This paper argues that fake news is a consequence of some flaws of the media system and suggests a direction to deal with this phenomenon. We begin by observing that the traditional media are regulated by two competing systems of values, truthfulness and proximity, and that the second is factually more important than the first. We argue that social media has maximised proximity and hence, has completed the separation of trust and truth. We propose to call "trust-conditional evaluation" the model of truth under which the social media operates today. We show that such a model eventually leads to idealistic and relativistic consequences, unless it is constrained by a pragmatic attitude. We suggest then a pragmatic constraint: Statements can be said to be true only relatively to an evaluation effort, therefore the truth value of a statement corresponds to the cost of falsifying it. We conclude by discussing some potential implementations in the design of algorithms using trust ties in social networks in order to evaluate statements.