Abstract
Constructive recommendation is the task of recommending object “configurations”, i.e. objects that
can be assembled from their components on the basis of the user preferences. Examples include: PC
configurations, recipes, travel plans, layouts, and
other structured objects. Recommended objects
are created by maximizing a learned utility function over an exponentially (or even infinitely) large
combinatorial space of configurations. The utility
function is learned through preference elicitation,
an interactive process for collecting user feedback
about recommended objects. Constructive recommendation brings up a wide range of possible applications as well as many untackled research problems, ranging from the unprecedented complexity
of the inference problem to the nontrivial choice of
the type of user interaction