Beyond 2030: Can AI Help Us Dismantle the Factory-School—and Build Something Human?
A brief tour of our paper on four AI-in-education grand challenges—and how I’m approaching them as an educator focused on culture, access, and equity.
Sanskriti is maintained by Sai Gattupalli, Ph.D., a learning sciences researcher and educator focused on AI-integrated learning environments, culturally responsive computing, and educational equity. Sai earned his doctorate in Learning Technologies from the University of Massachusetts Amherst, where he taught College Writing and collaborated on intelligent tutoring and multimodal learning projects. He writes at the intersection of culture, education, and technology and creates classroom-ready resources—including Equations & Echoes, a YouTube channel of AI-generated STEM music for young learners and teachers.
Written by Sai Gattupalli, Ph.D.
What if the real promise of AI in education isn’t faster grading or flashier apps, but a chance to redesign learning around people instead of classrooms?
TL;DR (Key takeaways)
Four grand challenges we propose: pedagogical innovation, closing the digital divide, global learning communities, and data-driven decision-making
Evidence matters: intelligent tutoring systems (ITS) reliably lift outcomes (often ≈50th→75th percentile), but must be deployed with human oversight (such as Kulik & Fletcher, 2016)
Embodiment & culture: learning blooms when it’s multimodal, social, and culturally responsive (e.g., the Wearable Learning computing education platform by Arroyo et al., 2017)
Equity by design: AI can widen gaps unless we intentionally build diverse datasets, transparent governance, and people-centred ecosystems (see the Troyes Declaration on aslerd.org)
I’ve spent the last year working with a brilliant team at UMass Amherst: Beverly Woolf, Danielle Allessio, Ivon Arroyo, and Boming Zhang—on a paper that asks a simple, hard question: what must AI actually do for learning by the 2030s? Our answer crystalizes into four grand challenges, each grounded in research and lived classroom realities. You can browse the issue and DOI here: http://dx.doi.org/10.55612/s-5002-064-001sp
Pedagogical innovations that put learners first.
Too much “AI in ed” still props up a transmissive, one-to-many model. We argue for tutors, copilots, and agents that adapt in real time—without displacing teachers’ judgment. Decades of work on intelligent tutoring systems shows consistent gains (think a shift from the 50th to ~75th percentile), but the point is not automation for its own sake; it’s personalized support with a human in the loop.
I’m especially excited by embodied, multimodal approaches that connect mind, body, tools, and peers—such as the Wearable Learning program, which turns math into collaborative, full-body problem solving. This isn’t edutainment; it’s cognition made tangible.Addressing the digital divide (intentionally).
AI can translate, scaffold, and personalize—but it can also scale bias if we train on narrow datasets or design for a single context. Our paper emphasizes inclusive data/UX practices and community-level capacity building. That aligns with the Troyes Declaration vision for people-centred smart learning ecosystems: technology serves wellbeing, not the other way around.Building global learning communities.
When tools are open, transparent, and co-developed with teachers, we can reach many more learners. Prior meta-reviews show computer tutors can approach human-tutoring effectiveness in specific contexts—useful when expert teachers are scarce. But community, culture, and teacher craft remain irreplaceable.
Data-driven decision-making—without losing the plot.
Learner performance logs, affect and problem solving signals, and learning analytics can surface who’s stuck, bored, or ready to fly. The goal isn’t surveillance; it’s better, earlier support. Data should inform teacher moves, not dictate them. In practice, I favor “explainable nudges” over opaque scores, and opt-in consent over blanket capture.
Where culture fits (my lens).
Long-time readers know this blog Sanskriti sits at the intersection of culture, education, and technology. So my north star is simple: Does this system help learners express their cultural identities and think critically—together? If yes, keep it. If not, re-design it. That’s why embodied tools, bilingual supports, and co-creation spaces feel non-negotiable.
Teacher time is sacred.
One practical takeaway from the paper is that AI should give teachers time back: generate draft questions, differentiate reading levels, summarize exit tickets—then get out of the way. In blended settings, that’s where I see the most durable wins.
The guardrails.
Privacy, bias, and governance aren’t footnotes. They’re design constraints. Start with diverse datasets, publish model cards, enable local fine-tuning when possible, and involve students/parents in the loop. None of this is “extra work”—it is the work.
Our Call to Action
If you’re an educator (K-12 or higher ed), try one small change: pick a unit and add one embodied or bilingual activity, plus an AI-supported feedback loop. Tell me what changed—in learning and in belonging.
Interested readers can read the full paper here: Grand Challenges in AI and Education Beyond 2030
Banner image source: Shubham Dhage on Unsplash
Until next time.
References
Arroyo, I., Micciollo, M., Casano, J., Ottmar, E., Hulse, T., & Rodrigo, M. M. (2017, October). Wearable learning: multiplayer embodied games for math. In Proceedings of the annual symposium on computer-human interaction in play (pp. 205-216). https://doi.org/10.1145/3116595.3116637
Kulik, J. A., & Fletcher, J. D. (2016). Effectiveness of Intelligent Tutoring Systems: A Meta-Analytic Review. Review of Educational Research, 86(1), 42-78. https://doi.org/10.3102/0034654315581420
