Cognitive friction is a framework introduced by Luiza Jarovsky in How AI Is Shaping Us (Jarovsky, April 2026) (April 2026, edition #289). It frames AI-induced cognitive deskilling by analogy to physical deskilling from sedentary work, and prescribes deliberately scheduled sessions of unaided cognitive work as a countermeasure.
The physical-friction analogy
Knowledge workers spend many hours sitting in front of screens. Jarovsky argues that human bodies were not designed for this, and that without physical friction — walking, running, cycling, swimming, weight training — hunter-gatherer-optimized bodies degrade.
By her account, generative AI presents a new kind of friction-loss: the first time a technology has automated general cognitive processes en masse. She notes that cheap, widely-available LLM-powered systems can perform many cognitive tasks faster than any human, making the temptation to delegate correspondingly large.
Jarovsky draws a parallel to existing technology-induced impairment: just as many people today cannot perform basic mathematical operations or effectively geolocate themselves without calculators and navigation apps, that type of cognitive impairment could spread to all the tasks that AI performs better than humans can.
The prescription
Jarovsky's prescriptive response is to add cognitive friction to compensate for AI-assisted cognitive work, mirroring the way physical friction compensates for sedentary work. For any domain a person considers part of their core professional expertise, she recommends deliberately scheduling weekly sessions to perform that task without any AI assistance, from brainstorming through final editing.
She gives parallel examples across professions: a writer who uses AI to write should write without AI weekly, a coder who uses Claude Code should code without AI weekly, and a researcher who uses AI for literature review should periodically do literature review unaided. The stated bar is staying competent at the work without AI, even while doing most of the work with AI.
Documented effects cited
Jarovsky frames cognitive friction as a prescriptive response to a body of measured effects, citing several:
- Work intensification rather than reduction. An HBR analysis (February 2026) describes workload creep leading to cognitive fatigue, burnout, and weakened decision-making, in which an initial productivity surge gives way to lower-quality work and turnover.
- Skill-formation degradation, especially among juniors. An arXiv paper Jarovsky cites argues that "the aggressive incorporation of AI into the workplace can have negative impacts on the professional development of workers if they do not remain cognitively engaged. Given time constraints and organizational pressures, junior developers or other professionals may rely on AI to complete tasks as fast as possible at the cost of real skill development."
- LLM fallacy. Jarovsky cites a 2026 paper coining the term, defined as "a cognitive attribution error in which individuals misinterpret LLM-assisted outputs as evidence of their own independent competence, producing a systematic divergence between perceived and actual capability." Jarovsky's accompanying social-media observation is that people who claim they have become much smarter and write better since starting to use AI may simply be observing that the AI is doing the writing.
- Cognitive debt. Referenced via the Kosmyna et al. arXiv preprint (June 2025).
Debates and positions
Jarovsky frames cognitive friction as a personal choice with stakes: each person decides how they want to use AI and what type of cognitive impairment they are comfortable with. She argues that without AI literacy and awareness of how AI affects users cognitively, psychologically, and physically, there can be no sound choices, and most people will simply be swayed by the AI wave. She links this argument to her broader literacy-divide thesis (see AI Divides (Literacy / Occupational / Ethico-Philosophical)).
The framework is positioned against the social-media discourse holding that, when AI does the work, the user becomes smarter — the view Jarovsky targets with the LLM-fallacy observation above.
Relationships
- introduced-by: How AI Is Shaping Us (Jarovsky, April 2026)
- depends-on: Luiza Jarovsky (her broader human-flourishing framing)
- related: LLM Fallacy, AI Divides (Literacy / Occupational / Ethico-Philosophical) (literacy divide), AI Mental Health and Psychological Harm, AI Acceleration Paradox
- contradicts: the "AI does the work and you are smarter" social-media discourse Jarovsky targets