User control in recommendation systems

At present where there are large volumes of information, recommendation systems are needed to help users find and evaluate articles of preference or interest. A number of investigations in this domain suggest that “best” recommendations, according to objective metrics, are sometimes not the ones that are most satisfying or useful to users. However, if the system assumes that user preferences are not in line with the recommendation made, mechanisms should be provided to put the user in control of the recommendations. The article investigates how to give control to the users in the systems of recommendation and the quality of these from a perspective centered on the user. We discuss a study that involved 4 scientific investigations related to the subject and consider parameters such as the type of recommendation system, the method used, the application area and the results obtained in those investigations. Which focuses on the accuracy and novelty of the recommended articles, and on the general satisfaction of the users. We have classified the recommendations considered with respect to these attributes and compared these results with measures of statistical quality of the algorithms considered. It is intended to generate new recommendations based on new preferences that could lead to greater user satisfaction and confidence in the system.