This is a major field of interaction between computer science and decision theory. Special focus will be given on issues such as the use of preference models in reasoning and inference methods, planning and revision as well as in machine learning. The design of automatic decision devices capable of real-time on-line decision making is a major challenge here as well as intelligent information retrieval and recommender systems. Argumentation theory, formal logic, preferential queries in data bases are expected to merge with more classic decision theoretical results. Specific subjects which are going to be considered include:
- Graphical Models for Decision Making. Graphical models in Decision Making have already been introduced through the use of Influence Diagrams in Bayesian Decision Theory. Further extensions have been studied in Artificial Intelligence in order to represent conditional preferences such as CP-nets and GAI networks. An important trade-off to take into account concerns, on one hand, the "expressivity" of the model used and, on the other, the possibility to have efficient algorithms to handle the model. Conjoint measurement theory is expected to provide an axiomatic basis for the study of such models.
- Computational Issues in Auction Design. Recent advances in information technology and its rapid acceptance by the business community have allowed for expediting of complex business transactions. The most prominent example involves use of auctions in corporate procurement and in government deregulation efforts. Auctions are important and unusually complicated games. In auctions, bidding functions maximising expected profit can be exceedingly difficult to compute and determining the winner can be extremely hard. More difficult problems arise in combinatorial auctions in which multiple goods are auctioned and bidders have to express valuations on which goods complement each other and which goods substitute for each other. Game theory and combinatorial optimisation have to be combined with the presence of partial information, limits to computation, and bidding incentives.
- Distributed Negotiation and Group Decision. Besides combinatorial auctions, which are centralised mechanisms for determining a suitable allocation of goods amongst bidders, it is also proposed to study distributed negotiation mechanisms, where allocations emerge as a consequence of stake-holders agreeing on a sequence of local deals. Like combinatorial auctions, such mechanisms raise problems that are combinatorial in nature and therefore require powerful preference representation languages. Negotiation may be seen as an interactive mean of aggregating preferences, while querying potential trading partners amounts, amongst other things, to preference elicitation.
- Large Databases and Inference. Several decision problems involve the use of very large data bases. Much of the data in these databases is in the form of sequences of information. It is often the case that sequences that occur frequently or are ``centrally located or form a ``consensus pattern are sought. Inferring a conclusion or identifying relevant data within a huge data base can be efficiently handled through the use of decision theoretic approaches based on the establishment of a ``consensus'' (about a similarity or a relevance) an issue studied extensively in social choice. Such approaches will be extended for the particular case of inferring conclusions from data bases, but also large knowledge bases.
- Graphical Models for Decision Making. Graphical models in Decision Making have already been introduced through the use of Influence Diagrams in Bayesian Decision Theory. Further extensions have been studied in Artificial Intelligence in order to represent conditional preferences such as CP-nets and GAI networks. An important trade-off to take into account concerns, on one hand, the "expressivity" of the model used and, on the other, the possibility to have efficient algorithms to handle the model. Conjoint measurement theory is expected to provide an axiomatic basis for the study of such models.
- Computational Issues in Auction Design. Recent advances in information technology and its rapid acceptance by the business community have allowed for expediting of complex business transactions. The most prominent example involves use of auctions in corporate procurement and in government deregulation efforts. Auctions are important and unusually complicated games. In auctions, bidding functions maximising expected profit can be exceedingly difficult to compute and determining the winner can be extremely hard. More difficult problems arise in combinatorial auctions in which multiple goods are auctioned and bidders have to express valuations on which goods complement each other and which goods substitute for each other. Game theory and combinatorial optimisation have to be combined with the presence of partial information, limits to computation, and bidding incentives.
- Distributed Negotiation and Group Decision. Besides combinatorial auctions, which are centralised mechanisms for determining a suitable allocation of goods amongst bidders, it is also proposed to study distributed negotiation mechanisms, where allocations emerge as a consequence of stake-holders agreeing on a sequence of local deals. Like combinatorial auctions, such mechanisms raise problems that are combinatorial in nature and therefore require powerful preference representation languages. Negotiation may be seen as an interactive mean of aggregating preferences, while querying potential trading partners amounts, amongst other things, to preference elicitation.
- Large Databases and Inference. Several decision problems involve the use of very large data bases. Much of the data in these databases is in the form of sequences of information. It is often the case that sequences that occur frequently or are ``centrally located or form a ``consensus pattern are sought. Inferring a conclusion or identifying relevant data within a huge data base can be efficiently handled through the use of decision theoretic approaches based on the establishment of a ``consensus'' (about a similarity or a relevance) an issue studied extensively in social choice. Such approaches will be extended for the particular case of inferring conclusions from data bases, but also large knowledge bases.