HAI Seminar: Learning by Creating – A Human-Centered Vision for AI in Education

Why It Matters

By reorienting AI from a shortcut to a collaborative partner, educators can foster deeper cognition and creativity, reshaping how technology supports learning outcomes.

Key Takeaways

• AI should augment, not replace, deep learning processes.
• Knowledge‑building emphasizes creation, not mere knowledge‑passing in educational contexts.
• Script & Shift offers layered, non‑linear writing interface with AI “friends.”
• Protecting cognition requires tools that scaffold, not automate, writing.
• Designing AI for teachers empowers them as learning environment designers.

Summary

The seminar, led by Stanford assistant professor Hari Subbanam, presented a human‑centered vision for artificial intelligence in education, arguing that generative AI should be used to deepen learning rather than merely increase efficiency.

Subbanam contrasted two historic paradigms: knowledge‑building, where learners create artifacts and co‑construct understanding, and knowledge‑passing, an efficiency‑driven model likened to factories. He warned that most current ed‑tech startups replicate the latter, automating quizzes, grading, and even homework submission, as exemplified by the “Einstein” AI that logs into Canvas and completes assignments.

Drawing on a 1934 geography exercise by Lucy Sprague Mitchell, he illustrated how students can engage in authentic inquiry. He then introduced “Script & Shift,” a layered writing environment that provides metadata, content, and scratchpad layers, and integrates AI “friends” for tone, ideas, detail, and structure, enabling non‑linear drafting and targeted feedback.

If adopted, such tools could preserve learner cognition, empower teachers as designers of learning experiences, and shift the ed‑tech market toward AI that scaffolds creation. This reorientation promises richer critical‑thinking skills and more meaningful student‑generated knowledge.

Original Description

Generative AI is entering classrooms at a breathtaking pace, often presented as a solution for efficiency by automating tasks that risk promoting a shallow, transmission-oriented model of education where teachers become content moderators, students become prompt engineers, and learning collapses into producing answers rather than developing understanding.

In this HAI seminar, HAI faculty fellow Hariharan Subramonyam offered a different vision which he calls learning by creating. Decades of learning sciences research show that understanding deepens when students actively engage in creative work such as modeling systems, designing solutions, and constructing artifacts that make thinking visible and open to reflection. Drawing on his work at the intersection of AI, design, and education, Prof. Subramonyam showed how human-centered AI systems can support learners as thinkers, creators, and problem-solvers by moving beyond shallow automation toward AI that supports learner agency in students.

This event was recorded on March 11, 2026 at Stanford University.

00:00:00 Introduction

00:01:32 Presentation

00:48:54 Q&A