Abstract. While various recommender approaches are increasingly considered in e-learning, lack of studies of actual use is hindering the development. For several years, we have used non-algorithmic recommender features on an undergraduate course website to help students find pertinent study materials. As students earn credit from adding and evaluating materials, some have chosen to evaluate materials dishonesty, i.e. without actually reading them. To improve honesty, in 2012 we coupled 5-star ratings with commenting (previously uncoupled) to increase the cost and complexity of evaluating and gave students individual presence with nicknames (previously anonymous) to increase social presence and enable reputation formation. Our results show that high enough cost of evaluating together with high enough social presence can lead to complete honesty in evaluations and enhance both user experience and student involvement. In effect, designing such e-learning systems includes not only designing the features but also their use, as the two are intertwined.