Concepts / Categorization

More like this

ahmed hussain IT

by Ahmed Hussain Salih

21st Century Social Studies

by Cynthia Tittle

Input output devices

by Doris McKnight

Google Data Studio Features

by Amit Puar

Concepts / Categorization

Neurological Evidence

With Alzheimers: Naming Superordinate: Horse--> Animal

Specific deficits for

body parts

Fruit & Vegtables

Living things & Food

Connectionist Explanations

IAC model: Can learn from specific instances and link comon characteristics: Links are knowledge / instances are strength of connections

Psychological essetialism (essential and constraining prop)

Con:

Braisby: Experts say it is a salmon: 25% modify their opinion to conform w. experts

Function and appearane are imp as well

Malt: We are essentialist for natural categories

Boat-Ship: Call it whatever

Trout-Bass: Ask an expert

Braisby: Is a cat still (essentially) a cat a cat even when robot from mars?

Concepts dirven by context and content

50% both true, fals

50 "Yes, cat

Malt: People know, the essece of water is H20

Categorizatio also driven by function, location socio-historical context

Pond is water although judged to contain only 63% H20

Tears are not water although rated to contain 87% H20

Pro

Gelmann and Wellman: a dog still dog if inside is taken (4-5 yrs: no)

Difference to classical view

A placeholder can change

Filling the placeholder is job of science

If we are not shure: we create a placeholder

We have a belief about essential prop

Comon-Sense "Theoroy"

Top down: with more complex situations we have to think

Bottom up: Facilitated by similarity

knowledge involved - not mere lists

Its rather knowledge than theory

Underspciefied: How are complex categories/ theories combined?: pet fish

Good because it hints to problems with "similarity"

Developemental Evi: Kiel: characterisic to defining shift

Murpy: They dont know about biological categories

Racoon disguised as a skunk

Looks and Behaves like Zebra: 4 yrs-zebra 7 yrs-horse

Kroska and Goldstone: 2 emo Szenario categorzed as fear but more similar to joy

Rips Pizza Dissotiation: More similar to quater but more likely (=categorized) a pizza

Problem with similarity: What properties are looked at? (Plumes and Lawnmowers-> knowledge makes the difference)

Rich Internal Structure

Typical members (Prototype views)

Demarcatino Line (Classical View)

Prototype theories (graded, some more typical than others)

Prototype: Other members determined by comp with typical member

Evaluation

Unclear Definition: Are prototypes lists or typical members?

Family resemblence often predicts typicality scores

Explains many categories that lack clear definitions (game, furniture)

Semanitc Transfer?: Complex Concepts Lead to wrong interpretaions in real life: pet-fish = dog trout

Barsalou: goal derived categories have no family resemlance: presents that john likes

Medin & Shoeben: typicality context dependent: kaffeelöffel paradox: large wooden>small wooden

Implies we only use lists of attributes but we also use reason: blue bird=probably "warrum," but fat man is prob not a "klatau"

Hampton: Some astract concepts (belief) has no prototype

Pros

Roesch and Mervis: Robins share more properties with other category member than penguins

Armstrong: defined cat seem to have typiclity: female: mother vs. policewoman

Dual Proces: conept core to judge generel membership, prototype to evlauate instances

Mervis & Roesch: < RT faster and less errors for typical examplars

How we do it?

Calculate family resemblecne: one shared by 16, one by 14=30

high typicality instances match on high weighted values

Statistacl distribution determines weight

More similar features - quciker match

comparing of features with stored rep

What is typicality: Having moreo cue valid features in common with prototype

Weighting=Cue Validity= Some features more important than others e.g. bird=feathers

In some natural objects, attributes cluster together

Categories are not single thing or process

Categorizers differ

Novices: Similarity Driven

Experts: Definition driven

Categorization differ according to action / purpose

Explanation based

Essentialism: Consistent with (current) expertise

Theory: If thing need explanations: This guy is intoxicated: Thats why he jumps the pool

Decision based

Classical: Exact definitions needed: law

Prototype: Fast and Superficial: Fast decisions under uncertainty

Smith & Sloman: Two routes: Similarity or Route based depending on the task

Barsalou: internal structure not fixed but viewpoint and task driven

Types of Categories differ

Fuzzy (red)

Well defined (a triangle)

Classical View (all or none), members are determined by a definition

Cons

Everyday Concepts fuzzy not exact

Borderline Cases

McCloskey & Glucksberg: yes/no

Where does red turn into orange

Possilbly: Lack of knowledge

Typical items: chair= furniture: high over time and people

bookends =furniture?: changes across people and time

Intransitivity: big ben = clock = furniture

Bad RTs: RTs not a good predictor

is a dog an animal < Is a dog a mamal

Varying Knowlege: Experts vs. Normal people

Does not explain Typicality effets

Some things dont have definitions - more holistic

Cats would still be cats even if turs out they are robots from mars

Some Categories have no def. features

Example: Furniture / chair

Wittgenstein: Games are similar

Typicality more important than defining features:

Defined by functional /not defined by pyhsical properties

No neccessary and sufficient properties for all chairs

Pros

Collins & Quillian: Categorization takes longer for 2 levals

A canary is an animal > A canary is a bird

Bachelor: male, unmarried, adult

Hierarchical. Food= Fruit or Vegatables

Transitivity: A=B= C and C= A

Inheritance: Bird inherits all from animal but adds feathers

1. Some propertes in beginning / Most at end

Generality claims

Everybody represents in this way

All concepts are represented this way

Members

are equally good

share a) neccessary and b) sufficient features

All or none are members

Divides the world in distinct classes: taxonomies

Methods

Connectionist models

Neuropsychology

Attribute Lising

Card Sorting

Is

Categorizations not only natural but depend on purpose: "Things to take in case of fire"

concept provides Meaning: Mental Lexicon

A level of recognition

Familiarity: lov level

Semantic Classifacation

Naming

Compression and facilitation

To guide action

Predictions of behaviour

Makes remembering more easiy: All cats have hearts

Mental economy

Mental representations

Forms of reps

abstract, by relations

by Prototype

by Rule / defintion

LTM (Concepts are basic units)

Aquired by autobiographical experience

Procedural Memory

Schemas

Rule based "What to do in a Restaurant"

Semantic Memory

Propositions: "Dogs" bark, Dogs have fur, Dogs are pets