Originally written in Traditional Chinese: 要看的不是平均，是沒跟上的是誰
Two hours of class a day, twice the learning. That is roughly what one American AI school promises. This week, The New York Times checked it against the school's own data. Among students who started behind in math, only 37 percent made it.
The school is Alpha School, a private network in the U.S. with about 50 campuses. Academic classes run roughly two hours a day. Most of the teaching is done by AI software, while adults guide and motivate. The pitch: learn twice as much in half the time of a traditional school.
On October 23, The New York Times reported on the school's own data. Only 42 percent of its private-school K-8 students progressed twice as fast in both math and reading over the last school year. About half did not reach the doubled pace in math. And the number from the top: about a fifth of Alpha's private-school students scored at or below the 50th percentile nationally in math last fall, and only 37 percent of that group met the promised 2x growth by winter.
Let me be fair first. Forty two percent of students doubling their growth in two subjects would be impressive, if the numbers hold. The problem is not the 42 percent. The problem is who the promise is about. It says "students." Parents hear "my child."
My take: at least in this data, AI most easily accelerates the children who are already ready. The ones who need help the most are the hardest to speed up. What matters is not the average. It is who got left behind.
To be clear, I am not saying AI learning does not work; that 42 percent suggests it may work for some students. I am not saying the school is lying; these are its own figures, compared by the Times against its own promise, and supporters argue AI-personalized instruction helps improve efficiency and lets students learn at their own pace. Critics raise a different question: high test scores do not necessarily mean a child truly understands. Every parent should take that question home.
You will hear more promises like this. "Finish a day of school in two hours." "Double your learning." So here is something to try today. I call it the three promise questions. The next time an AI tool, course, or program claims it will help your child learn faster, ask:
- Who measured it? An outside test, or the product's own system?
- What did they measure? Progress through the material, or real understanding?
- Who did not make it? Which children are hidden under the average?
The third question matters most. In my course I teach a three-layer verification habit, because the most dangerous thing about AI is being confidently wrong. I ask learners to put one question to AI: "Under what conditions would this answer not hold?" That question works on AI. It also works on everyone selling AI. Including me.
For parents: tonight, ask your child to explain, in their own words, one thing they learned today. If they can explain it, they learned it. If they cannot, they only finished it.
For educators: when your class uses an AI platform, skip the class average at first. Look at the bottom five.
For professionals: when your company says AI doubled productivity, ask whose productivity doubled, and whether anyone is coaching the people whose productivity did not.
Back to the promise. Speed was never the problem. But the real promise of education is not a faster average. It is that no child gets averaged away. The stronger AI gets, the more valuable the adult who will sit with a struggling child for five more minutes.
Get the Human-First practice card
Four steps, ten minutes, with copy-ready AI prompts. The card and my weekly letter are currently written in Traditional Chinese.
👉 Subscribe free to 碩學苑 (Traditional Chinese)
Follow in English
I also post in English on LinkedIn.
👉 Follow Haishuo Lee on LinkedIn