Ch7 Knowledge Representation

implicit learning

(aka nonanalytic concept formation) requires that people pay attention to individual exemplars, storing info about representations of them in memory

issues with Collins and Quillian's semantic network

-accounted for too little -inheritance -predictions not connected equally

issues with Collins and Loftus

-accounted for too much -impossible to test; no way of knowing

Prototype View Implications/Characteristics

-denies existence of necessary-and-sufficient features of lists -features/aspects are characteristic (typical) of members of the category -family resemblance structure of concepts

major characteristics of schemata

-exists in all levels of abstraction (can exist for small parts of knowledge) -active processes (people are thought to be constantly assessing and evaluating the fit between their current situation and a number of relevant schemata & subschemata)

Problem with Scripts

-other, script-related info intrudes into our memory

working memory

-part of declarative memory -most activated part of declarative memory

Exemplar View [of Concepts and Categorization]

-similarity based -assumes we store specific individual instances and use these stored representations to categorize

Schemata/Scripts View [of Concepts and Categorization]

-similarity-based -explanation-based -regards concepts as schemata, packets of information with specific parts, that fill in default values for aspects of the situation

Classical View [of Concepts and Categorization]

-similarity-based -features represented are... 1) individually necessary 2) collectively sufficient

Prototype View [of Concepts and Categorization]

-similarity-based -we categorize objects by comparing them to mental abstractions, called _________, which are idealized representations of some class of objects/events

Prototype View Problems

-too unconstrained -doesn't always lead to right category (i.e. hyenas ≠ kind of dog, = felines)

Exemplar View Problems

-too unconstrained (like prototype) -fails to specify which instance will eventually be stored as exemplars/not -doesn't explain how different exemplars are "called to mind" at the time of categorization

scripts

-what usually happens (action) -a schema for routine events (knowledge of occurring situation cues you for how to behave appropriately)

5 Views of Concepts and Categorization

1) Classical 2) Prototype 3) Exemplar 4) Schemata/Scripts 5) Knowledge-Based

types of ACT models

1) adaptive control of thought (ACT) model of memory

types of network models

1) cognitive economy 2) hierarchical semantic network model of semantic memory

Classical View Implications

1) concepts mentally represent lists of features 2) assumes that membership in a category is clear-cut.... (lacking even one feature disqualifies for membership) 3) all members within a category are created equal......(i.e. no such thing as a "better"/"worse" triangle)

3 types of models for organizing knowledge

1) network models 2) ACT Models (adaptive control of thought) 3) connectionist models

kinds of concepts

1) nominal-kind 2) natural-kind 3) artifact

two major types for categorization (classifying the 5 distinct views of concepts & categorization)

1) similarity-based 2) explanation-based

Brooks (1978) 5 factors that encourage people to store info about individual exemplars...

1) task requirements 2) nature of stimuli (complex/simple) 3) original learning situation 4) multiplicity of belonging 5) uncertainty about application

Classical View Problems

This view has no way to explain people's intuitions that some birds are "birdier" than others.

category

a grouping of items sharing one or more similarities; a class of similar things that share either an essential core or functional properties

family resemblance structure of concepts

a structure in which each member has a number of features, sharing different features with different members; the more features a member possesses, the more typical it is

similarity-based categorization

classical, prototype, exemplar, schemata/scripts includes approaches in which categorization is assumed to be based on the similarity of an instance to some abstract specification of the category or to one or more stored exemplars

Collins and Loftus

cognitive economy ≠ hierarchical structure; based on experience -shorter link [based on experience] -spreading activation -typicality effect

Just and Carpenter (1987) on a schema...

compared to a questionnaire with blanks that a person is supposed to fill in... labels next to the blanks indicate what sort of info to fill in (i.e. name, address, d.o.b.)

nature of stimuli complex vs. simple

complex: often, the relevant dimensions of variation are not apparent to the novice simple: stimuli may only differ in a small number of very obvious ways

nominal-kind concepts

concepts that have clear definitions (ex. of tiger jumping through flaming hoop---> hoop)

adaptive control of thought (ACT) model of memory

developed by John Anderson specifying a networked memory comprised of working memory, declarative memory, and procedural memory.

Knowledge-Based View [of Concepts and Categorization]

explanation-based people use their own theories to guide their classification of objects; categorization based on top-down knowledge of how interaction occurs (--> What binds things in one category?)

uncertainty about application

in natural settings, we learn about instances without knowing how we will be called on to use the info later

multiplicity of belonging

instances may belong to a number of categories all at the same time

concept

mental representation of a category; allow us to categorize, giving us mental "buckets" in which to sort the things we encounter, letting us treat new, never-before-encountered things in the same way we treat familiar things that we perceive to be in the same set

studies of artificial intelligence

most of our info about organization of info in LTM comes from...

hierarchical semantic network model [of semantic memory]

organized in terms of nodes and links that stores properties at the highest relevant node to conserve cognitive economy

network models

posit that different concepts or ideas are mentally represented as nodes, with connections among related ideas along which mental activation can spread

ACT models

posit the existence of nodes; they further postulate the existence of production (if-then) rules used to represent info in procedural memory

back propagation

procedure used to develop patterns of activation through many trials with training examples

cognitive economy

properties and facts are stored at the highest level possible; hierarchical structure, involving inheritance

connectionist models

represent concepts/ideas by a pattern of activation across different units -bunch of connected nodes -some nodes make others more/less likely to fire -nodes activate all at once (distribution)

exemplars

representations of previously stored instances

procedural memory

represents info in "production rules"

explanation-based categorization

schemata/scripts, knowledge-based sees people as classifying instances based on meaningful relationships among instances and categories

Collins and Quillian's semantic network

semantic memory = analogous to a network of connected ideas; consists of nodes, each of which are connected to related nodes by means of pointers, or links that go from one node to another

declarative memory

stores info in networks containing nodes

original learning situation

the same instance may appear repeatedly, affording the person a chance to get to know certain instances very well

artifact concepts

things constructed to serve some function or accomplish some task (ex. of tiger jumping through flaming hoop---> flames)

natural-kind concepts

things naturally occurring in some environment (ex. of tiger jumping through flaming hoop---> tiger)

task requirements

to learn info that distinguishes among individual instances

Which has a faster retrieval.... typical or atypical instances?

typical

schema (schemata=pl.)

usually refers to something larger than individual concept; what's usually present/fitting characteristics or features;

Rumelhart and Ortony (1977) view of schemata...

viewed as the fundamental building blocks of cognition, units of organized knowledge analogous to theories schemata= "packets of info" that contain both variables and a fixed part.