Economic rationality and neural networks.
Item
-
Title
-
Economic rationality and neural networks.
-
Identifier
-
AAI9325095
-
identifier
-
9325095
-
Creator
-
Garavaglia, Susan Berger.
-
Contributor
-
Adviser: Peter S. Albin
-
Date
-
1993
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Economics, General | Economics, Commerce-Business | Economics, Labor | Artificial Intelligence
-
Abstract
-
Artificial Neural Networks are constructed by linking primitive computing structures with weighted connections. The information content of a neural network develops through the adaptation of the connection weights by learning algorithms to meet some measurable criterion. This adaptability, combined with parallelism and distributed information processing, produce a highly effective paradigm of a decision making organization. Higher degrees of organizational complexity are represented through a larger number of primitive units and connections which can be decomposed into any meaningful subdivision. Because of the efficiency and modularity of this representation of a decision making entity, a neural network exemplifies Simon's procedural rationality.;To establish foundations, an overview essay includes a survey of economic applications, including predictive econometric models, classification, and behavioral models. Specific examples of neural networks developed to perform estimation of non-linear functions, discriminant analysis and clustering, and pattern recognition, are discussed. This essay also discusses some of the key statistical and mathematical properties of neural networks, reviewing the work of Halbert White and others.;The principal justification of the application of neural networks to economic problems is delivered in the context of procedural rationality (Herbert A. Simon). The use of neural networks to reduce the complexity of a model is demonstrated and compared with Peter Albin's work on complexity measurement and organizational committee structures. In addition, the theme of biological analogy is examined in the work of Kornai and others. An economic system is often compared to a biological organism to explain its structure and behavior. One essay demonstrates the application of a neural network to modeling Akerlof's reciprocal gift exchange theory.;A new neural network model is proposed for application to rational decision making in organizations, based on R. K. Sah's and J. E. Stiglitz' work on optimal consensus size in committees and hierarchies, and Nils Nilsson's committee machine. This neural network learns the optimal size of a consensus in a committee and the characteristics of good decisions by observing the odds ratio of bad to good options. The behavior of the network is instructive in understanding decision-making behavior in large organizations.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
