Welcome to CIS767 (Decision Support Systems)! This is the course's Spring 2001 entry page. This page contains the following information:
syllabus of the course and a course outline;
and this link to the dedicated CIS767 'ad valvas' page.
Important notice : please consult the
CIS767 ad valvas page regularly to check for announcements
(such as changes in the course schedule, new assignments, exam dates).
Course Summary: The main concern of this course is the role of information, preferences and their interaction in decision making activities of individuals and groups. We explore these topics along three themes:
The theme on decision analysis techniques provides basic models for dealing with information and the judgments of a decision maker. We also have a look at the many fallacies of human judgment. Second, we introduce the basics of fuzzy information engineering to deal with qualitative information, and apply this knowledge to information retrieval problems for collaborative filtering and preference modeling. Third, we investigate group decision and consensus models. The exchange of information in decision making groups deserves special attention, and finally apply our decision analytic knowledge obtained so far to the analysis of social networks.
Background: This course requires a knowledge of elementary mathematics and probability theory.
Book: There is no required book for this course. Throughout the course, a number of books will be brought to your attention.
Course Outline - Thematic View
Theme 1: Decision analysis and human judgment
The first theme's objective is to present an overview of fundamental decision analysis techniques. The very first lecture provides an introduction to this course. We motivate why the topics below are treated in this course, and how they relate. We discuss the basic terminology used throughout the course, and come to understand terms as "decision analysis", "decision aid", "decision making", "normative models", "judgment", "measurement scales", etc. The discussion of decision making under uncertainty (another such term we need to define) is the topic of the second lecture. We review the Bayesian framework, blended with examples of decision trees and influence diagrams, and focus on how we can assess the value of (a priori unknown) information in this framework. Uncertainty of the probabilistic type (yes, there are other types of uncertainty, and we come back on this within the second theme of the course) is left aside in the discussion of multi-criteria decision models in the next two lectures. We review multi-attribute value theory and applications, and the outranking methods which may lead to a partial ordering of the decision alternatives. The way our mind works in making decisions has been studied for more than half a century now. This research has revealed that we often use unconscious routines, or heuristics, to cope with the complexity inherent in most decisions. These heuristics, however, are not always flawless. We discuss judgment biases in the last lecture of this theme to expose a number of important psychological traps in decision making.
Theme 2: Fuzzy information engineering
Historically (-- if we may use this term to describe the rather short history of fuzzy sets since 1965 --), fuzzy sets have been mostly noted for their ability to model qualitative or linguistic categories. This is due to the ability of fuzzy sets to represent gradedness of concepts, a reality that is both in a decision maker's mind and the physical world. Zadeh, the founder of fuzzy set theory, considers fuzzy sets as tools for "computing with words". In addition, many of the basic notions in decision analysis are naturally graded, such as uncertainty, preference, similarity. In this theme, we first outline fuzzy set theory and then examine how this theory helps us in the analysis of decisions. The first and second lecture provide a rather thorough basis to the basic elements of the theory: we discuss in detail fuzzy sets and fuzzy relations. This allows us to devote two lectures two applications of fuzzy sets: the formulation of preferences in lecture 3, and similarity relations and their use for collaborative filtering in the fourth lecture of this theme.
Theme 3: Group decision making
In very general terms, the central problem of group decision making is to define "fair" methods for aggregating individual choices to yield a social decision. The first lecture of this theme is devoted to a crucial result in social choice theory: Arrow's impossibility theorem. We examine various voting schemes (de Borda and Condorcet, most importantly), and show where these widely used methods fall short in terms of Arrow's work. Reaching consensus is a key objective in many decision making groups. In the second lecture, we examine a number of consensus models, analyse their underlying modeling assumptions, and try to discover their possibly biased impact on a group's decision. One reason for which groups are often formed is the premise that the exchange of information by a group's members may lead to more efficient decisions. We examine a well-known group information sampling model, and discuss extensions of this basic model. We also take the opportunity to highlight basic issues in information theory. Finally, we focus on the analysis of (social) networks. Social networks form as individuals (e.g., decision makers) establish and maintain relationships. Yet being connected not only yields benefits, but also conveys costs -- two familiar decision criteria. We aim to apply some of the basic decision analysis techniques we have seen so far to the analysis of these networks.
Last update: January, 5th 2001.