AI in schools – The choice we keep making

Every new educational technology is pitched as an efficiency that will scale teaching, but history shows the pattern repeatedly fails: televised lessons in the 1950s, MOOCs a decade ago, and classroom smart boards all promised to replace or vastly improve in-person instruction and instead undermined core classroom practice. Education’s durable truth is that learning depends on a skilled, present teacher working closely with a small group of students; technology has never made that relationship cheaper without degrading it. A concrete smart‑board example: a school spent more than $50,000 on panels that covered most whiteboard space and rarely worked, and teachers reverted to ordinary boards and projectors or even chalk. That history reframes the current push for AI.

A 2008 rollout of open gradebooks shows how well‑intentioned tech can damage learning: live grades led students to check scores 50-80 times a day, incentivized point‑chasing over risk and growth, produced grade inflation, and left colleges facing overconfident students whose external test scores didn’t match transcripts. The crucial split with AI is clear: one version amplifies teacher judgment and saves prep time, the other hands curricular and assessment decisions to algorithms and reduces adults to monitors. The choice is to listen to teachers, preserve time, attention, and relationships, and accept that real education is expensive because it cannot be automated without cost to children.