Eliciting user preferences through a guided design personalization process

In building industry sound methods are needed to determine the design options that respond to the requirements of the potential house buyers. A new method is presented that consists of two main concepts: a user centred design approach through a non-designers interface and Bayesian networks for learning the housing preferences and responding to design choices. In a design session the network is used to collect data about the housing preferences and to check for design inconsistencies. An experiment with the system in a housing project shows how the subjects responded to the system feedback on their design decisions.