Research Directions and Programs
Research in preference handling studies computational methods for the elicitation, representation, aggregation, and satisfaction of preferences for computational tasks. These tasks not only include decision making, but also database querying, web search, personalized human-computer interaction, personalized recommender systems, e-commerce, multi-agent systems, game theory, social choice, combinatorial optimization, planning and robotics, automated problem solving, perception and natural language understanding, and other computational tasks involving choices.
This general research program is based on the idea that preference-based computational systems will produce results that are more satisfactory for humans and that can be more easily adapted to changing user preferences. However, there are other computational methods for decision making that do not use explicit preference models. They can nevertheless be analyzed from a decision-theoretic viewpoint as the decisions made by those methods may be compatible with a preference model. This perspective leads to particular research programs of studying those methods with concepts from preference handling. A first proposal concerns the usage of numerical machine learning methods for decision making and can be found below: