A Community-Centered Design Framework for Robot-Assisted Feeding Systems
Tapomayukh Bhattacharjee, Maria E. Cabrera, Anat Caspi, Maya Cakmak, Siddhartha S. Srinivasa · 2019 · Proceedings of the 21st International ACM SIGACCESS Conference on Computers and Accessibility (ASSETS) · doi:10.1145/3308561.3353803
Summary
This paper develops a community-centered design framework for evaluating robot-assisted feeding (RAF) systems, shifting the perspective from purely transactional (robot picks up food, delivers to mouth) to relational (feeding as a social, cultural, and identity-laden activity embedded in community contexts). The authors argue that existing RAF development focuses narrowly on the robot-user dyad, ignoring the broader social network that the technology impacts. Drawing on the Social-Ecological Model (SEM), the framework analyzes RAF systems across three nested social levels: intrapersonal (L1 — the care recipient and their caregiver), interpersonal (L2 — family members and visitors), and community/institutional (L3 — the care facility and its staff). At each level, six evaluation contexts are assessed: technical function (food manipulation, bite acquisition/transfer, safety), technological robustness (collision avoidance, anomaly detection, visual servoing), information gaps (multimodal communication, user interface suitability), usability (bite timing, size, sequence adaptability, customization), user empowerment and social acceptance (dignity, independence, self-esteem, perceived usefulness), and system integration (compatibility with wheelchairs, other assistive devices, alarm systems). The framework was informed by a multi-level contextual inquiry: observations and semi-structured interviews with five care recipients and five caregivers in an assisted-living community during actual meals, plus surveys with 15 domain experts (occupational therapists, physical therapists, feeding specialists, speech-language pathologists, assistive technology professionals).
Key findings
The contextual inquiry revealed that feeding has profound emotional, social, and identity dimensions beyond nutrition. Care recipients reported feeling embarrassed asking others for help eating, and one expert described it as "a degree of infantilization... it's such a personal thing to have someone shoveling food in your mouth." The greatest perceived benefit of RAF was increased independence and self-esteem, not efficiency. Experts emphasized that meals are social occasions — "when we feed, we talk, we laugh, we ask how they are doing" — and warned that a fully autonomous robot could increase social isolation by removing the caregiver interaction. Key design indicators emerged: bite size tuning (care recipients have specific preferences and swallowing needs), pacing (allowing natural eating pace rather than rushing), bite sequence choice (maintaining agency over what to eat next), multiple input modalities (voice, gestures, head movements, gaze, AAC devices — since the target population has diverse motor abilities), safety features (emergency stop, choking detection, collision avoidance), and physical properties (small, quiet, non-attention-drawing). The framework was applied to evaluate three existing RAF systems (Obi — a commercial table-mounted device; PR2 — a general-purpose research robot; ADA — the authors' wheelchair-mounted robotic arm) across all indicators at all three social levels. The comparison revealed that no system adequately addresses user empowerment and social acceptance metrics, all lack bite timing and sequence adaptability, and system integration with other assistive devices is uniformly weak.
Relevance
This paper makes a fundamental contribution to how assistive robotics should be designed and evaluated. By applying the Social-Ecological Model to feeding robots, it demonstrates that assistive technology does not exist in isolation — it reshapes relationships, social dynamics, and self-identity across an entire care network. The framework reveals that a technically perfect feeding robot could still fail if it undermines the social aspects of mealtimes, strips agency from the care recipient, or creates new burdens for caregivers. For accessibility practitioners and assistive technology developers, the key insight is that success metrics for assistive devices must extend far beyond task completion. The community-centered approach — studying not just the user but their caregivers, family, and institutional context — provides a model applicable to any assistive technology that operates within caregiving relationships. The comparative evaluation of three RAF systems using the framework demonstrates its practical utility: it systematically identifies where each system excels and where it falls short across multiple social dimensions, providing a clear research roadmap. The finding that user empowerment and social acceptance remain the least-addressed design dimensions across all three systems suggests the assistive robotics field has been too focused on engineering challenges at the expense of human factors.
Tags: assistive robotics · activities of daily living · motor disability · feeding assistance · community-centered design · contextual inquiry · participatory design · caregiving · social-ecological model · assistive technology