Human experts have traditionally been involved in the development of Artificial Intelligence approaches to real-life problem solving. The inclusion of expert knowledge can clearly show the difficulties in adequately finding solutions in huge search spaces, while the opposite usually leads to over-simplification and poor solutions. This paper shows how a proper combination of human learning, as developed by inexperienced students in a given domain, and teaching, as applied by the experts, may allow to find new ways for finding solutions of quality. A real-life problem is used to conduct the research: the 4-part harmonization problem that music students face. Results show how we may profit from the way humans teach and learn, thus allowing to consider teaching within the evolutionary problem-solving approaches, thus paving the way towards evolutionary machine teaching.