Preference-based search for Configurable Catalogs

We consider the problem of preference-based search for configurable products. We focus on example-based tools that interactively show both candidates (best options given current preferences) and suggestions to stimulate preference expression. Suggestions are generated to maximize the chance of eliciting further preferences using our lookahead principle. We present Probabilistic Constraint Optimization problems (ProbCOPs) that allow to reason about the uncertainty of the preference models and show how the generation of suggestions can be formulated as a single optimization problem.