Cognitive Debt in AI-Augmented Research: Evidence from Neuroscience and Implications for Knowledge Production

Artificial intelligence systems are rapidly becoming embedded in research and knowledge work, promising efficiency gains but potentially accumulating hidden cognitive costs. This paper proposes the concept of cognitive debt—the delayed cost to attention, learning, and mental health from chronic reliance on AI that reduces active cognitive engagement. Drawing on cognitive neuroscience, human factors research, and occupational mental health literature, this theoretical analysis synthesizes emerging evidence suggesting how AI-mediated work may erode sustained attention, displace effortful learning, and contribute to “silent burnout,” particularly among neurodivergent professionals operating near sensory and attentional limits. A threshold transition model formalizes the proposal that three mechanisms—attentional erosion, effort displacement, and affective depletion—operate additively until exceeding individual capacity thresholds, at which point acceleration toward burnout may become non-linear. This framework is situated within predictive processing accounts of cognitive control, where passive AI offloading is conceptualized as maladaptive precision-weighting over internal versus external models. Three patterns of AI use are distinguished—passive offloading, guided co-construction, and reflective scaffolding—that are predicted to differentially contribute to or mitigate cognitive debt. The paper concludes by reframing design principles as testable hypotheses, proposing a minimal measurement battery, and outlining a multi-site Cognitive Debt Observatory to build systematic evidence. Implications for theories of attention, methodological practices in cognitive science, and research integrity frameworks are discussed. Without deliberate intervention, unreflective AI reliance may risk systematic intellectual atrophy in research ecosystems.