Toward a “Deschooled” Society Through HCI: Envisioning the Future of Learning
AI-Guided Learning and Learner Autonomy
As part of my doctoral research, I’m exploring ways to enhance both time efficiency and learner autonomy in environments that use audio and video content. Central to this endeavor is the concept of AI-Guided Learning — an integrated framework that uses AI to optimize and support an entire learning environment, helping each learner access knowledge aligned with personal interests and goals, and design their own learning process.
Rather than focusing solely on the idea that people learn directly from AI, AI-Guided Learning emphasizes a broader design philosophy. AI actively adapts the learning environment to each individual’s needs, enabling them to access knowledge efficiently and flexibly, and to direct their own learning journey. One illustrative example is how professional shogi players in Japan already leverage AI software to explore new strategies and tactics. AI-Guided Learning seeks to expand such point solutions beyond specific domains — like shogi — to fields involving audio, video, and other diverse resources. By analyzing each learner’s understanding, preferences, and available content, AI can provide dynamic support throughout their learning process. Ultimately, this approach aims to free learners from the constraints of a single instructor or institution, bringing us closer to a self-directed learning model.
Deschooling in the Digital Age
When considering the growing autonomy and flexibility of modern learning environments made possible by AI-Guided Learning, it’s helpful to revisit the ideas of Ivan Illich. In Deschooling Society, Illich envisioned a world where individuals could pursue knowledge based on their own interests, unbound by school systems. Half a century ago, such a vision must have seemed purely idealistic, especially in an era without widespread personal computing or the internet. Yet today, thanks to rapid developments in information technology, elements of Illich’s deschooled environment are gradually taking shape in digital spaces. By returning to this intellectual foundation, we can situate HCI research within Illich’s philosophical framework, reconsidering how our work might add new value to his notion of a society unencumbered by institutional constraints.
A Learner-Centric Society and HCI
Ivan Illich criticized schools for their uniform curricula and credential systems, which channel learning in a strictly top-down manner. Instead, he envisioned a network for learning where individuals can freely explore and exchange knowledge rooted in everyday life, guided by their own interests and passions. This idea has been partially realized in learner-centered alternatives like unschooling and free schools, where teachers serve not just as knowledge providers but also as mentors, coaches, and navigators. At the core of this approach is the belief that learners themselves have the freedom to chart their own learning paths.
The notion of learner-designed learning resonates with the Human-Computer Interaction (HCI) principle of User-Centered Design, and becomes even more explicit in Learner-Centered Design (Soloway et al., 1994). Under this framework, system and software development is redefined around the question of how best to foster a learner’s growth and exploration, often by providing scaffolding — graduated levels of support — as needed. Illich’s vision of learning freed from institutional constraints strongly aligns with these learner-centric design philosophies, and modern HCI research and educational practices are already translating parts of that vision into tangible digital solutions.
Implementing Self-Directed Learning Environments
In recent years, HCI research has moved beyond theoretical discussions of learner autonomy and toward the practical design of systems and interactions. At conferences like UIST and CHI, we see numerous interactive dashboards that let learners choose their own pace or materials, as well as adaptive learning systems that dynamically adjust learning paths based on individual preferences and levels of understanding.
These efforts aim to give learners an environment where they can actively customize how they acquire knowledge. For example, Wambsganss and colleagues (2020) presented an adaptive system using natural language processing (NLP) to offer real-time feedback on a learner’s writing style and argument structure. Park and collaborators (2024) built a dialog-based tutoring framework that integrates student modeling and large language models (LLMs), tailoring guidance strategies to each learner’s unique characteristics. Such research demonstrates new possibilities for learners to reflect on their thought processes and adjust tutoring styles on the fly.
Additional technologies have emerged, such as personalized recommendation systems that suggest learning materials based on user logs, automated slide-generation tools for specific goals, and meta-cognitive support tools that visualize individual skill development. These innovations let learners access relevant knowledge resources according to their interests, and design their learning processes proactively — much like the deschooled network Illich once envisioned.
My Doctoral Work: AI-Guided Learning
Under the umbrella of AI-Guided Learning, my goal is to build an environment where learners can efficiently and proactively extract essential information from the vast amount of audio and video content available online. In Illich’s vision of a learning society, everyone could freely access knowledge. While MOOCs and OERs have significantly lowered the barrier to accessing diverse resources today, learners still face a new challenge: effectively filtering massive content and harnessing it within limited time.
To address this, my research proposes three technical approaches that enhance both time efficiency and learner autonomy — central pillars of an AI-Guided Learning framework where users can design and reconfigure their own learning paths as needed.
AIxSpeed
AIxSpeed focuses on audio content like podcasts or lecture recordings, dynamically adjusting playback speed so users can listen faster without losing comprehension. By analyzing speech via automatic speech recognition, the system slows down difficult segments or those prone to misrecognition, enabling listeners to quickly digest extended content. This boosts time efficiency and transforms online audio resources into a flexible source for on-the-go study.
FastPerson
FastPerson addresses video content, creating summary videos by extracting key points from long recordings. By integrating large language models, OCR, object detection, and other methods, it analyzes slides, diagrams, and spoken words to produce concise summaries. This approach saves time while letting learners choose which sections to explore in detail, strengthening both efficiency and autonomy. It also makes large libraries of recorded lectures more accessible, offering digest versions tailored to each learner’s level.
Profy
Profy analyzes user behavior, audio, and video data to highlight which parts differ from a skilled performance. This system extends beyond language learning — where you can compare your pronunciation with a native speaker — to fields like music or sports, comparing a learner’s actions to an expert’s. By leveraging self-supervised models, Profy operates without relying on a specific teacher or curriculum, supporting learners’ improvement through a self-driven feedback loop. It also introduces a new style of learning in which expert videos across the web can be directly integrated as study materia.
These three approaches convert diverse online media into open, navigable learning environments that reduce dependence on any one teacher or institution. Through AI-Guided Learning, learners can freely select content that matches their interests, goals, and learning styles, and construct an effective path of study. In this sense, it offers a glimpse of how deschooling principles might be reimagined in the digital era.
A Learner-Centric Ecosystem and Future Outlook
Technological advances — ranging from generative AI to AR/VR and IoT — are rapidly diversifying and extending today’s learning environments. Within these expansions, HCI research plays a crucial role, not just by introducing new technology, but by embracing human-centered design that foregrounds learner autonomy.
Here, Illich’s call to deschool serves as a valuable frame of reference. His vision of freeing education from rigid institutions aligns closely with the empowerment of learners to craft their own paths and access knowledge when and how they want. While current tech solutions only partially fulfill Illich’s broader agenda for societal transformation, they demonstrate promising steps toward a more learner-directed ecosystem.
My AI-Guided Learning initiative aims to propel this vision forward by integrating techniques for faster content understanding, autonomous material recommendation, and meta-cognitive support. Taken together, these innovations can reduce dependence on conventional schooling, paving the way for more open, self-directed learning environments.
Continuing to refine this learner-centric ecosystem will require collaboration across educational policy, instructional design, media literacy, and the humanities and social sciences. HCI researchers must go beyond technical breakthroughs to also address institutional and community-building aspects. Through such multifaceted approaches, we can reevaluate deschooling principles for the digital age — and continue to explore new forms of learning communities and ecosystems.
References
- Illich, I. (1971). Deschooling Society. New York: Harper & Row.
- Kawamura, K., & Rekimoto, J. (2022). DDSupport: Language Learning Support System that Displays Differences and Distances from Model Speech. 2022 IEEE International Conference on Machine Learning and Applications (ICMLA), 313–320.
- Kawamura, K., & Rekimoto, J. (2023). AIxSpeed: Playback Speed Optimization Using Listening Comprehension of Speech Recognition Models. Augmented Humans International Conference (AHs), 200–208.
- Kawamura, K., & Rekimoto, J. (2024). FastPerson: Enhancing Video Learning through Effective Video Summarization that Preserves Linguistic and Visual Contexts. Augmented Humans International Conference (AHs), 205–216.
- Soloway, E., Guzdial, M., & Hay, K. E. (1994). Learner-centered design: The challenge for HCI in the 21st century. interactions, 1(2), 36–48.
- Park, M., Kim, S., Lee, S., Kwon, S., & Kim, K. (2024). Empowering Personalized Learning through a Conversation-based Tutoring System with Student Modeling. In Extended Abstracts of the CHI Conference on Human Factors in Computing Systems, 1–10.
- Wambsganss, T., Niklaus, C., Cetto, M., Söllner, M., Handschuh, S., & Leimeister, J. M. (2020). AL: An Adaptive Learning Support System for Argumentation Skills. In Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems, 1–14.
About the Author
Kazuki Kawamura
Completed a Master’s in Informatics at Kyoto University’s Graduate School of Informatics in 2021. Currently a PhD student at the University of Tokyo’s Graduate School of Interdisciplinary Information Studies and a researcher at Sony Computer Science Laboratories, working on AI-Guided Learning from a Human-AI Interaction perspective.
Connect with Kazuki
His interests include leveraging deep insights into user behavior and effective interventions to enhance overall well-being and skill development at a societal level, integrating theories from the humanities and social sciences. He also contributes to AI education at multiple institutions, aiming to connect foundational research, real-world applications, and pedagogical outreach in a seamless way.
