# Will AI significantly improve education or is it exaggerated?: 21 Sourced Claims, Contested Facts & Evidence The strongest case that Will AI significantly improve education or is it exaggerated? will transform society. Authors: Noor HTML: https://moyomap.com/maps/will-ai-significantly-improve Markdown: https://moyomap.com/maps/will-ai-significantly-improve.md License: HTML and markdown are for reading and citation. Bulk copying and AI training need a written license. See /terms and /developers#maps-data. Cite a claim as: MoyoMap, "Will AI significantly improve education or is it exaggerated?: 21 Sourced Claims, Contested Facts & Evidence", claim {id}, https://moyomap.com/maps/will-ai-significantly-improve.md#{id}. “Moyo, a product of SenTeGuard Inc.” ### Benjamin Bloom's 1984 research established that tutored students significantly outperform those in convention… Id: claim_1d29beafed951cf34cf0e03d Parent: topic_yVn6XTyAelumC104407Z Depth: 1 Benjamin Bloom's 1984 research established that tutored students significantly outperform those in conventional classrooms, with tutored students scoring 2 sigma higher and mastery-learning students scoring 1 sigma higher. Labels: Source-linked, Low, known ### Since 2023, randomized controlled trials of LLM-based AI tutors have been conducted in Nigeria, Ghana, the Un… Id: claim_41e2f77082d01051ee6ddb93 Parent: topic_yVn6XTyAelumC104407Z Depth: 1 Since 2023, randomized controlled trials of LLM-based AI tutors have been conducted in Nigeria, Ghana, the United States, Turkey, and England. Labels: Source-linked, Low, known ### A World Bank RCT in Nigeria found that students using an AI tutor (Microsoft Copilot/GPT-4) outperformed a co… Id: claim_4b77c4d295360896665459e4 Parent: topic_yVn6XTyAelumC104407Z Depth: 1 A World Bank RCT in Nigeria found that students using an AI tutor (Microsoft Copilot/GPT-4) outperformed a control group by 0.31 standard deviations on a combined assessment. Labels: Source-linked, Medium, known ### AI adoption in higher education is widespread and occurring rapidly, often outpacing institutional policy. Id: claim_7424c1fe68a671b58ab0d385 Parent: topic_yVn6XTyAelumC104407Z Depth: 1 AI adoption in higher education is widespread and occurring rapidly, often outpacing institutional policy. Labels: Source-linked, Low, known ### Pedagogical capability of AI models is increasing rapidly while maintaining constant inference costs. Id: claim_c21f9af3aefa7973c190cc60 Parent: topic_yVn6XTyAelumC104407Z Depth: 1 Pedagogical capability of AI models is increasing rapidly while maintaining constant inference costs. Labels: Inferred, Low, known #### A controlled trial found that teachers using AI saved approximately 25.3 minutes per week on planning activit… Id: claim_01627a22bc233a1533c148ec Parent: claim_4b77c4d295360896665459e4 Depth: 2 A controlled trial found that teachers using AI saved approximately 25.3 minutes per week on planning activities, a 31% reduction compared to non-users. Labels: Source-linked, Low, known #### A systematic review of 68 studies found a moderate positive effect of AI on learning, with a standardized mea… Id: claim_0aa120c71aea6c2b241d71f3 Parent: claim_7424c1fe68a671b58ab0d385 Depth: 2 A systematic review of 68 studies found a moderate positive effect of AI on learning, with a standardized mean difference of 0.45 across 337 effect sizes. Labels: Inferred, Low, known #### Providing human tutors for every student in the United States would require tens of millions of additional te… Id: claim_135ba8a86666354b7ae0f320 Parent: claim_1d29beafed951cf34cf0e03d Depth: 2 Providing human tutors for every student in the United States would require tens of millions of additional teachers, illustrating the scale of the labor constraint. Labels: Source-linked, Medium, known #### Recent studies on AI-powered tutoring show mixed results, with some researchers noting difficulties in isolat… Id: claim_1c63eecb81f278fcf8c29d07 Parent: claim_7424c1fe68a671b58ab0d385 Depth: 2 Recent studies on AI-powered tutoring show mixed results, with some researchers noting difficulties in isolating the specific impact of AI chatbots from broader program effects. Labels: Contested, Medium, known #### In a partnership between Google DeepMind and Eedi, over 75% of AI-generated messages for tutoring sessions di… Id: claim_3032029966ccfcffbc98d33d Parent: claim_4b77c4d295360896665459e4 Depth: 2 In a partnership between Google DeepMind and Eedi, over 75% of AI-generated messages for tutoring sessions did not require major revisions by human tutors. Labels: Source-linked, Low, known #### A full year of Khanmigo costs $44 and allows for up to ten children per subscription, making it significantly… Id: claim_3b2356e843980a3d7567441a Parent: claim_4b77c4d295360896665459e4 Depth: 2 A full year of Khanmigo costs $44 and allows for up to ten children per subscription, making it significantly cheaper than one hour of human tutoring. Labels: Source-linked, Low, known #### Bloom identified the primary barrier to universal one-to-one tutoring as economic, specifically the high cost… Id: claim_412948f6c8c6f3ff2d2a1695 Parent: claim_1d29beafed951cf34cf0e03d Depth: 2 Bloom identified the primary barrier to universal one-to-one tutoring as economic, specifically the high cost of labor, rather than a pedagogical limitation. Labels: Source-linked, Low, known #### A 2025 survey of 2,232 US public school teachers found that 60% had used AI tools for work, with weekly users… Id: claim_4473beb476c1c7cb863a846c Parent: claim_4b77c4d295360896665459e4 Depth: 2 A 2025 survey of 2,232 US public school teachers found that 60% had used AI tools for work, with weekly users estimating they save 5.9 hours per week. Labels: Source-linked, Low, known #### A February 2026 review by the Brookings Institution found that generative AI in tutoring led to learning gain… Id: claim_48a2145cea5812ac43085bbe Parent: claim_41e2f77082d01051ee6ddb93 Depth: 2 A February 2026 review by the Brookings Institution found that generative AI in tutoring led to learning gains, improved motivation, and increased efficiency. Labels: Source-linked, Low, known #### Khan Academy's AI tutoring system required a dedicated calculator tool for math and pre-processing pipelines… Id: claim_987a7cbe61086bc1334a3f68 Parent: claim_4b77c4d295360896665459e4 Depth: 2 Khan Academy's AI tutoring system required a dedicated calculator tool for math and pre-processing pipelines for visual content interpretation to function reliably. Labels: Source-linked, Medium, known #### 86% of higher education students globally report using AI in their studies, according to a Digital Education… Id: claim_9a5ee4bebe4f5940949bc56a Parent: claim_7424c1fe68a671b58ab0d385 Depth: 2 86% of higher education students globally report using AI in their studies, according to a Digital Education Council survey of 3,839 students across 16 countries. Labels: Inferred, Low, known #### Google's Gemma-3n E4B, an open-weight model designed for local devices, scores 64% on the CDPK benchmark. Id: claim_a8a22e4aa632cff74618ad2a Parent: claim_c21f9af3aefa7973c190cc60 Depth: 2 Google's Gemma-3n E4B, an open-weight model designed for local devices, scores 64% on the CDPK benchmark. Labels: Inferred, Low, known #### In the Harvard physics RCT, students using the AI tutor achieved a higher median post-score (4.5) compared to… Id: claim_b88a5420c1d10e7f1a4dd279 Parent: claim_41e2f77082d01051ee6ddb93 Depth: 2 In the Harvard physics RCT, students using the AI tutor achieved a higher median post-score (4.5) compared to students in the active learning classroom (3.5). Labels: Source-linked, Medium, known #### Testing of 97 LLMs on The Pedagogy Benchmark shows accuracy ranges from 28% (Llama-3.2 1B) to 89% (Gemini 2.5… Id: claim_c69a41001288676a114426cd Parent: claim_c21f9af3aefa7973c190cc60 Depth: 2 Testing of 97 LLMs on The Pedagogy Benchmark shows accuracy ranges from 28% (Llama-3.2 1B) to 89% (Gemini 2.5 Pro). Labels: Source-linked, Low, known #### On the Cross-Domain Pedagogical Knowledge benchmark, model performance at $0.10 per million tokens increased… Id: claim_ea4ce30c9710e64f1c1cc9af Parent: claim_c21f9af3aefa7973c190cc60 Depth: 2 On the Cross-Domain Pedagogical Knowledge benchmark, model performance at $0.10 per million tokens increased from 50% in April 2024 to 82% in June 2025. Labels: Source-linked, Low, known #### Bloom's tutoring results were driven by mastery learning—ensuring full understanding before progressing—and i… Id: claim_f4d2ad6e69e8d1abb866e308 Parent: claim_1d29beafed951cf34cf0e03d Depth: 2 Bloom's tutoring results were driven by mastery learning—ensuring full understanding before progressing—and increased time on task, rather than solely by the relational aspect of tutoring. Labels: Source-linked, Low, known