Recommender systems in education aim to help students make good decisions about the direction of their learning. The design of such systems in conventional research has treated the decision making process of students as a black box and assumes the best recommendations to be those that accurately predict student choices. Such an approach overlooks potentially valuable use cases for supporting optimal decision making, especially in self-directed learning contexts which present such challenges as identifying all available options, accurately evaluating the options against selection criteria, and selecting the best choice. This qualitative study aims to understand the areas where students struggle in the context of planning an open-ended project in order to inform the design of educational recommender systems. Data from interviews with 7 students at an international engineering school in Japan are analyzed to examine choice behaviors, influences on choice, and difficulty to choose in a self-directed learning context. The results illustrate considerations for designing educational recommender systems that can support the divergent thinking and convergent thinking demands of decision making. We provide case-based examples where the use of different recommender metrics, such as novelty and diversity, may provide value to users with different approaches to the decisionmaking process.