AI Is Making Students Smarter (and Dumber) | Should AI be Used in Education? Future IQ

207 views • Jul 17, 2026

Concepts in this episode

Browse all concepts ›
  1. Standard Deviation mental-model

    Standard deviation describes the typical distance between observations and their mean, turning raw differences into differences relative to a population’s spread.

  2. Bloom's 2 Sigma Problem principle

    One-on-one tutoring can substantially outperform conventional classroom instruction, but its cost has historically prevented it from scaling. AI may loosen that constraint by making personalized tutoring cheaper to provide.

  3. Evaluate a technology against the realistic alternative it replaces, not against an unattainable ideal.

  4. Performance-Learning Gap mechanism

    Tools can improve measured performance while weakening genuine learning when they supply answers and remove the productive struggle that builds understanding.

  5. Illusion of Explanatory Depth mental-model

    Familiarity creates a false sense of understanding because the mind stores a usable outline while silently skipping the connections between parts. Trying to explain, reproduce, or operate the thing forces those missing connections into view.

  6. Retrieval Practice principle

    Durable learning requires actively recalling and applying an idea, not merely recognizing or understanding an explanation when it is presented.

  7. Motivation Bottleneck mechanism

    When high-quality instruction is readily available, access may stop being the main constraint on learning. The bottleneck shifts to whether learners choose to begin and return consistently enough for the tool to matter.

  8. Making excellent educational resources abundant does not guarantee learning because desire, discipline, and sustained engagement remain scarce bottlenecks.

  9. Active Recall mechanism

    Learning becomes durable when you pull knowledge from memory, restate it in your own words, apply it, and return to it repeatedly—not when you merely reread an explanation.

  10. Learning by Doing principle

    Making or testing an idea converts assumed understanding into observable performance. The attempt exposes missing details, forcing abstract confidence to answer to concrete feedback.

Description

When it comes to using AI for education, there's a lot of polarisation in public opinion. Some believe it should be completely banned, while some believe it should be embraced sooner than later. In today's episode, we deep dive into this topic to understand what are the challenges that come with the rise of AI especially in how children and students interact with it for learning. Based on multiple research papers, we explain that neither of the polarised camps are the right answer -- but finding a middle ground is what is needed. Remembering that AI is a tool, and not a companion is how one will find this middle ground. If used with an attitude of curiosity, with a focus on understanding the process rather than just getting the solution, using AI to actually solve more complex problems rather than asking it to do your homework, it has the power to influence education in a positive way. 💬 Join Our WhatsApp Community: http://tapthe.link/futureiqwa Find us on Substack: The FutureIQ Substack: https://futureiq.substack.com/ The AI IQ Substack: https://aiiq.substack.com/ Find Navin on Substack: https://substack.com/@ngkabra Do hit us up on Twitter: @ngkabra http://twitter.com/ngkabra @shrikant https://twitter.com/shrikant Videos you may like / referenced in today’s episode: The Illusion of Overconfidence: https://youtu.be/_ak0k7GNCjM Spaced Repitition: https://youtu.be/JAPwrsm5OeA Can you Bribe ChatGPT?: https://youtu.be/txFM43N8ePg ChatGPT can make you Smart or Dumb: https://youtu.be/McdWSLWQUkA Will AI take away my job?: https://youtu.be/3fOTvF8ReXA The Value of Branding: https://youtu.be/Bg5MJ9jOvO8 Why Idiots are Winning: https://youtu.be/CfpeFsxcB5g Books referenced in this video: Atomic Habits: Listen it on the podcast provider of your choice: https://tapthe.link/FutureIQRSS Follow FutureIQ on Instagram: https://www.instagram.com/thefutureiq/ Source / References: - Bloom’s 2-sigma problem: https://en.wikipedia.org/wiki/Bloom%27s_2_sigma_problem - AI in EdTech: https://futureiq.substack.com/p/ai-in-edtech-should-help-teachers - AI can help students or hinder them: https://www.oneusefulthing.org/p/choosing-to-stay-human - EdTech/AI products work when used as intended: -https://www.ixl.com/materials/us/research/The_Impact_of_IXL_in_New_York_State.pdf - https://f14f78b2-f764-4a7b-8742-57cf62de1b57.filesusr.com/ugd/2c805c_32541c8af0a743ceb4ed37a5696fe7b8.pdf - Khanmigo: [internal link removed] - But only 5% use them as intended: https://www.educationnext.org/5-percent-problem-online-mathematics-programs-may-benefit-most-kids-who-need-it-least/ - AI makes students dumber: - https://scale.stanford.edu/ai/repository/generative-ai-can-harm-learning - https://www.ie.edu/center-for-health-and-well-being/blog/ais-cognitive-implications-the-decline-of-our-thinking-skills/ - Tutor prompt: https://aiiq.substack.com/p/using-agents-without-becoming-dumb - Use the best paid models: https://aiiq.substack.com/p/use-the-best-paid-models-you-can - Dealing with AI sycophancy: https://aiiq.substack.com/p/llms-try-to-please-you-and-thats Chapters: 00:00 Intro 00:55 Standard Deviation & the 2 Sigma problem 04:10 But is AI really GOOD ENOUGH? 05:40 The small problem 07:22 The medium problem 09:31 The big problem 11:45 How to use AI efficiently? 14:36 Should I use AI for everything in life now? 16:33 Do's & Don'ts

Transcript

Subscriber transcript

Subscribe to @TheFutureIQ, then sign in with Google to unlock full transcripts and transcript search.