AI is transforming other industries. Why not education?
Other sectors are quickly and adeptly implementing AI to solve big problems. What's stopping the education sector from following suit?
I hope everyone had a restful, joyous holiday!
I was in South America the last few weeks and discovered this nice paper on AI in Uruguayan education. It’s a good AI overview and has a strong section on metrics that could be used to assess AI policies in schools. Uruguay has a basic literacy rate of nearly 100%, a good foundation for exploring how AI could close inequalities and accelerate learning. I continue to think that the U.S. should be learning alongside other countries on this topic.
Why Is AI Transforming Science Faster than Education?
I recently spent some time watching videos from the World Science Forum about the use of AI in science, and I’m fascinated by how AI is transforming other sectors. I think we have some real opportunities to apply lessons from these efforts to education.
One panel, led by Eric Topol (a leader in Precision Medicine), showcased how rapidly AI is advancing drug discovery, cell discovery, and the scientific process itself. Virtual labs and AI-powered scientists are accelerating breakthroughs that were once unimaginable. Another panel, while more technical, included a striking anecdote about a scientific discovery—one the human research team would have missed without AI. Artificial intelligence is clearly enabling a profound shift across scientific disciplines.
A particularly powerful example is AlphaFold, an AI system developed by DeepMind, which has solved a decades-old biological challenge: accurately predicting protein structures. Understanding how proteins fold is essential because their 3-D shapes determine their functions in the body. By accurately predicting these structures, scientists can better understand diseases, accelerate the discovery of new medical treatments, design more effective drugs, and even engineer plants to resist disease.
Imagine what a similar AI-driven leap would mean for education if we could predict how to help struggling students or optimize entire school systems.
Every time I watch videos like this, I wonder: why aren’t we having similar conversations in education? Why aren’t we seeing AI uncover equivalent breakthroughs?
The Data Problem
One major barrier is data—or lack thereof. Within education, privacy concerns, legal uncertainty, decentralized governance, and outdated data systems have created significant hurdles for AI experimentation and reliable modeling. By contrast, sectors like healthcare and drug development have made strides in developing data-sharing agreements and secure systems that allow AI to analyze vast datasets without compromising individual privacy.
At the World Science Forum, many discussions centered on innovative approaches to data development, such as advanced privacy protections, incentive structures for data sharing, and public-private partnerships that unlock large-scale benefits. These strategies could serve as a blueprint for education. The question is: why haven’t we tried harder to adopt them?
The Incentives Problem
Science attracts massive funding for AI research due to its potential for breakthroughs with immediate, measurable societal and economic impact. Education lags in funding for innovation, often focusing on small-scale studies or pilot programs. This isn’t by accident. Education doesn’t attract certain types of funders and investors because it is a publicly provided good and doesn’t lend itself to monetization.
Education systems are also burdened by entrenched policies, bureaucracy, and cultural resistance to rapid technological adoption. Teachers, unions, and policymakers often view new tools skeptically, fearing increased workload, inequity, or job displacement.
These are all major disincentives to tackle the same large-scale, focused research and coordinated data collection we see in science. But in the end, the stakes in education and also perhaps the opportunities are incredibly high. AI early adopter district leaders say teachers often embrace AI faster than other “novel” innovations because they quickly see how AI-enabled tools make their jobs easier. We may see that AI adoption mindsets are more open and malleable than we think, given AI’s ability to solve specific problems that have long plagued our field.
Can we find ways to shift the incentives, at least for the most critical questions threatening the next generation’s future? Can big federal agencies like the Institute of Educational Sciences (IES) or the National Academy of Sciences target federal funds to the most critical questions? Can private philanthropies be less risk-averse and more cutting-edge? Is there a role for public-private partnerships, and what would that look like?
Are We Asking the Wrong Research Questions?
Last year, I attended an international conference on AI and sat in on some fascinating research presentations. Most, however, were narrowly focused on the efficacy of specific AI tools in education—adaptive learning platforms, tutoring bots, and grading systems. Valuable, but incremental. Most AI applications in education (like adaptive learning platforms) target symptoms, such as test prep or content delivery, rather than root causes, like inequity, teacher burnout, or systemic inefficiencies. Where is the flood of research on how AI might crack the code on the foundational challenges that have long plagued public education? I believe that part of the problem is the lack of a coordinated learning agenda and funding strategy around AI in education, which should be transformative and targeted to the questions most urgent to closing racial and income learning divides. Susanna Loeb (at Stanford) and I have more ideas about this if anyone cares to chat.
The Opportunity for Education
Imagine if AI could:
Diagnose the root causes of disengagement in students.
Predict which interventions will work best for which learners and when.
Create scalable solutions for teacher professional development.
Identify systemic inefficiencies and propose actionable solutions to address them.
Help reconfigure student time and teacher, parent, and community assets to truly change how learning environments look and feel.
These aren’t far-fetched goals, but achieving them requires shifting how we approach education research and innovation. It also requires a collaborative effort to build the infrastructure necessary for AI to succeed.
A Call to Action
To catch up with other sectors, education needs to:
Invest in Data Infrastructure: Develop shared data systems that are secure, interoperable, and designed with input from educators and families to address privacy concerns.
Clarify and Incentivize Data Sharing: Create the “carrots and sticks” needed to encourage districts, states, and private organizations to contribute to shared data pools for the greater good. Provide more explicit guidance and template contract language for structuring data agreements so districts feel safe moving forward with more ambitious AI efforts.
Foster Cross-Sector Collaboration: Bring AI researchers, educators, and policymakers together to co-design solutions that address real-world challenges.
Fund Transformative Research: Support large-scale projects aimed at solving foundational problems rather than just improving existing tools.
Start the Conversation: Conferences, forums, and working groups should prioritize discussions about AI’s potential to revolutionize education—not just incrementally but fundamentally.
If we can mobilize the same level of urgency and creativity for education that we’ve seen in science, there’s no reason AI couldn’t help us solve some of the most intractable problems in public education. The breakthroughs are waiting—we just need the will to pursue them.
More Updates
A few more examples of how AI is transforming science and tech:
AI and Weather Forecasting
Accelerating Drug Discovery
Do we really need humans in the loop?
A new study found that ChatGPT was better at diagnosing medical problems than doctors, even those with access to ChatGPT. Here’s a fun podcast on Hard Fork and a NYT article about the study. The author opines that the future of what it means to go to a doctor and to BE a doctor may change dramatically. To my point above, this is an essential question in medicine and one we must ask in education. Can AI better identify students with learning disabilities than school counselors, for example? What other questions could we be asking at the intersection of education and AI but are not? Leave a comment or drop me a note.
More on Scientific Transformation via AI, but with a Cautionary Finding
A new study shows that AI can notably accelerate scientific discovery in an R&D lab: “AI-assisted researchers discover 44% more materials, resulting in a 39% increase in patent filings and a 17% rise in downstream product innovation. These compounds possess more novel chemical structures and lead to more radical inventions.” However, these gains come at a cost. More than 80% of scientists report reduced satisfaction with their work due to decreased creativity and skill underutilization. This is an issue worth watching and one with potentially significant implications for education.
The AI Gender Gap
New research suggests men are more likely than women to use AI. This might explain why teachers (a predominantly female profession) are slow to adopt. Schools may also want to keep this in mind as they develop AI literacy classes.
Does the World Need More Desert Orcas and Llamas? YES!
The Nazca Lines, sprawling ancient geoglyphs etched into Peru's desert, have baffled scientists for decades. These massive designs—ranging from hummingbirds to a fish-tailed cat—were first spotted from the air in the 1920s. Recently, AI has helped archaeologists discover 303 new geoglyphs in just six months, doubling prior finds. Using drones and machine learning to sift through 47,000 possibilities, the team pinpointed designs like knife-wielding orcas (center right above) and playful llamas. If the AI is hallucinating here, I am all for it.
Data privacy regulations and the ways organizations interpret them, are a significant source of hesitancy in K-12. Tools like this open source project from Vanderbilt might be the step needed for school districts to move forward with using AI for data analysis at scale: https://www.amplifygenai.org/.
The other significant barrier is data collection itself. Much of what we want to make decisions based on, like which intervention to prescribe, require fidelity data that isn't routinely collected. If we see success with an intervention, it can be currently difficult to know what was actually done as opposed to what was recorded or intended. Progress in making this data collection easier (and more accurate) is underrated.