State of the art recommendation techniques still cannot fully explain and predict the information needs of the user while is searching for items such as news, services or products. In fact, the specific ephemeral needs of the user, the context of the search, and the context of items' usage, do influence the user's response to, and evaluation for items. Hence, RSs should take into account this information to deliver recommendations that users would judge as appropriate to their situations. Context modeling and context-dependent reasoning is a complex subject and there are still major technical and practical difficulties to solve: obtain sufficient and reliable data describing the user preferences in context; selecting the right contextual information, i.e., relevant in a particular personalization task; understanding the impact of the contextual dimensions on the user decision making process; embedding the contextual dimensions in a recommendation computational model. These topics will be illustrated in the talk, making examples taken from the recommender systems that we have developed.