Recommender systems aim at recommending some item as most appropriate for the user. This article introduces a new way of measuring appropriateness for the user: an item is among the most appropriate ones if it is among the preferred items of the user when considering all arguments in favor or against all possible items. I describe precisely this goal and describe what a recommender system aiming for that goal could look like, called an Argumentative Recommender. I also provide a way of measuring whether a recommender system has achieved the goal, which can be used to compare such recommender systems, and briefly outline a way of building such a system.