Artificial Intelligence, Human Cognition, and Intellectual Dependency: Balancing AI-Assisted Productivity with Independent Thinking

The rapid diffusion of generative artificial intelligence (AI) into education, writing,programming, and everyday decision-making has raised a pressing question: ishabitual reliance on AI systems eroding the cognitive capacities — memory,reasoning, creativity, and independent judgment — that these systems were built toaugment? This paper undertakes a multidisciplinary review, drawing on cognitivepsychology, neuroscience, human-computer interaction, educational theory, andIslamic epistemology, to examine the evidence for and against an “AI dependency”hypothesis. We synthesize experimental findings on cognitive offloading, transactivememory, automation bias, and the neurophysiological correlates of AI-assistedwriting, alongside decades of research on deep reading, desirable difficulty, and skillacquisition. We find consistent, though still preliminary, evidence that unreflectiveAI use is associated with reduced memory encoding, weaker neural engagement,and lower confidence in unaided reasoning, while also documenting genuine, well-established benefits in accessibility, personalized learning, and productivity. In parallel, we examine how the Qur’an and authentic Sunnah frame knowledge (’Ilm), reflection (Tafakkur), reasoned inquiry (’Aql), verification (Tabayyun), and balance (Wasatiyyah) — concepts that anticipate, in a normative register, much of what cognitive science now describes empirically as the risk of passive, unverified reliance on external systems. We integrate these strands into a Balanced Human–AI Intelligence Framework (BHAIF), a set of practices for individuals and institutions that preserves AI’s productivity gains while protecting the active cognitive engagement on which durable learning depends. We conclude that the evidence does not support treating AI as inherently harmful to cognition; rather, harm appears concentrated in patterns of passive, unverified, effort-avoidant use, and is mitigable through deliberate practice, education, and — for Muslim users —engagement with an intellectual tradition that has long treated unreflective imitation (Taqlid without evidence) as an epistemic hazard in its own right.