A new paper about “Optimal learning paths in information networks” just appeared on SciRep. A novel approach to algorithmic education is here proposed, which combines well-established findings of research in cognition and memory with a complex system perspective. Indeed, the bits of information or knowledge the learner wants to learn are treated as nodes in a complex networks of associations, whose role is key in affecting the learning process. We propose a quantification of their effects and design a class of algorithms to efficiently sequence the introductions of the units and their reviews in a learning schedule, thus simulating the learner’s exploration of the knowledge space. How much the topological properties of the network underlying the items can affect the efficiency of the learning dynamics is investigated. Tests are reported on both synthetic and real graphs, like some free word associations graphs and subsections of the online encyclopedia Wikipedia. Our results indicate that structures where hubs and specific nodes are balanced lead to efficient learning schedules. Moreover, the topologies of the real graphs considered are close to be optimal with respect to slight perturbations, thus suggesting an interesting link between the ways humans organize knowledge in complex structures and the best ways to explore the same structures while learning.