<![CDATA[Generative artificial intelligence is rapidly becoming a cognitive partner in education, capable of planning tasks, monitoring progress, and evaluating solutions on a learner’s behalf. This conceptual synthesis paper examines the risk that such AI tools, while improving immediate performance, may erode students’ metacognitive abilities in their capacity to plan, monitor, and evaluate their own thinking. Drawing a parallel with GPS navigation research, where habitual turn-by- turn guidance has been shown to impair spatial memory and hippocampal engagement, we introduce the metaphor of AI as a “GPS for thinking". Through an integrative review of literature spanning cognitive psychology, neuroscience, and the learning sciences, we synthesize evidence that AI-assisted learning can lead to a form of cognitive disuse atrophy, specifically by short-circuiting the metacognitive loop. Emerging studies reveal that students who rely heavily on AI tutors often perform worse when the tool is removed, suffer from an illusion of explanatory depth, and struggle to articulate the reasoning behind their answers. To counter these effects, we propose a shift from a GPS model where the tool issues commands to a compass model, where the tool provides orientation while preserving learner agency. Five evidence-informed design principles are advanced: prompting planning before assistance, delaying and fading feedback, embedding mandatory reflection pauses, making AI reasoning visible, and calibrating learners’ confidence. The article argues that the long-term goal of educational AI must be to strengthen, not supplant, the student’s inner compass.]]>
Copyrights © 2026
