This thesis presents a new framework and a tool for decisionmaking using uncertan information. It is called DELTA after itsprimary decision rule. The objective is to present a method forevaluating choices under uncertainty. The nature of mostinformation available to decision makers is vague andimprecise, be it information for human managers inorganisations or for process agents in a distributed computerenvironment. In spite of this, many traditional models fordecisions disregard this state of affairs. Some more modernapproaches address the problem of vagueness, but many of themconcentrate more on representation and less on evaluation. Theemphasis in this thesis is more on evaluation, and therepresentation used is that of standard probability theory.
The approach is interdisciplinary and draws on ideas fromthe areas of operational research and artificial intelligenceas well as from statistical decision theory. It allows thedecision maker to be as deliberately imprecise as he feels isnecessary as well as providing him with the means forexpressing varying degrees of imprecision in the inputstatements. This leads to a more natural relationship betweenthe decision maker and the support tool. The purpose of themethod is not to replace decision makers with machines. On thecontrary, an objective is to increase the decision maker'sability to make sound decisions.
A longer summary and updated versions of the core chaptersof the thesis are available athttp://www.dsv.su.se/~mad/thesis.htm1
KEYWORDS:Decision Analysis, Decision Support,Multi-Agent Systems, Uncertain Reasoning
Kista: Data- och systemvetenskap , 1997. , 224 p.