AI and Cognitive Development: Why Children Face Greater Risk Than Adults
Summary
- • Adults lose AI-offloaded skills through atrophy; children may never build them at all.
- • Researchers term child AI dependency 'cognitive foreclosure' — potentially irreversible neural gaps.
- • Shen & Tamkin 2026: developers who delegated coding to AI scored 17% worse on understanding tests.
- • Children can't audit AI output in domains they're still supposed to be learning.
Details
Atrophy vs. foreclosure: two categorically different AI-induced cognitive harms
Adults who offload tasks to AI weaken skills they already have — atrophy, like an unused muscle: recoverable if AI access were removed. Children who offload tasks they've never learned may never form the underlying neural pathways. The article argues this 'foreclosure' may be irreversible in ways adult atrophy is not.
Gerlich study: younger users show inverse pattern — more AI reliance, less critical thinking
In Michael Gerlich's study on AI offloading and critical thinking, participants over 46 showed higher critical thinking alongside lower AI reliance. Participants aged 17-25 showed the inverse. The author interprets this as younger users offloading tasks they never learned, not tasks they already know how to do.
Shen & Tamkin 2026: AI-delegating developers scored 17% worse on conceptual understanding
A 2026 preprint by Shen and Tamkin tested professional developers learning a new coding library. Those who fully delegated to AI produced working code but failed conceptual quizzes afterward, scoring 17% worse than non-AI peers. They also couldn't debug the AI's code. These were adults with existing expertise — the effect would be compounded in learners without that baseline.
The AI Audit Problem: children can't catch errors in domains they're still learning
Auditing AI output requires the exact domain expertise the child is still supposed to be developing. Adults can notice oversimplification, catch omissions, and recognize overconfident AI claims. Children lack this capacity precisely because they are still building it — creating a feedback loop where AI errors go uncorrected and shape understanding instead.
Homogenization risk: one AI model routing a generation's reasoning may narrow intellectual diversity
When students route their reasoning and information synthesis through a single AI model, the article speculates that population-level convergence in intellectual patterns may result — narrowing the diversity of approaches across a generation in ways that have no direct adult analog.
Insight = author's conceptual argument; Research = cited empirical study findings
What This Means
This Psychology Today opinion piece argues that AI-in-education discussions are underweighting a critical asymmetry: cognitive risks to developing minds differ in kind, not just degree, from risks to adult users. If the 'foreclosure' thesis holds, the window for intervention is time-bound in a way that adult cognitive atrophy is not. AI practitioners and researchers building tools for educational contexts should consider whether design assumptions governing adult productivity AI are appropriate — or actively harmful — when applied to learners who have not yet formed the cognitive infrastructure those tools assume exists.
Sources
- Adults Lose Skills to AI. Children Never Build ThemPsychologytoday
