Towards Fluent Interaction with Cyber-Physical Architecture
New research explores walls that move and rooms that reshape themselves based on evolving user intent.
A team from Carnegie Mellon University's Human-Computer Interaction Institute and the University of Washington has published groundbreaking research on the future of interactive, robotic architecture. Their paper, 'Towards Fluent Interaction with Cyber-Physical Architecture,' investigates the design principles for environments where walls, furniture, and structures can physically reconfigure themselves. The work is grounded in two user studies involving a total of 32 participants, moving from speculative visions to practical interaction challenges.
The first study, a series of design workshops with 20 people, uncovered aspirational user desires but also exposed critical tensions. Participants grappled with the balance between proactive automation—where the space anticipates needs—and the preservation of personal control. A second tension emerged between highly personalized environments and the need for shared, public ownership of adaptive spaces.
A follow-up, task-based 'Wizard-of-Oz' elicitation study with 12 participants then grounded these visions in reality, revealing core interaction challenges. A key finding is the necessity for a 'modality-agnostic model of evolving user intent.' This means the AI system must understand a user's changing goals over time, regardless of whether they communicate via voice, gesture, or implicit behavior, to enable fluent collaboration.
Based on these findings, the researchers conclude with a set of grounded proposals for creating robotic environments that act as trusted partners. The work, accepted to the prestigious 2026 CHI Conference, lays a crucial foundation for a future where our built environment is a dynamic, responsive collaborator in daily life.
- The research is based on two user studies with a total of 32 participants, combining speculative workshops and practical task-based evaluation.
- A core proposed technical solution is a 'modality-agnostic model of evolving user intent' for the AI to understand changing goals across different forms of input.
- The work identifies major design tensions, particularly between proactive automation and user autonomy, which must be resolved for such systems to be trusted.
Why It Matters
This research defines the human-AI interaction blueprint for the next generation of smart homes, offices, and adaptive public spaces.