Is AI making us stupid?

Synthesis: A 3-page perspective (opinion/review, not an empirical study) addressing whether AI use erodes human cognition. The authors' answer: not inherently — but the risk is real and follows the Cognitive Psychology principle of cognitive offloading. When people delegate reasoning, writing, memory, or Problem Solving to AI, they forgo the mental practice that builds and maintains those capacities. The threat is use-dependent, not intrinsic to the technology: AI that augments thinking preserves the underlying processes; AI that replaces them outsources exactly the practice that builds expertise.

Core distinction: skills vs. basic cognitive abilities

The article's organizing framework separates two facets of our cognitive systems:

• Skills — learned, domain-specific behaviors supported by knowledge (arithmetic, flying, diagnosis, writing, programming). Acquired and maintained through practice [4]. Offloading practice to AI "will almost certainly compromise skill acquisition" [5] and can cause skill decay [5,7].
• Basic cognitive abilities — foundational, domain-general capacities (working memory, selective attention) that underlie skills. May be more resilient to erosion: cognitive-training research shows gains are highly task-specific rather than broad [9], so basic abilities appear "stubbornly resistant to substantial change." Open question whether long-term or developmental-stage offloading could still shift them [11,12].

Evidence cited

• Math learning (high school) [6]: Students with an AI that let them fully offload solving practice problems scored higher on practice but performed worse on a later no-AI test than students who never had AI. A third condition — a custom AI tutor that probed knowledge and filled gaps (rather than giving answers) — performed no worse than no-AI students, indicating the targeted skill was still acquired.
• Endoscopy skill decay [8]: After an AI detection tool was introduced, adenoma detection rates fell from 28.4% → 22.4% in cases where the AI was unavailable — evidence of deskilling when the tool was withdrawn.
• Knowledge acquisition (Box 1) [2,3]: Offloading to an external store during learning reduces later retrieval; an AI-summary study found advice was "briefer, less unique, and rated less helpful," with learners spending less time and feeling less ownership — shallower depth of learning than web search.

The "how we use AI matters" argument

Whether skills survive offloading depends on the form the offloading takes [13]:

• Completely offloading the task ("student has the AI submit the answer") → harm.
• Having the AI provide an explanation, a suggestion, or act as a collaborator giving feedback [14], or emulating a thoughtful tutor [15] → can preserve or even boost skill despite reduced effort ("Coach not crutch" [15]).
• Staying "in the proverbial cognitive loop" mitigates costs; costs are likely limited to the specific skills offloaded, not basic abilities.

The "coach not crutch" counterpoint

This perspective's answer — that offloading can harm but need not — is empirically sharpened by Lira et al. (2025), who show that AI can exert opposing effects on effort and learning environment, sometimes yielding "work less, learn more." In pre-registered experiments, adults who practiced cover-letter writing with an AI tool wrote better no-AI letters than those who practiced alone — and AI beat both googling examples and personalized feedback from experienced human editors. Notably, they found no illusion-of-mastery effect (AI-practice participants did not over-report skill), directly addressing this perspective's open question about metacognitive source-monitoring. The reconciliation with the "making us stupid" risk is the form of offloading: Lira et al.'s AI scaffolded (surfacing examples and feedback, keeping the learner in the loop) rather than replacing the writing act. Where it did the latter, their guardrails/performance-vs-learning literature still shows the crutch harm — so the two results are consistent, not contradictory: AI that coaches preserves or boosts skill; AI that substitutes risks decay. The boundary condition both camps converge on is whether the learner's cognitive act is preserved.

Concluding remarks & open questions (Box 2)

"It is far too early to say with certainty" the long-term effects. Open questions: prolonged offloading over years/decades; developmental-stage effects (children?); decay as a function of initial skill level and skill type; refresher-training strategies; designing AI to discourage harmful offloading (policy levers); whether people can learn to strategically offload; impacts on metacognition and source-monitoring (misattributing AI output as one's own); effects on dispositions toward thinking.

Notably, the article cites the PNAS Guardrails paper (Bastani et al. 2025, Generative AI without guardrails can harm learning: Evidence from high school mathematics) and the "Coach not crutch" preprint (Lira et al. 2025) — making the offloading/crutch mechanism a shared thread across the knowledge base.

Why this matters for the knowledge base

• The canonical cognitive-offloading citation Cognitive Offloading needed; this perspective synthesizes the mechanism and the skills-vs-basic-abilities resilience distinction.
• Conceptual bookend to the empirical RCTs already in the knowledge base: Generative AI without guardrails can harm learning: Evidence from high school mathematics (unguarded tutor cut exam scores via crutch/offloading), Faster Completion, Less Learning: Generative AI Reduced Study Time on Math Problems and the Knowledge They Build (population "Cognitive Surrender"), and contrasts with Artificial intelligence and feedback in university education: effectiveness and student perceptions (well-architected AI feedback matched teachers — the "coach not crutch" design).
• Reinforces Over-Reliance, Metacognition (source-monitoring), Self-Regulated Learning (deliberate vs. passive offloading), and AI Literacy (using AI to augment, not replace).

What this means for practice

• Instructors. Keep students in the cognitive loop: when AI is permitted on practice, require them to review or explain the AI's solution, or ask for a hint rather than an answer, because the form of the offloading — not its presence — determines whether the skill survives.
• Prefer tutor-style assistance over answer-giving. In the high-school math study the authors cite, a tool that solved the practice problems raised practice scores but lowered scores on a later no-AI test, while a tutor designed to probe knowledge and fill gaps left students no worse than those who never had AI.
• Assess the specific skill with AI withdrawn. Performance during AI-supported practice is an unreliable proxy; the endoscopy evidence the authors cite — adenoma detection falling from 28.4% to 22.4% when the tool was unavailable — shows how quickly a supported skill decays.
• Name the one skill each assignment is meant to build and protect it from offloading, since the costs appear limited to the skills actually offloaded rather than spread across basic cognitive ability.
• Build in metacognitive checks against source-monitoring error, so students can tell whether an insight came from their own reasoning or from the model.

Limitations

• This is a three-page perspective in Trends in Cognitive Sciences, not an empirical study: it reports no new data and reasons from a small set of cited studies.
• Its evidence is indirect for education. The strongest causal support is a single high-school mathematics experiment (Bastani et al. 2025) with three conditions, an adult cover-letter writing preprint (Lira et al. 2025), and a retrospective multicenter observational endoscopy study; none of these followed school learners over time.
• The authors state that it is "far too early to say with certainty" what the long-term effects of offloading will be, and their Box 2 lists the open questions they cannot answer: prolonged offloading over years or decades, effects at different developmental stages, and whether decay varies with initial skill level or skill type.
• The claim that basic cognitive abilities are more resilient rests on cognitive-training research rather than direct evidence about AI, and the authors concede that factors such as schooling are associated with modest changes in those abilities, leaving the developmental case unresolved.

Citation

Cash, T. N., Kelly, M. O., Macnamara, B. N., & Risko, E. F. (2026). Is AI making us stupid. Trends in Cognitive Sciences