Google Research has published results from an experiment that uses generative user interfaces (GenUI) to produce custom, interactive educational simulations. The initiative aims to make digital learning less passive by enabling multimodal, inquiry‑based practice that students can use to experiment, test hypotheses and solve problems.
What the project delivers
The research demonstrates a system that generates interactive, guided learning experiences tailored to a teacher’s objectives and curriculum. Google also released a sample library of more than 30 AI‑generated learning interactives in English for STEM subjects—physics, chemistry, biology and math—targeted primarily at middle and high school. All items in the sample library were produced by AI and reviewed by teachers.
Pedagogical rationale and background
The project centers on active learning, a principle supported by classical education theorists and modern cognitive research (e.g., John Dewey, Jean Piaget, and the ICAP framework): students learn more effectively when they engage directly in problem solving and experimentation. The work builds on Google’s prior education models and experiments, including LearnLM (a family of generative AI models fine‑tuned for education released in 2024) and the 2025 Learn Your Way experiment.
How the system is designed for classrooms
Key elements of the generation pipeline include:
- Teacher‑led learning objectives: educators propose topics; the system produces precise, editable learning objectives that the teacher must approve and that drive the generation process.
- Structured game levels: complex concepts are decomposed into a sequence of progressively harder levels so learners can practice deliberately and build competence over time.
- AI‑generated scaffolding: each interactive includes an introduction to prime prior knowledge, a toolbox of relevant formulas and theories, multi‑level hints, tailored feedback explaining why responses work or fail, and worked solutions for post‑exploration consolidation.
Generation is iterative and governed by pedagogical guardrails. The pipeline includes self‑correcting loops and automatic evaluations that check pedagogy (e.g., whether levels align with objectives and increase in difficulty), mechanics (e.g., are controls functional and is the level solvable?), and visual clarity (e.g., are there distracting duplicate elements?). Some automated checks are agentic: they open a Chrome instance and interact with the simulation to assess solvability and even attempt adversarial actions such as driving controls to extreme values. The loop repeats until all quality criteria are met.
Teacher vetting and early feedback
A central principle of the research is keeping teachers at the center. All interactives available today were vetted and approved by teachers. Sample topics include curricular items such as Kepler’s Laws of Planetary Motion, Data Visualization and Projectile Motion.
A collection of learning interactives was evaluated by STEM teachers in the UK, with overall ratings of good or excellent; physics and chemistry were identified as most amenable to simulation creation. In an initial U.S. study with 12 teachers, each teacher requested three custom interactives tailored to their classrooms. The average teacher rating for interactive quality was 8 out of 10. Teachers highlighted that dynamic generation helps differentiate instruction in ways static, off‑the‑shelf simulations cannot. They also praised design alignment with instructional goals and the multi‑tiered scaffolding that mirrors one‑on‑one teacher support.
Next steps and pilot program
Google will expand the library in collaboration with classroom teachers. Working with Google for Education, the team will pilot learning interactives in schools worldwide. Schools using Google Workspace for Education can sign up through the Google for Education Pilot Program so teachers can request custom STEM simulations aligned to curriculum, learning goals and grade level. Generated interactives go to the requesting teacher for review; only after teacher validation can they be added to the public library.
The project team will also conduct UX research and field studies to measure learning gains and student engagement when learning interactives are used in classrooms.
Acknowledgements
The announcement thanks contributors including Alex Moy, Alisa Kovshov, Anisha Choudhury, Anna Iurchenko, Ayça Cakmakli, Ayelet Shasha Evron, Brit Mennuti, Diana Akrong, Femi Olanubi, Ian Li, Ido Lerer, Julia Wilkowski, Lidan Hackmon, Michal Gordon, Nir Kerem, Preeti Singh, Rena Levitt, Rotem Yulzary, Sarah Smith, Shlomi Ben Shimon, Sophie Allweis, Tracey Lee‑Joe, Tzvika Stein, Yaniv Carmel, Yishay Mor and Yuri Lev. Executive champions named include Niv Efron, Avinatan Hassidim, Maureen Heymans, Amy Keeling, Katherine Chou, Ronit Levavi Morad, Yossi Matias, Chris Phillips and Ben Gomes.
By adapting generative technologies specifically for educational use, Google aims to make practice more active, effective and tailored, while continuing to refine the system in partnership with teachers.
