Handling inconsistency caused by a inadequate preferential model choice in uncertain preferences elicitation

In preference elicitation, the preferences of a user are collected in order to recommend one alternative. The user’s preferences are assumed to follow a specific preference model, modelled by a criteria aggregation function, which is chosen in advance by an expert. However, if the aggregation function is incorrectly chosen, the elicitation algorithm may recommend an alternative that is far from the user’s preferences. In this paper, we show that our elicitation approach based on possibility theory can detect a wrong model choice, and we study the consequences of changing the aggregation function during elicitation.