It's Not Who Gets AI, It's Who Guides You

Originally written in Traditional Chinese: AI 時代拉開差距的，不是誰拿到 AI，是誰身邊有人帶

My take: what will separate learners in the age of AI is not who has access to AI. It is who has someone beside them.

In October, EdWorkingPapers published a randomized trial from the University of Maryland. In fall 2025, 2,379 undergraduates and 30 instructors across several disciplines took part. Instructors were randomly assigned, and all students of the selected instructors were offered an AI study assistant built into their course platform, drawing on materials chosen by their instructors. Within sections of the same course, students offered access earned final grades 0.37 standard deviations lower, about 4 percentage points. Participation on the learning platform dropped by 0.90 standard deviations. The authors estimate the losses were greater for first-generation college students.

Days later, Gallup released a survey of 15,482 US workers. College graduates were more than twice as likely as non-graduates to use AI daily or weekly: 40% versus 17%. Among AI users, 68% of those with a bachelor's degree said it helped them work faster, compared with 51% of those with a high school education or less. Gallup's conclusion: these benefits are not evenly distributed.

Put the two together and I see one pattern. AI does not level the starting line by itself. My view: without someone to guide them, AI may make it even harder for people to move up.

I was one of the people who got guided. When I was in middle school in Taiwan, a teacher spent an entire day taking four of us to Taichung. He drove. He paid. It was a day trip, just for fun. The teachers who shaped me all had one thing in common: they did not care only about my grades.

Back to the trial. How did students use the assistant? Reporting on the study says about 74% of requests asked for information, explanations, or solutions, about 11% were practice or test prep, and fewer than 1% asked for feedback on the student's own attempt. AI kept answering. Very few students handed in their own thinking first.

It's not who gets AI. It's who has someone asking, "What do you think?"

To be clear, the lead author, Jing Liu, says the findings do not establish that purpose-built AI tutoring tools are not beneficial. Most instructors kept the default direct-instruction mode instead of the guided tutoring mode. Only about 15% of students offered the tool used it, and this is a working paper not yet peer reviewed. Nor does it mean first-generation students are less capable. It means the students who have the fewest people asking them what they think need that question the most.

One method anyone can use today, which I call the AI Feynman:

  1. Explain something you just learned to AI in three sentences, in your own words.
  2. Ask AI to play a curious ten-year-old: "Don't correct me and don't give me answers. Ask me only one 'why?' or 'can you give an example?' at a time."
  3. Circle the places where you get stuck. Those are the places to study.

For parents: after your child asks AI something, ask one more question: "What do you think?"

For educators: before adopting an AI assistant, ask whether its default is giving answers or asking questions, and who is watching the amount of practice.

For professionals: Gallup found 54% of US workers have not used AI in their role. Teach one colleague your best AI habit this week.

AI is a fast car. What a learner often needs is not more speed. It is someone in the passenger seat.

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

References