Isabelle C. Hau is the Executive Director of the Stanford Accelerator for Learning. Susanna Loeb is a Professor at the Stanford Graduate School of Education, and Faculty Director at the Stanford Accelerator for Learning, leading the SCALE Initiative. This think piece outlines four complementary pathways that enable research-informed digital learning at scale, using the Stanford Accelerator for Learning as an illustrative model. It is included in this chapter to emphasize that true contextual readiness requires building ‘circular learning systems’ and investing in the human capacity needed to continuously adapt digital education innovations to local realities.
Introduction
Across the world, education systems face a profound paradox. Knowledge about how humans learn has advanced dramatically, yet learning outcomes remain deeply unequal and often stagnant. New findings in development, cognition, motivation, feedback, social belonging and system improvement have expanded what educators and researchers understand about learning.172 Even so, many classrooms, training programmes and digital platforms still reflect older assumptions about how learning happens and what educational systems are designed to support.
AI has intensified this challenge.173 It is changing how people access knowledge, solve problems and demonstrate understanding. It is entering homes, classrooms and workplaces faster than most institutions can evaluate its effects or guide its use. It is also changing who and what we learn with. These developments create real possibilities for more personalized, creative and responsive learning experiences. They also create serious risks. Systems that adopt AI without a clear account of learning, development and equity may deepen dependency, reward superficial performance or widen existing inequalities. The challenge extends beyond innovation to translation and diffusion, at a pace commensurate with technological change.
Research produces insight; systems require implementation.174 The distance between the two has become one of the central bottlenecks of educational progress. Without mechanisms that connect discovery to practice, knowledge accumulates without improving lived learning experiences. In the age of AI, this gap risks widening further. What is needed is stronger infrastructure for moving knowledge into the settings where learning happens. Research findings must travel across disciplines, sectors and communities if they are to shape everyday practice.
The Stanford Accelerator for Learning is one such translational infrastructure. Launched in 2021 as a Stanford University-wide initiative hosted at the Stanford Graduate School of Education, the Accelerator was created to bridge the persistent gap between advances in the learning sciences and their application in real-world settings. From the outset, it has drawn on expertise across education, computer science, neuroscience, design, business and public policy to address learning as a system-level challenge.
This think piece uses the Accelerator as an illustrative example of what it means to move ‘beyond digital as usual’: creating connective infrastructure that links research, practice, technology and policy. By bringing together researchers, schools, communities, industry, policymakers and global organizations, it demonstrates how sustainable educational transformation emerges from the continuous co-evolution of evidence, innovation, human capacity and public systems.
The Accelerator organizes its work across four interconnected pathways that advance digital learning grounded in the learning sciences and infused with research. These pathways aim to turn knowledge about learning into knowledge used for learning, and, more fundamentally, to build digital learning systems that learn themselves, transforming emerging evidence into better learning experiences while adapting to a world in which AI is reshaping how knowledge is created, accessed and applied. Each pathway is addressed in turn below.
Funding discoveries: Investing in research that shapes digital learning
Universities play a unique role in generating new knowledge, yet discovery alone rarely transforms digital learning.175 Too often, advances in the learning sciences remain confined to academic journals, disconnected from the digital tools, AI applications and classroom practices that shape learners’ everyday experiences. Building digital learning beyond digital as usual therefore begins by investing in discoveries with clear pathways towards implementation, ensuring that evidence informs the design of technologies.
The Rapid Online Assessment of Reading (ROAR), developed by Professor Jason Yeatman and colleagues, shows what this looks like in practice. Decades of research have demonstrated that early identification of reading difficulties is critical for improving literacy outcomes,176 yet traditional assessments are often time-intensive, expensive and difficult to administer at scale. Through interdisciplinary collaboration among researchers in education, neuroscience and medicine, the ROAR translated decades of reading science into a rapid, engaging digital assessment capable of reaching large numbers of learners.177 In California, the ROAR has been one of the state’s approved screening instruments for identifying students at risk of reading difficulties since 2024. Today, the ROAR has reached more than 1.2 million students. Its widespread implementation continues to generate new scientific insights into literacy development and dyslexia, creating a virtuous cycle in which digital learning both applies and advances the science of learning.
The Accelerator supports research discoveries such as the ROAR through thematic seed grants that bring together researchers, educators, technologists, designers, entrepreneurs and policymakers around shared educational challenges. But funding research is only the first step. Through the Accelerator Studio, multidisciplinary teams receive support to transform promising discoveries into research-informed digital learning solutions. The Studio helps projects move from proof of concept to real-world implementation by providing expertise in human-centred design, product development, implementation, evaluation, entrepreneurship, partnerships and scaling. In doing so, it ensures that digital tools are learning science-enabled, grounded in evidence from the outset and refined through practice. This translational approach reflects a growing recognition that educational innovation requires sustained collaboration among researchers, designers, entrepreneurs, educators and implementation partners throughout the innovation life cycle.
Impactful digital learning starts with a deep understanding of how people learn. Technology becomes transformative only when research is intentionally translated into digital experiences, AI-enabled tools and learning environments that can be tested, improved and scaled in partnership with educators and learners.
Building solutions: Co-designing research-informed digital learning
Evidence alone does not transform learning. Decades of research on knowledge mobilization, implementation science, improvement science and research–practice partnerships have demonstrated that educational improvement depends both on generating high-quality evidence and on creating sustained mechanisms that enable research and practice to inform one another.180 This challenge has become even more urgent in digital education, where technological innovation often outpaces the generation and application of evidence. Recent scholarship argues that effective digital learning requires moving beyond evaluating technology after deployment towards embedding evidence throughout the entire innovation process, from design and development to implementation, evaluation and continuous improvement.
Building research-informed digital learning therefore requires moving beyond a linear model in which researchers generate knowledge and practitioners simply adopt it. Instead, evidence and innovation evolve together through continuous collaboration among researchers, educators, learners, designers, entrepreneurs, technology developers and policymakers. Digital learning environments become living laboratories that generate new evidence about teaching, learning and human–AI interaction.
The Accelerator has applied the same approach to AI-enabled instructional coaching, led by Professor Dorottya Demszky. Decades of research have demonstrated that instructional coaching is among the most effective strategies for improving teaching practice,183 yet access remains limited because expert coaching is costly and difficult to scale. The project set out to expand access to evidence-based coaching with AI, without displacing teachers’ professional judgement. From the earliest stages of development, learning scientists, computer scientists, instructional experts, designers and classroom teachers worked together to translate research on effective teaching into an AI-supported coaching platform.184 Teachers participated as co-designers throughout the development process, ensuring that both the technology and the pedagogical model reflected authentic classroom practice. Each implementation generated new evidence that informed subsequent iterations, creating a continuous cycle of research, design, implementation and improvement. Early evidence suggests that generative AI can strengthen teacher preparation and lesson planning when designed to complement, not replace, professional judgement.185
Successful digital innovation depends on ecosystems where research, design, implementation and ongoing learning reinforce one another. In the AI era, the most impactful digital learning solutions will be learning science-informed, educator-designed, evidence-generating and continuously improving.
Mobilizing knowledge: Translating discovery into decisions
Research creates public value only when it informs decisions.186 Yet educators and policymakers often face fragmented evidence, competing priorities and rapidly evolving technologies. Even when robust evidence exists, translating it into practical guidance remains one of the greatest bottlenecks in educational improvement.187 This challenge is especially pronounced in the context of AI, where the rapid pace of innovation and adoption often exceeds the speed at which rigorous evidence can be generated, synthesized and translated into policy and practice. As demand for trustworthy guidance grows, there is an urgent need for infrastructure that can curate, evaluate and synthesize the rapidly expanding body of AI-in-education research.
The AI Hub for Education’s Research Repository at Stanford, part of the SCALE Initiative at the Accelerator, curates and synthesizes the rapidly growing evidence base on AI in education.188 The Repository organizes emerging research into an accessible, continuously updated public resource. By making evidence easier to discover, interpret and apply, it helps decision makers distinguish between promising innovations and unsupported claims, accelerating the responsible adoption of AI in education while identifying important gaps for future research. In a field evolving as rapidly as AI, mobilizing knowledge means generating new evidence and ensuring it is accessible, actionable and available when decisions are made.
Developing talent: Building human capacity for research-informed digital learning
Digital transformation succeeds because people, not technologies, learn, adapt and innovate.189 Even the strongest evidence and the most promising AI-enabled tools cannot improve learning without educators, researchers, entrepreneurs and leaders capable of translating research into meaningful educational practice. Moving beyond digital as usual therefore requires investing both in digital infrastructure and in the human capacity to design, implement, evaluate and improve research-informed digital learning.
Teachers remain central to this transformation190 as co-designers.191 As AI becomes increasingly integrated into education, educators need opportunities to critically evaluate emerging technologies, understand their pedagogical implications and integrate them thoughtfully into teaching and learning. The Accelerator’s AI Tinkery provides hands-on professional learning experiences where educators can explore, build and test AI-enabled tools while grounding their work in the science of learning. It empowers them to become informed designers and critical users of AI in education. Research increasingly suggests that one of AI’s contributions to education may come by augmenting teachers’ expertise.192
Equally important is cultivating the next generation of innovators. Through the Education Entrepreneurship Hub, the Accelerator supports students and early-stage founders in transforming research into scalable digital learning innovations that respond to real educational needs, driving democratization in tool building.
These investments build more than individual expertise; they cultivate ecosystems in which educators, researchers, entrepreneurs, technology developers, policymakers and learners learn from one another. Human relationships, interdisciplinary collaboration and shared evidence become the infrastructure that enables digital learning to evolve responsibly over time.
Overall, sustainable digital transformation depends both on new technologies and developing the people, partnerships and institutions capable of ensuring that digital learning remains grounded in research, responsive to educators and learners, and improved through evidence and practice.
Towards circular digital learning systems
These four pathways illustrate a shift from digital as usual, where technologies are developed, deployed and periodically updated, to digital learning systems that learn and improve. Research, technology, implementation and capacity-building become part of a continuous cycle of evidence generation, adaptation and innovation.
Discoveries in the learning sciences generate new knowledge about how people learn. That knowledge informs the design of digital tools, AI-enabled learning experiences and educational policies. As these innovations are implemented in classrooms and communities, they generate new evidence about what works, for whom and under what conditions. Those insights, in turn, inform the next generation of research, technologies and professional learning. In this model, implementation is not the end of innovation; it is the beginning of the next cycle of learning.
This represents an important shift in how educational systems think about digital transformation.193 Too often, scaling is understood as replicating successful technologies across larger populations. Yet effective digital learning must regularly adapt to learners’ needs, educators’ expertise, local contexts, emerging evidence and the rapid evolution of AI itself.
Circular digital learning systems embrace this reality.194 They create continuous feedback loops that connect researchers, educators, learners, technology developers, policymakers and communities. AI can accelerate this cycle by enabling faster synthesis of evidence, more responsive personalization and richer implementation data, while the learning sciences ensure that technological innovation remains grounded in how people actually learn and develop.
The goals are both to spread digital innovations and to build learning systems that improve themselves, where research informs technology, technology strengthens teaching and learning, implementation generates new knowledge, and every cycle brings digital learning closer to being more effective, equitable and human-centred. This is what it means to move beyond digital as usual.
Lessons for building digital learning beyond digital as usual
The Stanford Accelerator for Learning shows one way to build this kind of translational infrastructure. Five lessons emerge:
- Invest in translational infrastructure. Scientific discoveries create value when institutions and platforms exist to translate learning science into digital learning experiences, practical guidance and policy.
- Build interdisciplinary ecosystems from the outset. The ROAR is a case in point: it took expertise in education, neuroscience and medicine, not any one field alone, to turn reading science into a usable assessment.
- Co-design digital learning with educators and learners. Teachers, students, families and communities are essential partners in shaping digital tools and AI applications that are relevant, trusted and responsive to local contexts.195
- Invest in human capacity as much as digital capacity. Successful digital transformation depends both on robust technologies and on educators, leaders, researchers, entrepreneurs and policymakers who can critically evaluate, adapt and improve research-informed digital learning.
- Build digital learning systems that continuously learn. The most effective digital ecosystems generate evidence through implementation, use AI and data responsibly to inform improvement, and create feedback loops that connect research, practice and innovation. They enable continuous adaptation as technologies, evidence and learner needs evolve.
These principles extend beyond any single institution. They offer a framework for universities, governments, philanthropies, industry and international organizations seeking to build digital learning ecosystems that are grounded in the learning sciences, informed by research and designed to evolve alongside advances in AI. The Accelerator’s model assumes resources, including in-house AI, design and neuroscience expertise, that most ministries of education and lower-resourced school systems do not have. These same functions can still be built at a smaller scale, through a research–practice–partnership unit, a regional intermediary organization or an existing government implementation research body.197 What a system needs is some structure that connects evidence to practice and feeds implementation back into further discovery, sized to what that system can actually sustain. Moving beyond digital as usual means moving beyond technology adoption towards systems that translate evidence into better learning for every learner.
Conclusion
This piece began with a paradox: we know more than ever about how people learn, yet outcomes have not improved. AI does not close that gap on its own. Technology alone will not transform education. The real opportunity lies in creating digital learning ecosystems that are grounded in evidence, designed around how people learn, and improved through research, implementation and responsible innovation.
The Stanford Accelerator for Learning is a single institutional example. The principles behind it are transferable across countries and education systems seeking to move beyond digital as usual. Its four interconnected pathways – funding discoveries, mobilizing knowledge, building solutions and developing talent – illustrate how learning science, digital innovation, implementation and human capacity can reinforce one another in a continuous cycle of improvement. Research, technology and practice as separate endeavours become part of a shared system for advancing learning.
These examples are also still at an early stage. Most of the evidence so far documents teacher practice, tool adoption or reach, rather than long-term causal effects on student learning, and the four pathways described here have so far operated within a single, well-resourced research university rather than as an integrated system tested elsewhere. Whether this approach improves student outcomes at scale, and does so outside a context with Stanford’s resources, remains an open empirical question.
No single institution or country has discovered the definitive model for the future of digital learning. Every context will require its own partnerships, priorities, technologies and pathways. Schools and systems need adaptive learning ecosystems that connect discovery with practice, evidence with innovation, technology with human development, and local implementation with global learning.
In the age of AI, the greatest opportunity is to build digital learning systems that learn – systems that evolve as rapidly as the science of learning, advances in AI and the needs of learners themselves. When research informs practice, practice generates new knowledge, and technology is intentionally shaped by evidence and human values, digital learning can move beyond delivering content to becoming a powerful engine for lifelong learning, equity and human flourishing.
