Cognitive Tribology: Friction, Wear and the Long-Term Effects of AI Assistance on Human Cognitive Capability

Version 1.2—June 2026 This version incorporates additions published after the original submission date. Section 5.1 adds clinical evidence from Budzyń et al. (2025) documenting a 6.0 percentage point decline in adenoma detection rates among experienced endoscopists performing unassisted colonoscopies following the introduction of AI assistance, described by the authors as the first real-world evidence of automation-induced deskilling linked to patient outcomes, and the most direct clinical confirmation yet of the reversibility standard's failure mode. Also, a Section 5.1 footnote adds Winthrop (2026), who independently coins the term "cognitive stunting" for the foreclosure phenomenon in children and calls for the development of metrics to track children's healthy cognitive development in relation to AI use, extending the atrophy/foreclosure distinction into policy terms. Section 5.1 includes a perspective from Ke et al. (2026) introducing never-skilling, trainees failing to develop foundational clinical reasoning through early AI reliance, and mis-skilling, internalizing AI errors as fact, as constructs that map directly onto cognitive tribology's foreclosure argument and phantom friction construct respectively. Section 5.2 adds Lodge and Loble (2026), who distinguish beneficial from detrimental cognitive offloading in education and find that the risk of foreclosure—rather than mere atrophy—is most acute for novice learners building foundational knowledge, providing policy-level confirmation of the atrophy/foreclosure distinction and the expertise gradient. Section 5.2 also adds Sakai (2026), whose fMRI evidence shows that reading on paper produces more efficient linguistic and narrative-structural integration than reading on digital devices, with digital readers requiring greater compensatory frontal activation to achieve comparable comprehension, a neurological parallel to the desirable difficulties effect. Section 5.3 adds findings from Hackenburg et al. (2026) demonstrating across four preregistered experiments with 6,923 participants that frontier AI systems reliably out-persuade expert humans, including world championship debaters, with the advantage stemming from information volume rather than rhetorical insight. The finding extends the social phantom friction argument from affirmation to active persuasion, establishing epistemic friction as a cognitive safety requirement rather than merely a design preference. Section 5.6 adds Mitchell (2026), who characterizes "jagged intelligence" as the defining architectural feature of current AI systems—an uneven capability landscape without reliable internal signals of failure—establishing that epistemic friction is structurally necessary rather than merely useful. Section 5.7 adds Moon et al. (2025), who introduce a "diversity growth rate" metric across an analysis of 2,200 admissions essays and demonstrate that AI-generated essays contribute fewer novel ideas than human-written ones, with the gap widening as more essays are included, providing empirical confirmation of collective serendipity loss through homogenization. Section 5.7 also adds Vatansever (2026) proposing mental navigation, the brain's spontaneous traversal of conceptual space through the default mode network, as the default computational mode underlying creativity and discovery, providing the neural mechanism through which frictionless AI suppresses the conditions for serendipitous connection. No substantive changes to the argument, framework, figures or conclusions. Version 1.1—May 2026 This version incorporates additions published after the original submission date. Section 5.1 adds neuroplasticity evidence from Rossi et al. (2026) establishing that passive AI reliance weakens activity-dependent brain plasticity while active co-creation sustains it, providing a neurobiological substrate for the atrophy mechanism. Section 5.5 adds empirical evidence from Chakrabarty and Dhillon (2026) documenting that expert writers experienced an identity crisis and eroded aesthetic confidence when confronted with convincing AI-generated writing, confirming the cognitive surface texture erosion argument in the creative domain. Section 5.6 adds findings from DosSantos et al. (2026) demonstrating that LLMs adhere strictly to policies even when counterproductive, and that aligning AI with human judgment requires training on how decisions are made rather than just which decisions are made, providing the mechanical basis for the artificial certainty construct and reframing epistemic friction as an alignment requirement. Section 5.7 adds Uzzi (2026) arguing that whether AI assists or substitutes for human creativity depends on whether it is asked how to think rather than what to think, mapping directly onto the productive friction construct. Section 5.7 also adds insight research from Becker et al. (2025) demonstrating that insight almost doubles memory for solutions and requires conditions of productive struggle that frictionless AI removes, together with a companion review by Becker and Cabeza (2025) providing the mechanistic account: insight functions as a prediction error signal triggering hippocampal and dopaminergic memory consolidation, a cascade frictionless AI prevents by delivering answers before the brain has formed a prediction. Section 5.10 acknowledges the Cognitive Atrophy Paradox framework (Kabashkin, 2025) as a parallel independent contribution addressing the same problem space, while distinguishing cognitive tribology's theoretical and design contributions. This version also includes minor formatting standardizations to in-text citations and typographical corrections in the reference list.No substantive changes to the argument, framework, figures or conclusions. Abstract The drive toward frictionless human-AI interaction has produced interfaces of unprecedented smoothness and capability. We argue this achievement carries an unexamined cost: the systematic erosion of human cognitive capability through the removal of the productive resistance that maintains it. We introduce cognitive tribology—a framework drawing on the science of friction, wear, and lubrication between interacting surfaces—to name, explain, and propose design responses to this problem. The framework makes three core contributions. First, it develops a tribological model of human-AI interaction in which cognitive surfaces are maintained by friction and degraded by its absence, extending Cooper's foundational cognitive friction concept by distinguishing between friction types. Second, it introduces a taxonomy of cognitive friction types—productive, parasitic, and phantom—with particular attention to phantom friction: the substitution of the subjective experience of cognitive engagement for its substance. Third, it proposes seven design principles for friction-conscious human-AI interaction, evaluated against a reversibility standard that asks not how well humans perform with AI assistance but whether they retain the capacity to perform without it. The framework identifies three consequences of over-lubrication at distinct time horizons: phantom friction within individual interactions, cognitive atrophy across months and years, and serendipity loss across careers and fields. The framework is grounded in converging empirical evidence: randomized experimental demonstrations that AI assistance impairs skill formation and reduces persistence on unassisted tasks; neurological evidence of reduced brain connectivity accumulating across months of LLM use; large-scale documentation of metacognitive laziness and cognitive surrender; organizational research on artificial certainty; and formal economic modelling establishing that rational AI adoption can produce a steady state in which workers end up less capable than before. Design implications are developed for AI product design, professional education, organizational policy, and regulatory frameworks. Cognitive tribology does not argue against AI assistance. It argues that AI assistance without friction design progressively diminishes the human partner and that the design philosophy and evaluative frameworks currently governing human-AI interaction are not equipped to detect or prevent this outcome.