Lecture 8 - Knowledge & Problem Solving Flashcards
Memory vs knowledge
Memory allows us to build knowledge as we learn from past experience
Learning and memory are closely related
Learning is the acquisition of skill or knowledge while memory is the expression of what we’ve acquired
Knowledge is the possession of information or the ability to locate it
Memory is part of learning, the ability to retain knowledge in the brain
Conceptual knowledge
Concept: a mental representation used for a variety of cognitive functions
(Example: what the essence of a cat is)
Conceptual knowledge: enables us to recognise objects and events and to make inferences about their properties
(Example: what you know about cats)
Categorisation: the process by which things are placed into groups called categories
Categories are all possible examples of a particular concept
(Example: cats are categorised as living things, mammals, pets, things we love)
Why are categories useful
Knowing something is in a category provides a great deal of information (pointers to knowledge)
Help to understand individual cases not previously encountered
Provide a wealth of general information about an item
Allows us to identify the special characterstics of a particular item
Example: you may not have a concept of a liger, but you know that it is a large feline (category) so you know it is an animal, predator, carnivore, likely dangerous, lazy, has whiskers, excellent hearing and smell, probably roars instead of meows and you should not get close
What makes a category
Not all members of everyday categories have the same defining features
Determined category membership based on whether an objec t meets the definition of a category does not work well
Instead of only relying on strict definition criteria items in a category resemble one another in a number of ways, such as family resemblance
How do we categorise concepts?
The prototype approach
Prototype: an average representation of the “typical” member of a category
Characteristic features that describe what members of that concept are like
An average category members encountered in the past
High prototypicality:
A category member closely resembles the category prototype - typical member of the b ire category = robins
Low protypicality:
A category member that does not closely resembles the category prototype - bird = ostrich
Differs among cultures and geography
There is a strong positive relationship between prototypicality and family resemblance
- items in a category that have a large amount of overlap have high family resemblance
Typicality effect: prototypical objects are
- processed preferentially
- processed more rapidly
- named for rapidly
- more effected by priming
The exemplar approach
A concept is represented by multiple examples (rather than a single prototype)
Examples are actual category members (not abstract averages)
To categories we compare new item to stored examples
Similarity to prototype view: representing a category is not defining it
Difference: representation is not abstract
The more similar a specific examplar is to a known category member, the more it will be categorised - family resemblance effect
Examplar vs prototype
Examplar:
- explains typicality effect
- easily takes into account atypical cases
- easily deals with variable categories
Prototype approach:
- fast and efficient
- facilitates categorisation
- easily deals with variable categories
Reality:
We probably use both (simultaneously and altering)
Examples may work best for small categories
Prototypes may work best for larger categories
Hierarchical organisation of categories
Three levels
Basic level (in the middle) is “psychologically privileged”
Going above basic level -> large loss of information (furniture vs table)
Going below basic level -> little gain of information (surgery theatre preparation table is still a table)
Semantic networks
Concepts are arranges in networks that represent the way concepts are organised in the mind (Collins and Quallian 1969)
- hierarchical model
- node = concept/category
- concepts are linked
- model for how many concepts and properties are associated in the mind
- bridges computer models of knowledge
Cognitive economy: shared properties are only stored at higher-level modes: exceptions are stored at lower nodes
Semantic dementia
Progressive neurological disorder in which people lose specific knowledge first and loss of memory follows the hierarchy from specific to general
Gradual disintegration of concepts and categories
Follow as opposite direction as in which children acquire knowledge
Cortical atrophy in semantic dementia
Selectively affects temporal lobes
Leads to progressive loss of
- word memory (mental lexicon)
- semantic categories (knowledge/recognition)
Spreading activation
Activation is the arousal level of a node
When a node is activated, activity spreads out along all connected links
Concepts that receive activation are primed and more easily accessed from memory
Activation spreading through a network as a person searched for a word (e.g. from ‘robin’ to ‘bird’)
Criticism of the Collins and Quillian model
This model cannot explain typicality effects
Cognitive economy?
Some sentence-verification studies have produced results that are problematic for the model
Solution: the connection approach
Originated in creating computer models for representing cognitive processes
Uses parallel distributed processing
Knowledge is represented in the distributed activity of many units
Knowledge can be activated by external stimuli and signals from other units in the knowledge system
Weights determine at each connection how strongly an incoming signal will activate next units
Advantages to the connection approach
Similar to human learning process, can explain how learning occurs (how humans build conceptual networks in our minds)
Training systems to recognise properties of one concept provides information about relates concepts (semantic networks, categorisation)
Can explain differences in typicality similarly to prototype models
Explains generalisation of learning
Can explain changing knowledge strcutre over time
Performance distribution occurs gradually as parts of the system are damaged. Network function not totally disk types by damage similarly as in semantic dementia
Very similar to brain - can model cognitive functions
