Co-STEAM: Redesigning AI’s Role for STEAM
About Jiarong Yu’s Research:
STEAM education, combining Science, Technology, Engineering, the Arts, and Mathematics, has become a vital approach to preparing learners for a world that increasingly calls for creativity, adaptability, and collaboration across disciplines. It values learning that is collaborative, embodied, and experiential. Yet as AI becomes more common in various learning contexts, it’s often designed to instruct or solve problems for people, a role that can sit uneasily with these collaborative values between learners and technology within education.
Jiarong’s PhD asks a different question: what if AI didn’t try to teach or solve, but shaped the conditions for people to think and create together, without replacing the human collaborators and human agency at the heart of learning? Through practice-based design research, she is developing Co-STEAM, an experimental AI system that works quietly in the background of group learning activities, supporting collaboration without taking over.
Co-STEAM has been tested and refined across several real-world settings, from university classrooms to public community events and exhibitions. This research contributes to how we think about AI’s role in human learning, education, and creative collaboration, and forms the centre of her doctoral research.
About Jiarong Yu:
Jiarong Yu is a PhD researcher in Interaction Design at the School of Design, University of Edinburgh, affiliated with the Edinburgh Futures Institute. Her research sits at the intersection of Human-AI Interaction, STEAM education, and applied AI system design, with a practice-based focus on participatory and embodied approaches to design. Her creative and research work has been supported by Immersive Arts UK, Innovate UK’s Immersive Tech Network, and UKRI CoSTAR, and exhibited at venues across Edinburgh, Glasgow, London, Cardiff, Shanghai, and Tokyo, including National Museum of Scotland.
Figure 1: Jiarong Yu, PhD researcher in Design (Interaction Design), University of Edinburgh. Credit: Photograph by Chris Scott.
STEAM education, combining Science, Technology, Engineering, the Arts, and Mathematics, has become a vital transdisciplinary approach to preparing learners for a world that increasingly calls for creativity, adaptability, and collaboration across disciplines. At its heart, STEAM values learning that is collaborative, embodied, and experiential, positioning technologies and materials as partners in the learning process rather than instructional tools. Learning, in this view, is not simply the transfer of knowledge from one person to another, or from one system to the people, but something that emerges through shared making, conversation, and the back-and-forth of ideas between people.
Yet as AI becomes more common in learning contexts, it is often designed to instruct or solve problems for people, stepping into the role of tutor, assistant, problem-solver, or answer-generator. This is a role that can sit uneasily with the collaborative values at the heart of STEAM, where the relationships between learners, and the process of learning itself, often matter as much as the outcomes they produce. When AI takes over the task of generating answers, something of that shared process risks being lost.
Jiarong Yu’s PhD research asks a different question: what if AI didn’t try to instruct or solve, but instead meaningfully shaped the conditions for people to think and create together, without replacing the human collaborators and human agency at the heart of learning?
Through practice-based design research, she designed Co-STEAM, an experimental AI system that works quietly in the background of group learning activities. Rather than directing a conversation or offering answers, Co-STEAM listens to the shape of a group’s discussion and gently reflects fragments of it back, through generative imagery and emerging words that surface nearby. No one is asked to respond to what appears. It simply exists alongside the group, a quiet infrastructure for their ideation rather than a voice competing for attention within it.
Figure 2: A sample from Co-STEAM’s generative visual stream during a collaborative ideation session.
The design of Co-STEAM has moved through several stages of thinking about what role AI should play in this kind of space: from a tool learners might pick up and use, to a collaborator that takes part more actively, and now toward something quieter still, a presence that holds space for collaboration. It asks not only what AI can do in a learning space, but what kind of presence it should have within it. Just as importantly, it asks what we don’t want AI to do in this learning context.
Co-STEAM has been tested and refined across several real-world settings: in university classroom contexts within University of Edinburgh’s Design Informatics programme, at public community events including creative technology showcases hosted by the Royal Conservatoire of Scotland. Each setting has offered a different lens on how people respond to Co-STEAM working alongside them. These observations, gathered through hands-on practice and reflection rather than formal testing alone, are now feeding into the next stage of Jiarong’s PhD research.
Figure 3: Co-STEAM in use during a university tutorial session within University of Edinburgh’s Design Informatics programme
An earlier phase of this research was presented and published at the British Computer Society’s Human-Computer Interaction Conference (BCS HCI 2025), under the title “Human–Human–AI Co-design for Person-Centred STEAM Futures.” Since then, the research has continued to develop through presentations at events including the Design Research Society’s 2026 conference in Edinburgh, where Jiarong co-led a PhD hub session exploring AI in practice-based research more broadly. Co-STEAM’s core module has also been applied in I&AI: Mirror, an AI-enhanced immersive installation and performance that runs as a parallel creative practice alongside Jiarong’s PhD research, exhibited in Edinburgh with support from Immersive Arts UK.
Looking ahead, this research will continue to develop over the coming months, with further findings to be shared in due course.
CONNECT WITH Jiarong Yu
See more about this project via Instagram