In robust incremental elicitation, it is quite common to make recommendations and to select queries by using a minimax regret criterion, which corresponds to a pessimistic attitude. In this paper, we explore its optimistic counterpart, showing this new approach enjoys the same convergence properties. While this optimistic approach does not offer the same kind of guarantees than minimax approaches, it still offers some other interesting properties. Finally, we illustrate with some experiments that the best approach amongst the two approaches heavily depends on the underlying setting.