A preview from MasteryLoopHQ, an AI tutor built around explaining it back.
Kids and teens can now get fast, polished answers from AI. A parent looks at the finished work, and it looks fine. Addy Osmani wrote about this problem for engineers and product people, in an X Article titled “Critical thinking when agents do the work.” The same pattern shows up at the kitchen table. He points out that thinking used to happen while you waited for an answer, and that wait has almost disappeared: “In the past, most of our cognitive and aesthetic processing happened in the space between question and answer. That space is now measured in seconds.” He also warns: “Many (most?) outputs will be well-written and mostly correct, but they won't demand a second look.”
Osmani describes three levels of leaning on AI. In his words: “At the top: cognitive offloading, the act of delegating tasks to the AI, but maintaining ownership of the result. I do this all day. You should, too. In the middle: deferral, where the output of the tool becomes your new input; the result has ceased to be a hypothesis. At the bottom: cognitive surrender - the AI's output quietly becomes your own, you feel no need to check it.” Here is what each level can look like for a student: Offloading: the child uses AI as a helper and can still teach the idea back. Deferral: the AI's draft becomes the starting point, and the child's own first guess disappears. Surrender: the child pastes the answer in and moves on.
It would be easy to turn this into a scare story. Osmani doesn't. He writes: “I'm a big fan of these tools. I'm not writing this to advocate for less agentic workflows.” He explains his aim: “My goal is to help you maintain critical thinking in a world where it's easy to offload anything to an agent.” And he draws a line by stakes: “Where the stakes are low, feel free to delegate entirely.”
Osmani gives a simple test. He writes: “When you can't explain the answer to a question, even by way of your agent, you've ceded ownership.” And: “If you can't explain the answer, even with an agent's help, you've surrendered it.” He adds: “Be honest about your place on the spectrum.” At home, the check looks like this. Close the AI tab. Ask your child to explain the answer in their own words. If they need the AI to explain “their” answer, the answer isn't really theirs yet. MasteryLoopHQ, the site this study is published on, is built around the same idea: explaining it back.
Osmani asks himself a question before he uses AI: “Did I formulate a hypothesis before asking? Before I run an agent on any non-trivial task, I write down, even just in my head, what I think the answer might be. It's OK if it's a bad guess.” He explains why the guess matters: “Once you've read a well-written response, though, it's hard to imagine any other possibility.” For a student, that means writing down or saying a rough answer, or the next step, before opening AI. Then they compare their guess with what the AI says.
Osmani warns that people can hand over the checking, too: “You can also surrender verification. ‘Is there evidence that supports my conclusion?’ becomes ‘Did the check turn green?’” His advice: “When a check goes green, ask what it proved. Then ask what it didn't.” For schoolwork, a correct multiple-choice click or a clean grammar check shows one thing went right. It doesn't show the student understands. A good follow-up question: what would a wrong answer have looked like, and why?
Osmani credits the term to two researchers: “Cognitive surrender is a term coined by Steven Shaw and Gideon Nave at Wharton.” Shaw and Nave have a working paper titled “Thinking—Fast, Slow, and Artificial: How AI is Reshaping Human Reasoning and the Rise of Cognitive Surrender.” We did not open the full paper for this study, so we don't report its numbers or findings here. If you see percentages about cognitive surrender shared online, check where they came from before you rely on them.
Osmani's habits fit in a few lines. “Have a hypothesis. Formulate one before asking, even if it's bad.” And: “Before you ask, write down what you think will happen, and after the agent is done, write down one thing it didn't check.” Here is a short checklist for home, built from his ideas. First, decide whether the task is low-stakes or high-stakes. Second, have your child make a guess before asking AI. Third, close the AI and have them explain it back. Fourth, ask what the “green check” didn't prove. For more on using AI well while learning, see How to Ask AI Good Questions While You Learn and The Learning Mode Prompt. Our “Why This Works” page explains why explaining it back helps learning.
Cognitive surrender is when an AI's answer quietly becomes your own and you never check it. Addy Osmani puts it at the bottom of his three-level spectrum. The top level, offloading, is useful. Homeschool parents can use his test: if a child can't explain the answer with the AI closed, the answer isn't really theirs yet. Use AI for low-stakes help, and protect understanding with a guess before asking and an explain-it-back afterward.
In your own words: (1) What's the difference between offloading and surrender on Osmani's spectrum? (2) Your teen used AI on a history paragraph. What would you ask them to do next, with the tab closed? (3) Why does Osmani want a guess before asking the AI? Answer key. (1) Offloading means handing a task to the AI but keeping ownership of the result. Surrender means the AI's output quietly becomes your own and you feel no need to check it. (2) Ask them to explain the paragraph's main claim and its evidence in their own words. If they need the AI to explain it, the answer isn't theirs yet. (3) Because once you've read a polished answer, Osmani writes, “it's hard to imagine any other possibility.”
Reading a good explanation feels like understanding it. Usually it isn't the same thing — and you don't find out which one you've got until someone asks you to explain it back.