Master of Science Thesis
Department of Computer Science
University of Crete
October 1999

Abstract

Voting is perhaps the most commonly used decision selection procedure that communities have. Modern technological developments have permitted voting to be conducted electronically; computer scientists are also familiar with the concept of voting , since it is used to solve various problems- such as assuring the consistency of a distributed file system, or performing action-selection while trying to navigate a robot - that appear in various sections of Computer Science.
It has been proved mathematically that it is impossible to find a simple, "fair" - according to defined Fairness Criteria - and consistent procedure that could specify the result of an election in which more than two candidates perticipate ("Arrow's Impossibility Theorem"). This fact has led to the development of Voting Theory, which is the mathematical treatment of the process by which democratic societies or groups resolve the many and conflicting opinions of the members of the group into a single choice of the group. So far Computer Science has used relatively simple voting schemes, despite the complexity of Voting Theory and the fact that the definition of a voting system and of appropriate voting procedures imposes major influence on the result of a voting session.
The objective of this master thesis was to design, analyse and implement a quite general architecture, namely the Voting Conduction Architecture, that permits the conduction of voting by use of any appropriate voting system. The Voting Conduction Architecture is an architecture of multiple co-operating distributed software agents that represent real entities. These entities were traced and considered important for the conduction of voting, following the study of Voting Theory and of a generic, real-world voting paradigm. The Voting Conduction Architecture was designed according to the principles of object-oriented systems' design, and is an extensible architecture. It allows voting sessions to be conducted in parallel, and it can be used as an infrastructure for the conduction of voting by agent-based architectures, or other applications that may utilise voting. Image retrieval by content is suggested as such an application.

Supervisor:
Stelios Orphanoudakis
Professor
Computer Science Department
University of Crete