Problem solvers who routinely distribute cognitive processing between their minds and the external environment often overestimate their ability to perform tasks without outside help. This bias—commonly referred to as knowledge illusion, illusion of competence, or foresight bias—has been demonstrated in multiple contexts such as anagram solving, where participants who had received the solution’s first letters (e.g., “C O O _ _ _” for “OEOCKI”, solution: COOKIE) for a while led them to overestimate their future performance without outside help (e.g., “_ _ _ _ _ _” for “F E O E F C”, solution: COFFEE). Metacognitive miscalibrations are problematic, as they can shape decision-making and foster unwarranted reliance on one’s own abilities when seeking support would be advantageous or even critical. In this talk, we will review established and introduce novel interventions to recalibrate or prevent the development of knowledge illusions. Specifically, we will present novel work regarding an intervention involving five questions designed to reflect on the contribution of outside help (e.g., “How confident are you that you could have solved the anagrams without any hints?”). In addition, we will describe an intervention in which the outside help was provided by agents with either software-like or robot-like appearances, which was aimed at enhancing the distinction between internal and external contributions to problem solving and thereby mitigating the emergence of knowledge illusions.