Knowledge Flashcards
Consider the word “chair”. Name some attributes that might plausibly be included in a definition for this word. But, then, can you describe objects that you would count as chairs even though they don’t have one or more of these attributes?
Has four legs, you sit on it, they are wooden. A metal stool however has one leg and is not wooden.
What does it mean to say that there is a family resemblance among the various animals that we call dogs?
Wittgenstein proposed that there are defining features within a family. Nobody in the family has all of these features, but imagining they did, this person would be the “ideal” for each family. In many families, there is not one person who shares every single one of these features, but each member of the family shares some of these features. Wittgenstein proposed that ordinary categories like “dog” or “game” work in the same way. There may be no features shared by all dogs or games however we can identify “characteristic features” which many, perhaps most, category members have.
What is the prototype theory?
One way to think about definitions is that they set the “boundaries” for a category. If a test case has certain attributes, then it’s “inside” the boundaries and vice versa. The prototype theory, in contrast, begins with a different tactic: perhaps the best way to identify a category is to specify the “centre” of a category. In this case. The ideal is an average of all the dogs for example that you’ve see .
Why is graded membership a consequence of representing the category in terms of a prototype?
An implication to represent a category in terms of a prototype is membership in a category depends on resemblance to the prototype, and resemblance is a matter of degree (some are more likely to represent the prototype closer than others). Therefore membership is not a simple yes or no decision, it’s a matter of more if less. This weighing up against the prototype is called graded membership, such that the objects closer to the prototype are better (eg sole dogs are doggier than others and some books are bookier than others)
What task shows us that concept judgements often rely on prototype and typicality?
The sentence verification task. One way to test the proposal that mental categories have a graded membership is the sentence verification task, where participants indicate whether a sentence is true or false (e.g participants were slower for sentences asking if a penguin is a bird than for sentences like is a robin a bird). According to the prototype theory this is because they were comparing the things to their prototype of that category. A production task is where people are asked to name as many birds or dogs as they can. According to the prototype theory, they would firstly identify their prototype for the category then think about the things that resemble it. By this logic, the first birds mentioned should yield the fastest responses in the verification task, and his is exactly what happened. Another illustration of this is I’m rating tasks where people are asked to rate how “birdy”, for example, that birds are. The ones that are closer to their prototype are rated higher.
What are basic level categories?
Basic-level categorisations are represented generally with a specific word. For example if one was asked how people get to work, the reply would most likely be “a bus, car or train” rather than “some people drive a Toyota”. Some basic-level categories include “apple” or “chair”. If you were asked to explain what members of a category have in common with one another, you would have an easy time with basic level categories such as what do chairs have in common, but a harder time with more inclusive categories (what does all furniture have in common). Children learning to talk often firstly Aquire basic level terms which shows that basic level categories seems to reflect a natural way to categorise the objects in our world.
What is similar in the processes of categorising via a prototype and the processes of categorising via an exemplar? What is different between these two type of processes?
Exemplar based reasoning is when categorisation relies on knowledge about specific category members (asked what a chair is, you may think that it looks similar to an object in Jerry’s room which he calls a chair therefore it must be a chair): when an exemplar is defined as a specific remembered instance. It is similar to the prototype view because you still categorise objects by comparing them to a mentally represented standard and the process of assessing the similarity between the object and the standard is to determine whether the resemblance is great or not. Exemplars provide information that’s lost from the prototype- including information about the variability within the category. It is useful to merge the prototype and exemplar proposals.
Give an example in which something is definitely a category member even though it has little resemblance to the prototype for the category.
A lemon that is painted with red stripes, run over by a truck and injected with sugar is still a lemon, though it’s not yellow, round or sour. This shows that typicality can be separate from category membership. On the other side of the scale, a counterfeit $20 is not an actual $20 note though it possesses the exact same properties except who it was made by.
In judging similarity, why is it not enough simply to count all of the properties that two objects have in common?
Because with a little creativity, you could potentially count thousands of shared properties. Therefore, resemblance does depend on shared properties, but more precisely, on whether the objects share important, essential properties. The beliefs about which properties are important depend on your beliefs about the concept in question
Why is an informal, usually unstated “theory” needed in judging the resemblance between two objects?
It seems clear that theorising needs to include more than prototypes and exemplars, for example resemblance. Resemblance depends on other knowledge. The idea is that we each have an individual theory about what characteristics are important in deeming something as part of a category (for example, we may presume that someone jumping into a pool at a party is drunk, though jumping in a pool is not a prototype for a drunk person). This is because we each have a theory of drunkenness. We can also draw on theories when thinking about new possibilities for a category (eg could a wooden plane fly?) a
What’s different between your informal, usually unstated theory of artifacts and your theory of natural kinds?
Theory of natural kinds (groups of objects that exist naturally in the world such as bushes, animals or stones) are as they are because of nature and the properties of these objects are relatively stable.
Theory of artifacts (objects made by humans) are different because we could change certain properties if we wanted to (eg. Give a chair 7 legs). Therefore people will reason differently with categories that come under natural vs artifacts (eg children believe that a toaster could be made into a coffee pot, but not that a raccoon could be changed into a skunk). The diversity of concept can also be explained through goal-derived categories (diets foods) relational categories (rivalry and hunting) and event categories (shopping and visit).
What does it mean to say that knowledge can be represented via network connections?
fMRI scans show that different brain sites are activated when people are thinking about living things in comparison to non living things, and different parts are activated when people are thinking about manufactured items vs natural items. In some brain damages cases, people lose the ability to name certain objects or to answer simple questions such as if whales have legs - this is termed anomia. This shows that even the brain separates certain categories. This carries into sensory and motor areas too. For example, when presented with the word kick, we detect activation in the brain areas that control the movement of our legs, and when someone is presented with the word rainbow, we detect activation in the brain areas usually associated in colour vision. This suggests that conceptual knowledge is intertwined with knowledge about the processing of an objects sensory or motor functions. Some theorists argue “embodied” or “grounded cognition”, which is that the body’s sensory and action systems play an essential role in all our cognitive processes. When participants do a sentence verification task, the obviously true sentences were a lot faster then the not obvious sentences. This is presumably because the longer the activating nodes have to travel to find a connecting mode, the longer the answer will take. This theory was argued by Collins and Quillian. More recent data adds some complications to this because of the prototype theory. However, even with these complications we can still generally predict the speed of knowledge access by counting the number of nodes participants must travel through when answering a question, showing that associative links play a pivotal role in knowledge representation.
What is a propositional network?
To represent the full fabric of our knowledge we need more then just simple associations. We need to be able to decide the difference between “Sam is a dog” and “Sam has a dog” because if the only association we had were the nodes SAM and DOG, we would not be able to tell the difference between the two sentences. One widely endorsed principle which solved this focuses on propositions - defined as the smallest units of knowledge that can either be true or false. For example, “children love candy” is a proposition, while “children” is not. Anderson’s model represents through arrows each nodes role within the proposition.
Why do distributed representations require distributed processing?
Local representations: each node represents a different idea so when you’re thinking about that particular idea that node is activated. In contrast, connectionist networks rely on distributed representations in which each idea is represented, not by a certain set of nodes but by a pattern of activation across the network. In a distributed network, when asked about what type of computer you have to manage to collectively activate the many nodes representing computer to activate the ones representing MacBook . In short, a network using distributed representations must use processes that are similarly distributed, so that one widespread activation pattern can avoid a different (but equally widespread) pattern. In addition, the steps bringing this about must all occur simultaneously with each other so that an entire representation can smoothly trigger the next. This is why connectionist models are said to involve parallel distributed processing (PDP). Learning involves the adjustment of the connections among nodes.
What did 20th century philosopher Ludwig Wittgenstein argue?
That the simple terms we all use every day don’t actually have a definite definition.
