Towards Considerate Embodied AI: Co-Designing Situated Multi-Site Healthcare Robots from Abstract Concepts to High-Fidelity Prototypes

Paper Title

Towards Considerate Embodied AI: Co-Designing Situated Multi-Site Healthcare Robots from Abstract Concepts to High-Fidelity Prototypes

Publication Info

• Topic area: Co-design of embodied AI systems for healthcare applications.
• Keywords: Embodied AI, healthcare robotics, co-design, multidisciplinary collaboration, iterative prototyping, non-value-added tasks, high-fidelity prototypes, stakeholder engagement, educational scaffolding, workflow integration.

Background and Problem

• Problem / challenge: Existing co-design efforts for healthcare robots often lack contextual depth, focus on single settings, remain at low-fidelity stages, and fail to integrate multidisciplinary collaboration effectively. These gaps hinder the development of robots tailored to diverse healthcare environments and real-world deployment.
• Significance: Addressing these limitations is critical to designing robots that reduce healthcare workers' (HCWs) burdens, improve patient care, and align with the unique workflows and constraints of different healthcare settings.
• Motivation and related work: Prior research has explored co-design in healthcare robotics, but often in isolated contexts or with limited stakeholder integration. This paper builds on these efforts by introducing a sustained, multi-context co-design process that progresses from abstract concepts to high-fidelity prototypes, enabling deeper insights into robot roles and deployment challenges.

Solution

• Proposed approach: A 14-week multidisciplinary co-design workshop series to develop embodied AI solutions for three healthcare settings: Emergency Departments (EDs), Sleep Disorder Clinics (SDCs), and Long-Term Rehabilitation (LTR) facilities.
• Novelty:
  1. Contextual task analysis across three distinct healthcare settings to identify non-value-added (NVA) tasks.
  2. Integration of multidisciplinary collaboration to shape robotic designs through shared understanding.
  3. Iterative prototyping from storyboards to high-fidelity prototypes, revealing design evolution and deployment challenges.
  4. Educational scaffolding to support non-technical participants in contributing to robot design.
• Procedure and key techniques:
  • Conducted need identification and workflow mapping through stakeholder panels and dollhouse activities.
  • Progressed through iterative prototyping stages: storyboards, cardboard prototypes, and full-scale prototypes.
  • Integrated educational sessions on robotics, design, and fabrication to support participant engagement.
  • Evaluated outcomes through artifact analysis, interviews, and post-educational session discussions.

Results

• Concrete findings:
  • Developed 19 storyboarded robot roles, 7 cardboard prototypes, and 3 full-scale prototypes tailored to ED, SDC, and LTR settings.
  • Identified four primary robot roles: delivery, storage, tour guide, and comfort/entertainment, with context-specific variations.
  • Speech, facial expressions, and touchscreen interfaces were the most common interaction modalities.
• Advantage over baselines:
  • Demonstrated cross-context learning by comparing shared and setting-specific robot roles.
  • Progression to high-fidelity prototypes revealed real-world constraints (e.g., spatial fit, ergonomic design) often overlooked in low-fidelity stages.
  • Multidisciplinary collaboration fostered richer, more grounded design outcomes.
• Experiments / evaluation:
  • Conducted artifact analysis across design stages.
  • Thematic analysis of interviews and discussions to evaluate collaboration, design evolution, and educational impact.
  • Participants included 22 individuals from diverse backgrounds (HCWs, artists, engineers, makers, patients).
• Limitations and future work:
  • Limited representation of HCWs due to high commitment requirements.
  • Constrained exploration of alternative robotic platforms.
  • Focused on Global North healthcare contexts, limiting generalizability.
  • Future work will expand to additional settings, platforms, and geographic regions, and explore multi-robot coordination.

Summary

This study introduced a 14-week co-design workshop series to develop embodied AI solutions for three healthcare settings (ED, SDC, LTR). By integrating multidisciplinary collaboration, iterative prototyping, and educational scaffolding, the study revealed how robots can address NVA tasks while aligning with facility-specific workflows. Key contributions include identifying shared and context-specific robot roles, fostering mutual learning among participants, and refining designs through progressive prototyping. The findings highlight the importance of grounding robot design in real-world constraints and propose eight guidelines for developing considerate embodied AI systems that are socially acceptable, trustworthy, and integrated into healthcare environments.