Following up on this morning's EU Council post, here is the cognitive science version of the same argument.

Four 2026 studies say AI use is not the variable. How you use it is. Active engagement preserves cognition. Passive offloading accumulates "cognitive debt" (the term Kosmyna's MIT team coined). The Stanford research found ONE intervention reliably preserved evaluative thinking during AI use: a five-minute self-assessment rubric BEFORE the AI prompt.

Honest question for the room: think about the last time you used AI for something that mattered. Did the thinking happen before the prompt, or did the prompt do the thinking?

There is no wrong answer here. The research suggests most of us are oscillating between both, and the goal is to notice which mode we are in.

Full breakdown of the research:

https://www.harvestkernel.com/cognitive-debt-research-active-vs-passive-ai-use/