THE ORGANIZATION OF INFORMATION IN TERMS OF ITS USE: A DEONTIC MODEL OF KNOWLEDGE REPRESENTATION (COGNITION, CATEGORIZATION, ARTIFICIAL INTELLIGENCE).
Item
-
Title
-
THE ORGANIZATION OF INFORMATION IN TERMS OF ITS USE: A DEONTIC MODEL OF KNOWLEDGE REPRESENTATION (COGNITION, CATEGORIZATION, ARTIFICIAL INTELLIGENCE).
-
Identifier
-
AAI8629729
-
identifier
-
8629729
-
Creator
-
RIFKIN, ANTHONY JOSEPH.
-
Contributor
-
Katherine Nelson
-
Date
-
1986
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Psychology, Experimental
-
Abstract
-
The representation of featural information for natural categories should be based on how features are most commonly asserted as descriptions, rather than in terms of fixed relations to the categories and instances they can describe. Such general "norm" relations would not be invalidated if they are not instantiated in particular contexts, and the use of features could be adapted for any number of contexts. The present study investigates a deontic model of representation in which features bear general "norm" relations to categories and their members, and do not formally delimit category boundaries or the instances they can describe.;Two experiments identify these relations. In one, the more typical instances of a category are found to be classified more often as belonging to these categories, while less typical instances are more frequently classified as belonging to contrasting categories. In the second experiment, it is found that features that can describe all members of a category ("Obligatory" feature norms) are used to define the memberships of the categories' more and less typical instances. In line with the instances' family resemblances, the membership of less typical instances in contrasting categories are defined by features that are "Impermissible" as membership definitions in the target categories, and the membership of more typical instances in the target categories are defined more often by "Permissible" feature norms that describe some, but not all, of the categories' members.;A third experiment examines the context-sensitive retrieval of this information, when instances are compared and the categories they belong to are not mentioned. It is found that descriptions of the instances' similarities are based on their least-upper-bound shared membership, and that descriptions of their differences are generated from the level immediately below their least-upper-bound membership. A taxonomic organization is therefore identified from within which the feature norms are generated as descriptions. Contexts are also identified in which instances are excluded from the description of the same norms, and in which instances are described by the categories' norms while being excluded from the categories. In addition, the deontic model is used in a computer implementation demonstrating how these descriptions are generated within particular contexts.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
-
Program
-
Psychology
