Concept Formation and Categorization

Learning Goals

• Define categorization
• Detail different categorization approaches and theories
• Analyze their strengths, weaknesses
• Explain the purpose of categorizing

Concept Formation and Categorization

• Concept: The basic building block of semantic knowledge (e.g., 'dog')
• Knowledge of concepts: Essential for interaction with the environment; allows predicting behavior/identifying properties
• Category: A class of concepts sharing common properties (e.g., 'animal' = organisms capable of voluntary movement)
• Natural Category: A group of entities existing in the natural world
• Artifact Category: A group of man-made objects designed for a specific function (e.g., 'computer')
• Nominal Category: A group of objects/ideas categorized by an arbitrary characteristic (e.g., 'positive numbers')
• Classical Theory of Concepts: Defines a category via defining features
  • Individually necessary: All category members have each feature
  • Jointly sufficient: Presence of all features makes something a member

Classical Theory of Concepts (continued)

• Example: A bachelor is an adult male who is not married
• Binary membership: An object is either a member or not a member
• Context-independent: Classification depends solely on the set of rules in the definition
• Hierarchy: Categories organized in hierarchical inclusion relations (e.g., biological taxonomy)

Empirical Data Supporting Classical Theory

• Study: Focuses on geometric shapes (color, size, number) to understand concept acquisition
• Simple Concepts: Defined by a simple attribute (e.g., "triangles")
• Conjunctive Concepts: Defined by several attributes (e.g., "square and black")
• Disjunctive Concepts: Require at least one attribute (e.g., "square or black") - These concepts often are the hardest to learn.

Problems with Classical Theory

• Difficulty in identifying necessary and sufficient conditions for many concepts
• Difficulty in explaining how incomplete or inaccurate information leads to clear cut concepts

Prototype Theories

• Prototype: The most typical member of a category (e.g., German Shepherd for 'dogs') or a set of such members (e.g., golden retriever, etc)
• Similarity-based categorization: Matches a concept's features to the object's properties, creating a similarity rating.
• Characteristic Features: Features correlated with category membership, not necessary for all members
• Typicality Gradient: Some members are closer to the prototype than others

Evidence Supporting Prototype Theory

• People tend to list characteristic features, not defining ones
• Concepts have a gradient level of typicality (some members are seen as 'better' examples)
• Categorization speed is influenced by typicality
• Typicality ratings sometimes fluctuate based on context

Evidence Against Prototype Theory

• Certain theories suggest that categorization may use superficial attributes only
• Typicality may be a by-product of other categorization processes, not essential
• There are cases where categorization involves factors beyond typicality

Exemplar-Based Theories

• Exemplar: Memory representation (instance) of an external object
• Categorization: Matches an object to existing instances in memory, emphasizing similarity
• Potential for variability: Allows for variability in category instances
• Strengths: Capture the variability of category members, better accounts for inconsistent concepts

Explanation-Based Theories

• Common-sense explanations: Used to categorize objects based on underlying theories about the world
• Categorization: Object features are compared to those expected based on the explanation used
• Context-dependent: Categories can be formed based on specific explanations

Hierarchical Structure and Basic Level

• Basic level: The intermediate level (e.g., "bird") which offers a tradeoff: dissimilarity between categories and within-category similarity.
• Expertise: Basic-level categories differ based on individual expertise. Experts tend towards more specific categories (e.g. "robin," "sparrow")