, its physiology, or anatomic diversity, the person can feel that defining of what a dog is would be easy: "it barks, it has pointing nose, it wags its tail" etc. etc. whereas someone who is an animal scientist would have a hard time because of all the inflow of information that can contradict one another
In organizing these posts I realized I didn't post myself.
The knowledge approach was presented as being akin to theory theory. I learned about theory theory in reference to TOM. It suggests that human beings rely upon folk-psychology to ‘mind-read’. In other words, individuals observe a behavior and, using theories about human behavior, infer the psychological state of others. Such theories of human behavior are thought to be stored in memory as mental representations (although it is unclear as to the exact nature of those mental representations). For example, one may have a theory concerning what it means when someone is frowning (i.e. frowning indicates sadness). In this case it seems as if we replace the 'individual' with 'the world' we get the knowledge approach. However, theory theory is often presented in contrast to simulation theory. In this case would exemplars/prototypes be more akin to simulation theory? This seems to be contrary to Benjamin's suggestion that knowledge theory is more perceptually grounded than the others.
I would agree that this was long, but still a good read. As I read through the second chapter, I was waiting for ideals to come up in the discussion of the prototype and exemplar theories, and was glad to see it finally touched on in the discussion of knowledge theory. I think this reinforces the fact that none of these approaches on their own can sufficiently explain how we categorize.
ReplyDeleteOne thing the article doesn't address very clearly (although it is hinted at in the knowledge approach) is the role of context. It explains all these theories as if we observe the surface features of these items without taking into account contextual information. For example, a knife might be more readily classified as a kitchen utensil or a weapon depending on the context (e.g. if it's lying on a countertop next to a fork or being wielded by a guy in a hockey mask). The article seemed mostly concerned with how we use surface features to form concepts, but I think context is also an important factor.
It is a well-representative introduction to how researchers strive to define categories (or more specifically, meanings) using logic(the classic view)/semantic categories/features and weights/statistical(frequency of occurrence) in theoretical terms. Throughout these two chapters, I agreed heartily with author’s statement saying that the nature of meaning itself is like a continuum, a continuous stream of concepts! Therefore, the fuzziness among categories is something inevitable. For example, like categories used in infant vocal productions--vowel-like sounds and consonant-like sounds. How much is one vowel-like sound produced by a baby similar to a vowel produced by an adult? Are they 50% alike, or 52% alike? In what dimensions they are similar, articulatory, acoustic, both, or others? And how each dimension be evaluated in terms of likeness? Is it fair if we attempt to categorize infants’ sounds in terms of adults’ mature categories? If not, why not? If yes, why so?
ReplyDeleteNice, a reading that I actually understand! I hope this is true for forthcoming readings as well.
ReplyDeleteI have a decent background in Barsalou’s recent works and I typically find his ideas pretty convincing. Especially convincing is an article from 1998 in which he and Goldstone stress a connection between “lowly” perception and the archetypical example of “pure cognition”, conception. I wish this grounded theory would have been discussed because I feel it coheres to an organic conception of the mind. The closest equivalent mentioned in the article is the knowledge approach which is more ecological in my opinion. Concepts do not need to be derived from overly complicated logical inferences or abstract equations if they are simply grounded in perceptual experience. In the world in which we evolved we needed to quickly categorize and conceptualize in order to survive. We could quickly filter out heuristic essences based upon the context in order to aid in achieving rapidly changing goals. Concepts are therefore a first principle in that they are built into the very structure of experience rather than being something that we need to apply afterward experience.
That’s my opinion at least
While reading this I found myself recalling tenets of connectionism. (Are the ideas expressed here similar to those founded there? Is it a stretch to view them as similar?) I am fascinated with the idea of patterns. Ben's statement "an organic conception of the mind" really hit home. Borrowing from Physics, if form follows function, could it be that we seek categorization and patterns because our brains, on a molecular level, process information as such? Like circuits, represented as patterns, that can be overlain onto new concepts? As well, once a circuit is formed, can shortcuts be administered when new information is presented? If negative information is processed slower than positive information, could it be the mind is searching for a pattern that isn't there? Positive information is being processed faster because the pattern is pre-established?
ReplyDeleteI really enjoyed this reading! I’ve read about a lot of the concepts before, so it was a nice refresher and it also gave more detail and discussed things I hadn’t learned about before. I found it very interesting when it was noted that those who try to salvage the classical view are more often philosophers than psychologists. I am in accord with the author in wondering why this may be, and I also wonder if there is anyone in the class who thinks the classic view should/can be salvaged. I’m also curious to see how the class feels about the prototype view (either original or a modified one to include feature lists for example) versus the exemplar view. I’m really just exited to hear everyone’s opinions on the topic. Also, the last part of the text, the General Context Model, kind of threw me. I thought the material before that section was more clearly written. Perhaps that’s just because I am somewhat familiar with the theories mentioned earlier in the text rather than this last one. Although I am usually interested in practical applications, when it comes to this topic, I am much more content to ponder it from a theoretical viewpoint, rather than mathematically calculating the distance between similarities and such.
ReplyDeleteI agree that this was a great read (I lol'd a couple of times!). I also agree with previous statements that context is extremely important. David's comment about kitchen knives summed that up pretty well. I think that how how we directly interact with the item/category is a main factor of categoriness (Barsalou would probably dig that). We (Wisniewski, Clancy, & Tillman, 2005, to properly give credit) looked at categorization of plural nouns (multiple entities), and found 4 distinct categories based on grammatical "behavior." For example, you can't have "a soapsud" or "a Rockie," a star out of the constellation Sagittarius is not in itself a constellation-nor is the rest of it with that star removed, and so on. So, while I see the wileyness of putting our finger directly on what is/not a category, a lot can come from indirect evidence like "seeing" the wind move leaves. We were able to discern conceptualization (role-governed, feature-based, etc.) of a category by its grammatical context.
ReplyDeleteIt seems to me that while all of these theories deal with how people make the decision that an object belongs to a specific concept, the more interesting decision making is for the borderline objects. Specifically, the borderline objects that oscillate between two or more concepts for the same person. What are the factors that alter a person's classification of the same object? Does this classification change based on new experiences (an alteration to the prototype or exemplar)? Or are there some individual differences that cause people to alter their classification? For me, this type of decision making is more interesting than how people go about deciding that an object definitely belongs to one concept.
ReplyDelete-Blair
***This post is from Haiying**
ReplyDeleteTypicality and the Classical View of Category
This article is much easier for me to understand than the topic model one. However, the figure 2.1 seemed a little bit confusing. For example, what do “words” of “oo”, and “ror” mean? At first, I thought it represented the “concept” of “氵”and “歹”, but it seems not. And what do the letter series above “I”, “II”, “III” etc. represent? They did not represent the meanings or concepts of the Chinese words below. Therefore, I could not connect this figure with the topic that the author elaborated.
Was this example fabricated to illustrate Hull’s argument?
Besides “necessary” and “sufficient”, the experiments that reseachers conducted to test learning normal concepts in the perspective of classical view actually depend on an essential characteristics of individual such as the perception and language expression accuracy etc. How were these factors excluded to merely test concepts learning.
I extremely like the theoretical section of “In-principle Arguments”, for it approves that the world is not absolute entity, but has the relativity.
This ability of categorization, especially of the fuzziness without the explicit boundary, relies on partly the human cognition. I remember a story about the different capabilities of color identification that male and female have. Female can name more colors; Does that indicate that the color difference in male and female lies in sex choromosome or just the perception and attention difference?
I really liked the read (any article with hermaphrodite slug as examples has to be good right ... right ..). I had come across the different theories before and this was a great expansion in particular showing the more recent developments. My only request in terms of content would be to expand on schemata theory. He seemed to treat it as a subsection of prototype theory and covered it briefly.
ReplyDeleteI like Benjamin's point about the importance of categorization. It for this reason I am not surprised at the complexity of the field and that various theories seem to all have explanatory power. Each theory seems to have identified individual effects of categories and I would not be surprised if these were separate processes. Which brings me to the question, is there any neuroscience data on this subject? Have they identified localized areas for categorization?
My last couple question are. The examples and experiments concentrate on physical categories; how are abstract concepts handled? Have these been explored by neural network theorists. The probabilistic nature seemed to include certain threshold in putting something in a category, and this reminded me of activation energy in a neural network. Might be a stretch but thought I'd ask?
As I was reading, I think that some of the similar themes that have been touched on were coming up in my head. In terms of how we go about classifying a particular "object" into a concept, I am wondering if in any of these proposed models, if the class may see emotions (more specifically, affective network theories) as possibly having any impact? For example, does the emotional experience that a person has with the exemplar or prototype effect how that is remembered and how new objects are compared to the existing schema/exemplar/prototype? And what about the emotional charge of an old object vs the emotional charge of a new topic...what happens if they are different?
ReplyDeleteI would be interested to hear what factors the class sees as the most important or if anything is missing..context related or not.
I believe my question(s) are somewhat related to Blair's line of thinking about objects that change categories. As I was reading, I was thinking about humans categorizing things instinctively. We see early categorization in children as they come to understand the world - all large things with legs get called cows for awhile (or in my case, hippos), etc. I think we can see the features children use to categorize things in their drawings. Children's drawings are largely symbolic; they attempt to represent the features of things without making the drawing actually resemble the object (e.g. 2 stick figure legs). Then their drawings seem to change as they get older. I wondered, is this change a symptom of the changes taking place in their concepts? Are the important "features", "resemblances", "prototypes", etc. changing as they acquire more information about the world? If we were to track a few concepts over time in the same people, would what differences would we see over time and what would trigger those differences? Does our knowledge evolution also change our concept/categorization evolution?
ReplyDeleteIf I were to summarize my understanding of the text in one phrase it would be: These chapters attempt to describe why the Classical View of operating with knowledge categories was largely abandoned in favor of approaches that include category fuzziness in their core and then describes what these alternative Views of knowledge representation are.
ReplyDeleteIn a shorter phrase: Why and how to get rid of the "law of the excluded middle".
My first comment is this was an interesting reading. My second comment is that I am looking forward to the next views that Science comes up with, this time with the major contributions from Neuroscientists and AI researchers. I am looking forward to views that try to integrate actual neuronal activity, as observed in human/animal brains and confirmed by experiments on subjects, including Artificial Neural Networks (which when I was 10 years younger I was certain would happen by now).
How does the awesome pattern recognition machine that our brain is construct the notion and the instances of a "concept"? A sheep can recognize her offspring's call in a flock of thousands of lambs, all born just half an hour before. How does that impressive machine gradually change its nature and then steps on the shaky grounds of representing and operating with mathematical notions (as abstract as "38" and even more)?
I am convinced, however, that these views that I read about are a necessary stepping stone in order to understand the next ones, the "brain" views. That's why I consider this reading (and hopefully the subsequent presentation) are extremely important knowledge, although I do understand perfectly why they might seem "a bit too philosophical" to my computationally-oriented Computer Science classmates (if any). A top-down approach is, therefore, crucial for the future understanding and integrating of the huge amounts of bottom-up data and hypotheses that Neuroscientists are continuously offering.
I have a lot of minor questions and comments, but they are all about non-critical components of the main theses of the paper. I will provide just a few examples.
-on page 23, "if eagles became sick" example is overlooking people's common representation of bird interaction
-on page 24, "4 is a more typical even number than 38", i believe they wrongly leave out memory, which is something else then processing. I believe the "frequency" explanation on page 37 does not work here. My impression, backed by no scientific reference that I know of, is that the number 4 is generated mentally, represented and stored in a different manner than number 38. I also believe that the number 38 IS generated similarly to number 42, but not to number 4. (That is unless you are in some special categories of autistic savants, for which 391 is hugely different from 396, since the first is 17x23 while the second has 18 factors)
-on page 62, with Barsalou's discovery that "ideals" are important to determining typicality I would have enjoyed to be reminded of Plato's "eidos" philosophy
---Cristian
This comment has been removed by the author.
ReplyDeleteIn the paper it addressed the issue of concepts, in which it says that we are very bad at addressing, for example, the definition of what a "dog" is, even though we can recognize a dog easily. I think this is rather fallacious, because it did not take into the consideration of knowledge affects, in which people tend to have a harder and harder time to define something actively using reductionist techniques. So, if someone doesn't know much of anything about dogs, its physiology, or anatomic diversity, the person can feel that defining of what a dog is would be easy: "it barks, it has pointing nose, it wags its tail" etc. etc. whereas someone who is an animal scientist would have a hard time because of all the inflow of information that can contradict one another
ReplyDeleteIn organizing these posts I realized I didn't post myself.
ReplyDeleteThe knowledge approach was presented as being akin to theory theory. I learned about theory theory in reference to TOM. It suggests that human beings rely upon folk-psychology to ‘mind-read’. In other words, individuals observe a behavior and, using theories about human behavior, infer the psychological state of others. Such theories of human behavior are thought to be stored in memory as mental representations (although it is unclear as to the exact nature of those mental representations). For example, one may have a theory concerning what it means when someone is frowning (i.e. frowning indicates sadness). In this case it seems as if we replace the 'individual' with 'the world' we get the knowledge approach. However, theory theory is often presented in contrast to simulation theory. In this case would exemplars/prototypes be more akin to simulation theory? This seems to be contrary to Benjamin's suggestion that knowledge theory is more perceptually grounded than the others.