Google Research is testing an artificial intelligence system that lets teachers design custom science and math games tailored to their lessons. The experiment, known as Learning Interactives, uses "generative user interface" technology to create step-by-step challenges instead of simple text or images. This system helps educators quickly generate learning tools without needing advanced programming skills.
What Happened
According to the Google Research Blog, the system builds multi-level simulations featuring math and science challenges, hints, and feedback. A public library of more than 30 AI-generated simulations is already open to the public, covering physics, biology, and chemistry for middle and high school students. To prevent the AI from generating incorrect math or science concepts, commonly called AI "hallucinations," Google built a specialized GenUI framework that prioritizes accuracy. Before a simulation is built, teachers must set and approve the educational goals. Google also uses automated testing loops to check the games. For example, the system launches a browser to interact with the simulation like a student would, which catches bugs and checks logic before a teacher reviews it. Teachers in the UK and US gave the tool positive marks in initial tests, with an average of 8 out of 10 for quality. Google is now expanding testing through its Google Workspace for Education Pilot Program.
The Bigger Picture
This tool builds on Google's "LearnLM" technology, which is designed to align with learning science principles by adjusting to a student's level and providing hints instead of direct answers. Research shows that interactive simulations can boost student comprehension. A meta-analysis published in the AsTEN Journal of Teacher Education found that interactive physics simulations had a large positive effect on student learning, particularly when paired with inquiry-based teaching. A study in the Jurnal Inovasi Fisika dan Edukasi showed that students using digital simulations improved their understanding scores compared to those in traditional classrooms. Still, some academic researchers warn against relying too heavily on automated tools. A paper in Communications Psychology points out a persistent gap between AI engineering and real-world child development, arguing that generative AI can prioritize efficiency over collaborative learning. As we previously reported on Bangladesh's partnership with UNICEF, schools worldwide use Google tools to scale instruction, but finding tools that actively engage students remains a challenge.
What This Means for Families
For parents and educators, this technology shows how homework and classroom exercises can be customized. Instead of a student getting stuck on a generic worksheet, the system adjusts to offer specific hints and formulas based on incorrect answers. Because this is an early experiment, families should not expect these games to replace traditional lessons yet. Teachers must still review the content to make sure the AI does not generate math or science errors. While Google’s automated testing caught many bugs, studies on software testing show that human oversight is still vital to ensure safety and accuracy.
What You Can Do
You can try out the 30+ free English simulations in biology, math, and chemistry to see how your child responds to the digital practice. If your school uses Google Workspace for Education, you can ask administrators if they plan to join the Google for Education Pilot Program to test these simulation tools. When helping children with homework, you can also adopt this teaching method by offering hints and asking guiding questions instead of giving them answers immediately.