Learning own preferences through consumption

This paper provides theoretical foundations for preference discovery theory. We propose to relax the assumption that the consumer has perfect knowledge of their own preferences, so that the consumer knows only the subjective probability of those alternatives being in any given relation, which is conditional on the information available to the consumer. To achieve that, we construct probabilistic measures on the space of all permissible preference relations and consider the consumer to be equipped with one such measure, instead of a preference relation. These measures are intrinsically linked by construction to the information structure available to the consumer and allow for indirect learning. We visualize how these measures correspond to the choices of the consumer, we consider three distinct decision procedures. These procedures formalize how under different assumptions regarding the underlying probability measure, the consumer guesses their own tastes. Finally, we use these measures to define the value of the information provided by the consumption of a chosen alternative and study the properties of the preference ranking induced by it.