Preface

Preferences provide a declarative way to specify the most desired candidates among a large set of alternatives even if it is not known a priori which alternatives are feasible and which are infeasible. Preferences guide human decision making from the earliest child age (“which ice cream flavor do you want?”) up to complex private and professional decisions (“which house to buy?”; “which flight trip to choose?”). Traditional decision theory models preferences in form of (additive) utility functions and thus defines the optimization objective for economic and logistic decision-making problems. Although those numeric preference models have a high impact on nowadays economic and logistic decisions, they have several limitations. For example, it is not natural to elicit numeric preferences and weights. Artificial intelligence (AI) promises a new perspective on preferences and addresses certain limitations of traditional methods. In fact, AI brings in several knowledge representation formalisms that provide more qualitative and more natural preference models. Moreover, AI can understand other roles of preferences such as control of reasoning, belief and hypothesis formation, that may play an important role in human thinking. Based on this perspective, preferences recently gained a lot of interest in different subareas of AI such as qualitative decision theory, nonmonotonic reasoning, constraint programming (CP), and reasoning about action and time. The AAAI-2002 workshop on preferences in CP and AI allowed researchers from these different areas to draw a state of the art and to discuss topics of common interest. This special issue of Computational Intelligence on preferences contains the most interesting contributions to this workshop. It gives an overview on current preference handling methods in AI and poses new questions that encourage and inspire future research on preferences. In particular, this volume covers the following topics: