{"id":52379,"date":"2026-04-01T14:02:06","date_gmt":"2026-04-01T12:02:06","guid":{"rendered":"https:\/\/blinklearning-blog.wildoomlab.com\/ai-in-education-why-access-isnt-enough-and-can-be-counterproductive\/"},"modified":"2026-07-10T07:45:27","modified_gmt":"2026-07-10T05:45:27","slug":"ai-in-education-why-access-isnt-enough-and-can-be-counterproductive","status":"publish","type":"post","link":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/ai-in-education-why-access-isnt-enough-and-can-be-counterproductive\/","title":{"rendered":"AI in education: Why Access isn\u2019t enough and can be counterproductive)"},"content":{"rendered":"\n

What scientific evidence reveals about the real impact of AI on learning (and what schools and publishers need to decide)<\/strong><\/h2>\n\n

EduIA pulse #1 | research on AI and learning in education by BlinkLearning<\/h2>\n\n

What Research Tells Us About AI and Learning This Month<\/strong><\/h2>\n\n

AI in education is transforming learning, but there\u2019s a question everyone in education is asking right now: does AI really teach, or does it simply provide answers?<\/strong><\/p>\n\n

This month, four key studies bring us closer to an answer. Spoiler: AI can be transformative in the classroom, but only if it\u2019s well designed and used. And what \u201cwell\u201d actually means is more complex \u2014 and more interesting \u2014 than it seems.<\/p>\n\n

1. AI improves outcomes\u2026 but are students actually learning?<\/h2>\n\n

“The Evidence Base on AI in K-12: A 2026 Review”<\/strong> <\/strong>Fesler, Martinez, Agnew & Loeb \u2014 AI Hub for Education, Stanford University (2026)<\/em><\/p>\n\n

Stanford has just published the most comprehensive review to date on the impact of AI in K-12 education. They analyzed more than 800 academic studies and found that only 20 provide strong causal evidence<\/strong>. A finding that calls for collective humility.<\/p>\n\n

What do those 20 studies tell us? That AI in education improves student performance while they have access to it\u2014in math, programming, and writing. But when they are assessed without that support, the effects fade or disappear. Researchers call this the difference between assisted performance and durable learning.<\/p>\n\n

\n
\n
\n
\n
\n
\n
\n

Two findings deserve special attention. First: pedagogical design matters more than the technology itself. <\/strong>Chatbots that guide students step by step\u2014rather than giving direct answers\u2014lead to stronger learning outcomes. Second: AI in education works very well for teachers. Teachers who used ChatGPT to prepare lessons reduced their planning time by 30%<\/strong> without losing quality, according to Roy et al. (2024). And less experienced tutors were the ones who benefited most from AI-powered feedback systems.<\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

\n
\n
\n
\n
\n
\n
\n

Current evidence suggests that tools designed to foster independent reasoning are more likely to support durable learning.\u201d \u2014 Fesler et al., Stanford, 2026<\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

2. Beyond the grade: how to measure learning with AI<\/strong><\/h2>\n\n

New tools to understand AI and learning outcomes\u201d<\/strong> \u2014 OpenAI + University of Tartu + SCALE Initiative, Stanford (March 2026)<\/p>\n\n

OpenAI published this month the results of a study involving more than 300 university students who used ChatGPT\u2019s Study Mode to prepare for microeconomics and neuroscience exams. In microeconomics, students with access to the tool scored 15% higher<\/strong> than the control group.<\/p>\n\n

But the most valuable part of the article isn\u2019t those results\u2014it\u2019s the honesty about its limitations. The OpenAI team acknowledges that an exam score is not enough to determine whether AI in education improves learning<\/strong>. That\u2019s why they developed the Learning Outcomes Measurement Suite, a longitudinal measurement framework designed together with Stanford and the University of Tartu (Estonia), where it is being tested with nearly 20,000 students aged 16 to 18<\/strong>.<\/p>\n\n

This system measures things that, until now, no one was measuring: autonomous motivation, persistence in the face of difficulty, metacognition, and long-term retention. A paradigm shift the sector urgently needed.<\/p>\n\n

\u201cWhat truly matters is whether improvements persist over time.\u201d \u2014 OpenAI team, 2026<\/p>\n\n

3. The real impact lies in the learning sequence<\/strong><\/h2>\n\n

“Effective Personalized AI Tutors via LLM-Guided Reinforcement Learning”<\/strong> Chung, Zhang, Kung, Bastani & Bastani \u2014 Universidad de Pennsylvania + National Taiwan University (2025)<\/em><\/p>\n\n

This study is one of the most rigorous published to date: a randomized controlled trial with 770 high school students in Taipei over five months, teaching Python. All students had access to the same AI chatbot. The only difference: half received practice problems in a personalized sequence using reinforcement learning<\/strong>, adapted in real time to their level.<\/p>\n\n

\n
\n
\n
\n
\n
\n
\n

The result was an improvement of 0.15 standard deviations<\/strong> on the final exam without AI assistance, which\u2014according to some studies\u2014is equivalent to an additional 6 to 9 months of schooling<\/strong>. And without increasing instructional time or teacher workload.<\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

The most surprising part: students in the personalized group didn\u2019t solve more problems or harder ones. What changed was their engagement<\/strong>: they were more involved, persisted longer, and used the chatbot more productively\u2014asking questions to understand rather than requesting direct answers.<\/p>\n\n

The benefit was greater for beginner students and those from lower-performing schools<\/strong>, opening up significant opportunities for educational equity.<\/p>\n\n

\n
\n
\n
\n
\n
\n
\n

\u201cPersonalizing the learning experience\u2014not the chatbot itself, but the sequence of problems\u2014can sustainably improve engagement.\u201d \u2014 Chung, Bastani et al., 2025<\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

4. The risk of AI: cognitive offloading<\/strong><\/h2>\n\n

\u201cThinking\u2014Fast, Slow, and Artificial: how AI is reshaping human reasoning and the rise of cognitive surrender\u201d<\/strong> \u2014 Shaw & Nave, The Wharton School, University of Pennsylvania (2025)<\/p>\n\n

The most conceptual article of the month\u2014and perhaps the most important for any educator. Wharton researchers propose the Tri-System Theory<\/strong>: humans no longer think only with System 1 (fast intuition) and System 2 (slow, deliberate reasoning), as Kahneman suggested. There is now a System 3: AI.<\/p>\n\n

And System 3 has a serious problem: it produces what the authors call \u201ccognitive surrender<\/strong>.\u201d In three experiments with 1,372 participants, when the AI provided an incorrect answer, participants still followed it in 80% of cases<\/strong>. Most concerning, access to AI increased participants\u2019 confidence<\/strong> in their answers\u2014even when they were wrong.<\/p>\n\n

Who is most vulnerable? Individuals with higher trust in AI and a lower tendency to question their own thinking. What helps? Incentives and immediate feedback reduce cognitive surrender\u2014though they don\u2019t eliminate it entirely.<\/p>\n\n

For educators, the key question is: are we teaching students to think with AI, or to delegate to it without thinking?<\/strong><\/p>\n\n

\u201cWe don\u2019t just use AI\u2014we think with it. And that changes who authors our decisions.\u201d \u2014 Shaw & Nave, Wharton, 2025<\/p>\n\n

The big idea of the month: access to AI is not enough<\/strong><\/h2>\n\n

\n
\n
\n
\n
\n
\n
\n

These four studies, published between 2025 and 2026, all point in the same direction: access to AI is not enough\u2014and can even be counterproductive if not managed properly.<\/strong><\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

What makes the difference is design: tools that guide without giving answers, that adapt difficulty to the learner, that measure learning over time, and that keep critical thinking active. This is not technological magic. It is intelligent pedagogy.<\/p>\n\n

How to integrate AI with intention: the Blinklearning approach<\/strong><\/h2>\n\n

\n
\n
\n
\n
\n
\n
\n

At BlinkLearning, we have spent years working on exactly that: technology in service of learning, not the other way around. <\/strong>And each month, the research reminds us why this principle matters.<\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

\n
\n
\n
\n
\n
\n
\n

The key question for your school<\/strong><\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

<\/div> \n
\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/h2>\n\n

\n
\n
\n
\n
\n
\n
\n

If your school or institution is using AI in the classroom, how do you know whether students are actually learning more\u2014or simply performing better with support? <\/strong>Do you have a way to measure it?<\/p> \n \n<\/div> \n<\/div> \n<\/div> \n<\/div> \n

\n
<\/div> \n<\/div> \n<\/div> \n<\/div> \n<\/section><\/div> \n
<\/div><\/p>\n\n

Nos encantar\u00eda leer tu experiencia en los comentarios. <\/p>\n\n

\n\n

EduIA Pulse es la publicaci\u00f3n mensual de BlinkLearning sobre investigaci\u00f3n en IA y educaci\u00f3n.<\/strong> Si te ha resultado \u00fatil, comp\u00e1rtela con alguien de tu equipo.<\/em><\/p><\/blockquote>\n\n

Fuentes:<\/em>\u2014 Fesler et al. (2026). The Evidence Base on AI in K-12. Stanford \/ AI Hub for Education.<\/em>\u2014 OpenAI + U. Tartu + SCALE (2026). Learning Outcomes Measurement Suite.<\/em>\u2014 Chung, Bastani et al. (2025). Effective Personalized AI Tutors via LLM-Guided RL. U. Pennsylvania \/ NTU.<\/em>\u2014 Shaw & Nave (2025). Thinking\u2014Fast, Slow, and Artificial. Wharton, U. Pennsylvania.<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"

What scientific evidence reveals about the real impact of AI on learning (and what schools and publishers need to decide)…<\/p>\n","protected":false},"author":2,"featured_media":52378,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[343,345],"tags":[238,226,247],"tipologia":[],"class_list":["post-52379","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","category-responsible-technology","tag-blinklearning-en","tag-eduia","tag-educacion-en"],"_links":{"self":[{"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/posts\/52379","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/comments?post=52379"}],"version-history":[{"count":1,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/posts\/52379\/revisions"}],"predecessor-version":[{"id":52952,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/posts\/52379\/revisions\/52952"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/media\/52378"}],"wp:attachment":[{"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/media?parent=52379"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/categories?post=52379"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/tags?post=52379"},{"taxonomy":"tipologia","embeddable":true,"href":"https:\/\/blinklearning-blog.wildoomlab.com\/en\/wp-json\/wp\/v2\/tipologia?post=52379"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}