Preparing the model…

Interactive guide / Semantic explanation

Can AI help my child learn?

See how answers, hints, tutor support, immediate performance, and later unaided learning are different outcomes.

What this experience shows

  1. Start with the goal

    A right answer is not always learning.

    Immediate performance and later unaided mastery are different outcomes.

    Physical focus: School

  2. Change the help

    Answer, hint, and coaching are different tools.

    The design determines whether AI replaces student thinking or supports it.

    Physical focus: AI model

  3. Keep the human

    AI can make a tutor more capable.

    Tutor CoPilot supported human tutors and reported a percentage-point mastery gain—not a replacement of the tutor.

    Physical focus: A person

Evidence connected to this pathway

  1. randomized trial / Study context

    Students can perform better while AI is available yet learn less for later unaided work when assistance gives away answers; guardrails and tutor-like design can change the outcome.

    Students can perform better while AI is available yet learn less for later unaided work when assistance gives away answers; guardrails and tutor-like design can change the outcome.

    Proceedings of the National Academy of Sciences: Generative AI can harm learning when it gives away answersOECD: OECD Digital Education Outlook 2026
  2. observational finding / United States

    Students whose human tutors had AI assistance were about four percentage points more likely to master a topic; gains reached about nine percentage points with lower-rated tutors.

    Students whose human tutors had AI assistance were about four percentage points more likely to master a topic; gains reached about nine percentage points with lower-rated tutors.

    Stanford National Student Support Accelerator: Tutor CoPilot: a human-AI approach for scaling real-time expertiseStanford SCALE: Tutor CoPilot publication

Continue through the evidence graph