This is a short summary of an essay I’ve just published, entitled Beyond text boxes: Exploring a network-based user interface for AI-supported learning.
The limitations of container-based learning
Courses, modules, and folders are the familiar structures we use to treat knowledge as if it exists in discrete, separate units. This approach, rooted in historical methods of organising information (based on books and libraries), fragments information that, in professional practice, needs to be deeply integrated.
Healthcare (and other) professionals must constantly synthesise information from various domains – anatomy, ethics, communication – to address complex situations that don’t fit neatly into curriculum silos. The container model makes it difficult to develop this essential integrative thinking, often leaving graduates feeling unprepared for the complexities of practice, despite success in compartmentalised assessments. The system tends to prioritise easily measurable knowledge within containers over the contextual, interconnected competencies required in the real world.
The advent of AI in education, primarily accessed through text-entry fields and chronological chat histories, hasn’t resolved this issue and, in many ways, reinforces it. Chat interfaces become linear containers, trapping information in sequences rather than networks. Retrieving related information involves scrolling or searching past interactions, mirroring the limitations of older systems. This interface model implicitly suggests that learning is a simple question-answer exchange, which misaligns with how experts build rich, interconnected mental models. While AI can make connections, the interfaces often obscure these relationships, increasing cognitive load as learners must mentally reconstruct the network. Improving text boxes alone doesn’t address this underlying structural problem. Furthermore, the ease of obtaining answers can create an illusion of understanding without fostering deeper integration or the crucial skill of formulating insightful questions.
Networks as a more authentic model
Networks are more effective and authentic organising principles for learning, particularly in professional contexts. Professional knowledge operates dynamically, with concepts gaining meaning through their relationships. Expertise involves navigating these complex webs, identifying patterns, and integrating information across traditional boundaries. Network models can better represent the contextual nature of knowledge and prepare students for the ambiguities of practice by making connections explicit during the learning process.
Envisioning a network-based learning interface
To support this network-based approach to learning interfaces, I want to explore a significant shift away from text-based systems towards interactive knowledge graphs. Some of the key features of such an interface would include:
- Visual navigation: Learners would navigate spatially (zooming, panning, focusing) through a visual knowledge map, allowing them to see the structure of their understanding develop as a network. Learning journeys become visible paths.
- Integrated AI: AI functions as both a conversational partner within the visual context (with dialogues anchored to specific knowledge nodes) and as a “network weaver,” dynamically suggesting and visualising potential connections emerging from discussions.
- Multimodal nodes: Nodes would contain diverse content types – text, video, simulations, 3D models – appropriate for representing the specific concept.
- Collaborative construction: Multiple learners could co-construct and navigate shared knowledge networks, fostering social learning and interprofessional understanding.
- Adaptive complexity: The interface could adjust the visible complexity of the network based on the learner’s stage, providing scaffolding for novices while revealing greater nuance for experts.
Conclusion: A necessary paradigm shift
Moving from container-based to network-based learning interfaces represents a fundamental shift. It reframes AI’s role from a mere answer provider to a facilitator of human connection-making. And it explicitly targets the development of integrative thinking skills central to professional expertise by making knowledge relationships visible and navigable. This approach holds the potential to bridge the persistent gap between education and practice, creating learning environments that extend more naturally into professional contexts.
As AI becomes increasingly integrated into education, we face a critical choice: reinforce outdated container models with new technology, or grab this opportunity to reimagine how we work with AI based on how knowledge truly functions. Exploring alternative user interfaces for AI-supported learning isn’t a question of aesthetics; it’s a fundamental question about how learning happens and how professional education needs to change.