Evolutionary algorithms used for multi-objective optimization mostly prioritize fitness over diversity to achieve a single optimum fast, or a region in the Pareto-front. In this paper, we argue on that diversity should be a primary objective as well, and we propose a novel approach called EGAL to solve a well-known problem: to generate very different exercises to test students' knowledge in a specific range of topics. We show that focusing on diversity and fitness at the same time result in a better quality of solutions in the resulting population.