Responsibly building AI for the future of education

Google focuses on key outcomes for educators, students, and the educational ecosystem. It builds solutions using LearnLM capabilities, learning science principles, and responsible AI guardrails. Impact is measured in a variety of ways and time horizons such as observational studies, randomized controlled trials, ongoing user feedback, and expert evaluations.

• Encourage active learning: Helping learners actively construct understanding through pedagogical approaches like retrieval practice, dialogic inquiry, and problem-solving: tasks in which the learner makes connections, explains concepts, or creates artifacts.
• Manage cognitive load: Directing learner cognitive efforts on the focal learning goal while managing distractions unrelated to the learning goal; using well-structured segments and purposeful dual coding to support working memory and schema construction.
• Stimulate curiosity and motivation: Fostering perceived task value and expectancy of success by anchoring new concepts in prior knowledge, authentic real-world contexts, and learner goals.
• Adapt to the learner: Assessing what the learner knows and can do, providing calibrated challenges and fading scaffolds to build proficiency, and changing subsequent interactions to build on what the student knows while addressing areas of growth; helping the learner achieve their goals while offering the autonomy to explore.
• Deepen metacognition: Providing explicit opportunities for learners to set goals, plan strategies, monitor comprehension, and evaluate their learning process.

At Google, we are committed to measuring and understanding how our tools affect teaching and learning. We rigorously evaluate the impact and quality of our products, features and models using various methodologies and time horizons. Some examples include Randomized Control Trials, quasi-experimental studies, and observational studies.

We work alongside the people who know education best — leading experts, educators and institutions — to ensure our tools respond to the real needs of learners. Some of our collaborators include Columbia Teachers College, Arizona State University, Khan Academy, Fab AI, and Eedi.