AI & Cognition Glossary
Short, sourced definitions of the terms used to describe what heavy AI use does to thinking - written for developers, not for a psychology exam. Each entry links to the research behind it. For the full evidence base, see every study on AI and cognitive decline.
AI Brain Rot
AI brain rot is the colloquial label for cognitive decline attributed to heavy AI use - weaker memory, thinner attention, skills that fade as the AI takes over. The science underneath it is real (cognitive offloading, skill atrophy), but researchers avoid the phrase itself: rot implies passive decay, and this is neither passive nor irreversible.
AI Skill Atrophy
AI skill atrophy is the decay of professional skills - debugging, reading code, recalling APIs, holding a mental model of a system - that happens when an AI assistant performs those tasks for you. The skills fade through disuse, usually without the person noticing until the AI fails.
Automation Complacency
Automation complacency is the well-documented human tendency to under-monitor an automated system that is usually right - attention drifts, checks get shallower, and errors slip through precisely because the automation rarely fails. The aviation industry has studied it for decades; AI coding tools inherit it wholesale.
Cognitive Debt
Cognitive debt is the accumulated cost of letting an AI think for you: each offloaded task feels free in the moment, but the skipped learning compounds, and the bill arrives later as weaker memory, shallower understanding, and skills you no longer have when the AI is wrong or absent.
Cognitive Offloading
Cognitive offloading is the use of external tools or actions to reduce the mental effort of a task - writing things down instead of remembering them, or asking an AI instead of thinking. The work gets done, but the brain skips the processing that would have built memory and skill.
Metacognitive Laziness
Metacognitive laziness is the tendency to stop monitoring and evaluating your own thinking when an AI is available - accepting its output without checking it against your own judgment. You stop asking "do I understand this?" because the answer no longer seems to matter.