Measuring what actually matters in Edtech

Why we can't teach AI Literacy yet

We asked one of the most respected education technology researchers in the world a simple question: how should schools teach students to use AI? His answer? We don't know yet, and pretending we do is the problem. Justin Reich is an Associate Professor at MIT and Director of the MIT Teaching Systems Lab [https://tsl.mit.edu], author of Failure to Disrupt [https://tsl.mit.edu/books/failure-to-disrupt/] (Harvard University Press), host of the TeachLab podcast, and the force behind The Homework Machine — a landmark 7-part podcast series investigating what's really happening with AI in classrooms across the US. In this conversation with Svenia Busson, Justin explores: * Why "AI literacy" follows the same broken playbook as digital citizenship and computational thinking — and will likely fail students the same way * What the history of web literacy teaches us: it took 25 years to find strategies that actually work * Why domain expertise — not AI knowledge — may be the most critical factor in using AI well * What to make of "AI-powered" schools like Alpha School * What students themselves are saying: two-thirds of US students say AI is harming their critical thinking * Why "the homework machine" is the most honest name for what's happening in classrooms today Also mentioned in this episode: * Mike Caulfield's SIFT framework for teaching web literacy: https://guides.lib.uchicago.edu/c.php?g=1241077&p=9082322 * A Guide to AI in Schools: Perspectives for the Perplexed — MIT Teaching Systems Lab guidebook (August 2025) https://tsl.mit.edu/wp-content/uploads/2025/08/GuideToAIInSchools.pdf * "Stop Pretending to Know How to Teach AI" — Justin's article in the Chronicle of Higher Education (November 2025) https://www.chronicle.com/article/stop-pretending-you-know-how-to-teach-ai * The Homework Machine podcast: https://podcasts.apple.com/us/podcast/the-homework-machine-what-ai-is-really-doing-in-classrooms/id583456652?i=1000747231954 (a must listen)

Measuring the real impact of AI in education

What does it actually take to know if an AI tutor is helping kids learn? Bibi Groot [https://www.linkedin.com/in/bibi-groot/], Chief Impact Officer at Eedi Labs [https://eedi.com/], has spent her career answering exactly that question — first at the Behavioral Insights Team (aka the Nudge Unit, co-founded with Nobel laureate Richard Thaler), then in classrooms across the UK and Latin America. In this episode, Bibi walks us through how Eedi's diagnostic engine works — 60,000 carefully designed multiple-choice questions, each distractor linked to a specific misconception — and why understanding why a student gets something wrong matters as much as knowing they got it wrong. Bibi also introduces a concept that should alarm everyone in edtech: cognitive surrender — the risk that when AI does all the thinking, students stop learning altogether. Her solution is architectural: don't ask students to self-regulate, build the constraints directly into the system. She references a striking study by Poulidis and Bastani on chess students — those who received AI hints at system-chosen moments improved 64% vs. only 30% for those who could ask for help whenever they wanted. This is a rare, rigorously evidence-based conversation about what responsible AI tutoring actually looks like — and how far most of the field still has to go. References mentioned in this episode: * Behavioral Insights Team (the Nudge Unit) [https://www.bi.team/] * Eedi Labs [https://eedi.com/] — including the free Eedi School platform [https://www.eedischool.com/us] * Google DeepMind's LearnLM [https://cloud.google.com/solutions/learnlm] * Learning Engineering Virtual Institute (LEVI) [https://learning-engineering-virtual-institute.org/] — created by Schmidt Futures & Renaissance Philanthropy * David Yeager, [https://casbs.stanford.edu/10-25-science-motivating-young-people-groundbreaking-approach-leading-next-generation-and-making]10 to 25: The Science of Motivating Young People [https://casbs.stanford.edu/10-25-science-motivating-young-people-groundbreaking-approach-leading-next-generation-and-making] * AI Hub for Education [https://scale.stanford.edu/research-in-action/understanding-evidence-base-ai-k12-education] (Stanford) — reviewed 800+ papers on AI in education; only 20 had causal evidence * Poulidis & Bastani chess study [https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5128584](system-chosen AI hints → 64% improvement vs. 30% for on-demand help) * London EdTech Week [https://www.londonedtechweek.com/] — Meet Bibi & Svenia at the London AI & Education Meetup on June 18, 2026

Making AI safe for children before it's too late

The tech industry is building powerful AI tools for children, often without understanding how children actually learn and grow. That's the gap Anne-Sophie Seret set out to close. Anne-Sophie is the co-founder and Executive Director of everyone.ai [https://everyone.ai], a Silicon Valley nonprofit bridging artificial intelligence and developmental neuroscience. She is also the Chief Program Officer of iRAISE (International Research-driven Alliance for AI Serving Every child), the global coalition she launched at the Paris AI Action Summit alongside 11 governments, UNESCO, UNICEF, and companies including OpenAI, Anthropic, and Google. In this episode, she and Svenia explore why children's brains are not mini adult brains, and why that changes everything for AI design. They discuss the critical developmental windows AI is currently disrupting (0–6 for language acquisition; 12–14 for social skills development), what the research on teenagers and anthropomorphic AI actually shows, and where the line is between AI as a scaffold and AI as a crutch. Anne-Sophie also shares the story of how iRAISE was built in just three months, what a "proactive" approach to AI safety looks like in practice, and why regulating AI is actually easier when children are the focus. She also previews the AI Safety Builder, a new science-backed tool launching at VivaTech that helps EdTech founders evaluate how their conversational AI interacts with children, detecting anthropomorphic, interactional, and relational risk cues based on the work of 30+ researchers. Resources mentioned: * everyone.ai — nonprofit at the intersection of AI and child development * iRAISE Coalition — launched at the Paris AI Action Summit (February 2025) https://parispeaceforum.org/initiatives/beneficial-ai-for-children-coalition/ * Research: "Adolescents & Anthropomorphic AI: Rethinking Design for Wellbeing" https://everyone.ai/wp-content/uploads/2026/02/Adolescents-Anthropomorphic-AI-Rethinking-Design-for-Wellbeing-.pdf * Research: "Mapping of generative AI impacts on child development" — mapping of risks and opportunities by age group, contributed to the G7 agenda https://everyone.ai/wp-content/uploads/2026/05/Mapping-of-GenAI-impacts-on-child-development-1.pdf * Book recommendation: Love to Learn by Isabelle Hau (Stanford) https://www.isabellehau.com/

Future of work: A Gen Z wake-up call

Kashyap "Kash" Rajesh is 20 years old, a Junior at Cornell University studying Information Science and Government with a minor in AI, and he's been working in AI policy since he was 14. He supported the founding of Encode, a non-profit originally founded by young people, focused on how AI is impacting the public and particularly the next generation, which grew to 40 states and every inhabited continent. As VP, he helped lead and grow the organization, which advised the White House Office of Science and Technology Policy's AI Bill of Rights and filed an FTC complaint against AI companion app Replika. He now supports the Rithm Project, a research and movement-building org focused on pro-social AI and human connection, and is involved in research at the Cornell Brooks Tech Policy Institute and the JEM Lab for Generative AI at Work. In this episode, Kash talks with Svenia Busson about: * Why entry-level jobs may outlast middle management in the AI transition — and what Gen Z should do about it * The Game Plan playbook Encode created to help Gen Z navigate the future of work (four archetypes: the Sleeper, the Anchor, the Tactician, and the Shaper) * The loneliness crisis that preceded generative AI — and how AI is amplifying, not creating, it * The Rithm Project's youth research report identifying nine portraits of how young people relate to AI chatbots * AI sycophancy — and what it quietly does to a generation's capacity to be wrong * The wave of state-level AI safety legislation: California's SB 53, the New York RAISE Act, and Illinois House Bill 315 * Why the Take It Down Act matters and how non-consensual deepfake imagery is already a crisis in schools A rare, honest, and deeply informed voice from inside the generation most affected by AI. Links mentioned: * ENCODE https://encodeai.org/ * The Rithm Project: https://www.therithmproject.org/ * Cornell Brooks Tech Policy Institute: https://publicpolicy.cornell.edu/btpi/ * Surgeon General's Advisory on Loneliness and Isolation (2023): https://www.hhs.gov/sites/default/files/surgeon-general-social-connection-advisory.pdf * California SB 53 / Illinois House Bill 315 / New York RAISE Act

Sorbonne's AI college for humanities students

What if AI education wasn't just for engineers and computer scientists, but for every student, regardless of their field? That's exactly the bet Camille Salinesi is making at one of the world's most iconic universities. Camille is a full professor of computer science at the University of Paris 1 Panthéon-Sorbonne [https://www.pantheonsorbonne.fr/], where he has been based since 1999. A specialist in requirements engineering and applied AI, he has published over 250 peer-reviewed papers. He heads the university's AI Observatory [https://www.pantheonsorbonne.fr/] alongside legal scholar Célia Zolynski. This autumn, he co-launches the Collège de l'IA [https://www.pantheonsorbonne.fr/actualite/le-college-de-lia-formation-inedite-tous-etudiants-la-licence] — France's first undergraduate-level AI diploma designed not for STEM students, but for students in law, history, philosophy, economics, and the arts. The programme, backed by France 2030 [https://www.pantheonsorbonne.fr/actualite/paris-1-pantheon-sorbonne-obtient-5-millions-deuros-developper-projet-aisorb], will give bachelor students 200 hours of AI training layered on top of their existing degree. In this conversation with Svenia Busson, Camille discusses: * Why AI literacy is as urgent for a law student as for a software engineer * The critical shift in information systems engineering from reliability to trust * How the Sorbonne is rethinking assessments in the age of AI — and why students themselves are demanding it * The difference between students who use AI to cheat and those who use it to learn * What the future of software engineering jobs actually looks like