While training on several examples is known to be a decisive step in learning how to flexibly use mathematical concepts in different situations, selecting the optimal sequence of examples remains a significant challenge for mathematics education (De Bock, Deprez, Van Dooren, Roelens, & Verschaffel, 2011). Specifically, identifying which examples may help learners develop their cognitive flexibility - the ability to consider the same problem from different perspectives and find its solution regardless of its wording - is a crucial stake for educational psychology (Clément, 2022). This study investigates whether the cognitive load attached to a problem may either facilitate or hinder one’s ability to find an optimal path of resolution. We analyze the interactions between cognitive flexibility and cognitive load using arithmetic word problems whose solution requires a perspective change (Gros, Thibaut, & Sander, 2021). In two online experiments, we asked adults to solve such problems while varying the extraneous cognitive load of the task, either by means of a dual-task paradigm, or by changing the difficulty of the calculations leading to the solution. We hypothesized that reducing a problem’s cognitive load may be a powerful lever to help learners find a problem’s optimal solution. We analyzed the rate of use of the different solving strategies to assess the extent to which the cognitive load imposed on a problem influences the expression of cognitive flexibility among adults. By providing a finer understanding of the interactions between cognitive flexibility and cognitive load in arithmetic reasoning, we intend to shed a new light on the crucial learning processes involved in problem solving, while providing guidelines regarding which examples to use to foster cognitive flexibility in mathematics classrooms.