Concepts and Categories Part 1

A set of things in the world (usually regarded as having particular shared characteristics)

Concepts are mental constructs a hypothesis or description of what those shared characteristics are of a category a stored generalization A concept doesn’t need a word/linguistic label

A label for a concept

Generalization Inference Prediction Composition

We have percepts that are simple and indivisible We have concepts that are complex and divisible

A concept is a definition The features associated with the concept are individually necessary and jointly sufficient to assign an observation to membership in a category

Easy to find similarities between concepts Allow you to see opposites Allows you to see superordinate - concept that has a subset of the features of the subordinate Allows you to interpret phrases though lawful union of features

It faces the problem that some concepts do not inherently have a singular definition that encompasses the concept in its entirety EX. game

The idea that a concept is characterized by a central or idealized representation of the category members (prototype) Your speed and accuracy for categorization are determined by distance to prototype Allows graded membership in categories

Production tasks (list members of a category) Rating tasks (rate members of a category) Sentence verification Picture identification

They are vectors Word-word co-occurrence matrix And the distance between the vectors of the words relates to the frequency, rating and typicality present when referring to the prototype An embedding for a word is just a vector (list of numbers) that represent its meaning, and capture something about its similarity to other words.

The average misses many key facts about variability

A newly encountered item brings to mind other stored exemplars in proportion to their similarity to the new item Evidence that supported prototype can also be explained with exemplar theory

Representation of prototypes and exemplars Consider category learning effects Rely on exemplars early in training and rely on prototype later in training