AI Doesn’t Need to Replace Humans to Weaken Human Capability
August 21, 2026 - 4:05 pm
AI Learning and Artificial Intelligence
Summary
AI boosts productivity (25% less email time, 26% more completed tasks), but research from MIT shows LLM-assisted writers had lower brain engagement and worse performance when the tool was removed. Dr. Vishal Kapoor introduces the concept of "cognitive erosion" and proposes an "AI-fed, human-led" approach: think before prompting, don't outsource initial questions or final decisions. He advocates for rationing AI adoption by task type, especially in education and original research.
Artificial Intelligence and Knowledge Work
Artificial intelligence is streamlining knowledge work, removing friction from tasks. A 2025 field experiment found that access to generative AI reduced email time by 25% for some workers. Another study involving software developers showed AI assistance increased completed tasks by 26.08%.
Measuring Productivity vs Cognitive Capacity
Productivity is the easiest metric in AI discussions, visible and attractive to businesses. However, cognitive capacity is harder to quantify. Researchers at MIT's Media Lab discovered that while LLM users produced faster essays, they showed lower brain engagement and performed worse without the tool.
Cognitive Debt and Offloading
The MIT study revealed a pattern of "cognitive debt," a deficit that accumulates slowly and becomes due when assistance disappears. Cognitive offloading, delegating mental tasks to external tools, isn't new, but generative AI distinguishes itself by reasoning on a person's behalf, delivering conclusions instead of raw material.
Dr. Vishal Kapoor's Concerns
Dr. Vishal Kapoor, a researcher at major banks, raises concerns about "cognitive erosion" from excessive reliance on AI. He advocates for considering the condition of the human mind when using AI, promoting an "AI-fed, human-led" approach where humans maintain control over reasoning while utilizing AI to accelerate or strengthen work.
Personal Experience and Observation
Kapoor personally observed a decline in his cognitive profile on a brain training tool after using AI for reasoning, measuring a drop of roughly 40 to 50 points on a 1,000-point scale. This experience underscores his concern about cognitive erosion from over-dependence on AI.