Brian Yu

Misalignment between high school and college math expectations can slow down or even derail students who are otherwise prepared to succeed. Repeating content, being placed in unnecessary remediation, or facing inconsistent requirements can chip away at students’ confidence and momentum. Just Equations’ new report, "Aligned by Design", highlights five cross-system strategies states can use to expand options, enhance preparation, and remove barriers that get in students’ way. A thoughtful resource for anyone focused on smoother and more equitable transitions from high school to college: https://lnkd.in/gvkRjK2v #K12Math #HigherEd #MathEquity #CollegeAccess

25

3 Comments

📊 Crack the code of student thinking in math We’ve launched a Kaggle competition with Vanderbilt University to push the boundaries of NLP and machine learning in education. The challenge: - Build models that detect student math misconceptions from open-ended written responses. - Use a rich dataset from |= Eedi capturing real student reasoning. - Help create tools that deliver faster, more personalized feedback. Winning models won’t just classify errors—they’ll help educators better understand how students think, ultimately improving math learning for thousands. 👉 Learn more and join the competition: https://lnkd.in/eHQniJjv #AI #NLP #MachineLearning #EdTech #MathEducation

28

2 Comments

Could the way a student interacts with educational software in the first few hours predict their success months later? A new study suggests it's possible. “In education, we often are interested in delayed outcomes like end-of-the-year assessments, but it would be useful if we could predict those outcomes using shorter amounts of data from educational software platforms,” says Stanford HAI Faculty Affiliate Emma Brunskill. Read more about the study: https://lnkd.in/eA2fnpGS

33

I think one of the most powerful, and often misunderstood, outcomes of computer science education is computational thinking. Students who experience CS as foundational instruction aren’t just learning about computers or code. They’re developing durable ways of thinking that transfer across disciplines: breaking down complex problems, recognizing patterns, modeling systems, and iterating toward better solutions. As researchers note, “Computational thinking is about searching for ways of processing information that are always incrementally improvable in their efficiency, correctness, and elegance.” That kind of thinking supports success in any pathway students choose. Yet CS is still too often framed as narrow or technical. Centering CS around computational thinking can help to break this narrative and open access to skills all students need, not just those headed into tech fields. I appreciate this article by Li et al. as it works to clearly articulate and define computational thinking in ways that are helpful for broader educational conversations about teaching, learning, and student success.

43

2 Comments

Those involved in tutoring and edtech are familiar with the "5% Problem"-- Only a small fraction of students typically engage with tutoring systems long enough to see real learning gains. So, how do we reach the other 95%? My latest article, featured by the Learning Engineering Virtual Institute (LEVI), explores how PLUS Tutoring, Khanmigo, and |= Eedi are getting students to engage more in tutoring. https://lnkd.in/eHsriKnr Simon Woodhead, Bibi Groot, Kristen Eignor DiCerbo, Ken Koedinger, Vincent Aleven, Emma Brunskill, Lee Branstetter, Conrad Borchers, Erin Gatz, PhD, Shivang Gupta, Ashish Gurung, John Whitmer, CMU Human-Computer Interaction Institute

34

4 Comments

Higher ed email is not failing because students stopped checking their inbox. It is failing because most institutions still send the same message to thousands of students and optimize after the fact. Batch sends, shallow personalization (Hi {first name},), and post-send A/B tests destroy engagement and ignore real intent signals. The issue is not subject lines or volume. It is that email strategies are built on guessing instead of predicting relevance before the send. If we want better engagement and yield, individual-level decisioning has to replace traditional campaigns. We break this down in The 2026 Higher Ed Email Optimization Guide. https://lnkd.in/ghT-a2Ea

23

I spoke with Vauhini Vara for her Bloomberg Businessweek article on AI in education. The article maps the ecosystem of Big Tech and K-12 education, detailing the many ways private interests are working hard to capture education. You should be able to read the article at: https://lnkd.in/g9xSQ22Q And I'll say it again: a future determined by Big Tech and its AI technologies is not inevitable. Over at Civics of Technology Project, we're developing a community for resisting, refusing, repurposing, and reimagining technology in education. Join us!

62

2 Comments

The AI Innovation Index launched with a clear purpose: give school systems a coherent AI leadership roadmap and the ability to measure progress — relative to peers across the country — on leader, staff, and student metrics. This past fall, a cross-sector mix of school systems participated, generating important context for these school systems and first-of-its-kind system-level data for the ecosystem. Now, we're excited to launch our spring data cycle and grow the national community of school systems working to ensure all students flourish — personally and professionally — in a world of accelerating AI integration. With today’s launch of our spring data cycle, school systems can expect: ➡️ More options for staff and student data collection, including flexible survey distribution methods ➡️ Customizable fields for deeper disaggregation (by department, participation in specific training programs, and more) ➡️ New analytics featuring usage and impact data for AI software and use cases implemented by peers across the country ➡️ A new resource library with guidance documents and leader video testimonials, all aligned to our leadership roadmap The Index is both public infrastructure for AI innovation data and a community of leaders committed to learning from each other. Are you a school system leader interested in learning more? Participation is free. Sign up and explore the leadership portal, with no obligation to participate. And check out the video below for what to expect. Know a superintendent, CIO, or cabinet-level leader who should be part of this? Tag them in the comments! Thanks for helping spread the word, friends and partners: Yusuf Ahmad, Mera Dougherty, Alex Kotran, Emma Doggett Neergaard, Vera Cubero, Amy Chen Kulesa, Emily Shisler, Dr. Michelle Watt, LeeAnn Lindsey, Ed.D., Eric Chan, Elena Thurman, Ian Connell, Bree Dusseault, Dr. Beth Holland, EdD, Beth Rabbitt, Chris Gibbons, Travis Franklin, Margo Roen, Jessica Paulsen, Davis Zaunbrecher, Shawn Rubin, Michelle Culver, Nate Kerr, Mike Yates, Sunanna Tara Chand, Ross Lescano Lipstein, Elena Sanina, Caitrin Wright, Jessica Wilson, Erica Murphy, Ji Soo Song

55

14 Comments

📣 New guide just dropped: How to use GenAI to actually support math learning (not replace it). AI for Education + Student Achievement Partners nailed it with this one, two thumbs up! Key takeaways: – Use LLMs for planning, examples, and explanations. – Use math-specific tools for anything involving calculation. – Focus on productive struggle, AI should support thinking, not short-circuit it. – At the elementary level, keep GenAI use mostly on the teacher side for now. ✅ Use GenAI for: – Drafting differentiated tasks – Generating warm-ups, translations, or common misconceptions – Tutoring if it’s scaffolded well (oh hey, Flexi.org) – Designing real-world, student-relevant problems 🚫 Avoid: – Relying on LLMs for math accuracy – Turning struggle into shortcut – AI-first instead of student-first planning Big question to ask: What are students learning from this task—and how can AI support that learning, not replace it? 📚 Link to the full guide: https://lnkd.in/gFgMCeVD #AIinEducation #MathEducation #Teaching

16

Yesterday’s NYT explored Alpha School as one model of AI in education. Here’s what I'd argue that model misunderstands -- and gets right. Two concerns, based on Alpha’s video description of their AI-assisted academic model called "2 Hour Learning" (video link in comments): 1 Alpha’s student dashboard includes a “waste meter” to ensure students compress their academic work into two hours a day. This measures the amount of time students waste so they can get more efficient in their learning. Does that imply to students that learning means constant attention and that anything else their minds do during those two hours is wasted time? After serious thinking, I often need to take a walk or just take a break, but that never strikes me as wasted time. I’d learn less if I didn’t “waste” that time. 2 The video ends saying “by mastering their studies in just two hours a day, our students have more time to develop crucial life skills and pursue their passions.” Does that imply to students that academics are just a hoop AI can help them jump through on the way to life skills and passions? If so, what does that leave them thinking about the relationship between academic skills and life skills? What about the long journey toward mastery in any endeavor, which requires lots of hard-won skills, academic and otherwise? What if some of those skills take more than two hours a day to learn? One place where it looks like we agree: High school students at Alpha work on “‘masterpieces’, a time-intensive project coinciding with the student’s passion.” (Link to NYT article in comments.) I run a purpose- and career-exploration program for high schoolers in which they spend months working on “purpose projects” that sound very similar. The goal is to help students graduate knowing more about the intersection of their skills, interests, ethical compass, and potential career paths. I incorporate both AI tools and AI literacy into those purpose projects, so students also graduate knowing how and when to use – and not use – AI. I hope students leave my classroom knowing learning can sometimes be AI-assisted but is always human-powered. And I hope they leave knowing that masterpieces are made when we see learning and growth as a long path we’re walking down, not as a series of hoops we need to jump through, expecting to find our motivation and our masterpiece waiting on the other side. 

74

19 Comments

If 20 years and $100MM+ in research and development builds an online math practice problem system that supports typical learning gains of 0.03SD, what can we expect from other edtech products? What can we expect from AI improvements to edtech products? https://lnkd.in/eGK-DN9p Khan Academy deserves major kudos for partnering with a very talented group of education researchers to study the effects of KA on typical students. The researchers had access to 200,000 students, their KA usage, and their MAP scores. There are two headline findings: First, typical usage is very low. "In our final panel, 24% of students have zero platform usage, 65% spend under 10 min per week (6 h per year), and only 10% meet KA’s recommendations of 18 h a year (approximately 30 min per week)." [[10% completion rate for the MOOC fans out there...]] Second, learning gains are minimal and disproportionately benefit more high achieving students. The students in the minimal group are predicted to get about .03 SD gains (.2SD in algebra is like moving a student from the 50th percentile to the 58th percentile, so about 1/10th of that). Students in the recommended group see about .08 SD gains. It's not nothing! Let's celebrate the incremental gains where we find them! But man, we've been working on this online practice problem model for 40 years, and I'm not sure the ROI is panning out here. If think AI can help, what's your bet on the value add? If you think AI can make something like KA 2x better, which is probably impossible, then you are still measuring effects with microscopes. You really need to be getting 5x or 10x gains to see viability, which-just as with magical ponies-is very fun to imagine. Huge kudos again to KA and the research team for enabling this research and being transparent. It's enormously helpful, even if it's ultimately a sad paper to read. If you've got another spin on the findings, I'd be very interested to hear...

379

25 Comments

I'm excited to offer a new course at Harvard University extension school this coming Spring: Mathematics for AI and Machine Learning (Math E-142). The goal of this semester-long online course is to teach the math needed to understand how AI works "under the hood". We start with an intensive review of concepts from linear algebra, analytic geometry, vector calculus, optimization, and probability, and then apply them in detail to machine learning methods such as regression, dimensionality reduction, density estimation with Gaussian mixture models, and classification with support vector machines. https://lnkd.in/eeGPJNgx

30

1 Comment