Nonverbal Human–Robot Interaction in Companion Robots for Older Adults: Scoping Review of Interaction Contexts, Behaviours and Outcomes

As the global population ages, companion robots have emerged as a pivotal technology in elderly care and health support. They are tasked with providing emotional support, social companionship, and health management, with applications spanning home environments, communities, nursing homes, and rehabilitation centers . As quintessential socio-technical systems, their acceptability and sustained user engagement fundamentally depend on their efficacy in conveying social cues, particularly the quality of non-verbal interaction (NVI). In interaction design for older adults, non-verbal behaviours occupy an irreplaceable position. Compared to cognitively demanding verbal communication, non-verbal modalities provide more intuitive, lower-cost perceptual input, which is crucial for seniors with hearing impairments, language barriers, or cognitive decline. Companion robots construct social presence through dimensions such as facial expressions, gaze, dynamic haptics, proxemics, and paralinguistic cues. These behaviours not only drive users' immediate emotional responses but also form the foundation for fostering long-term trust and attachment. Despite a surge in empirical research, existing reviews remain largely confined to macro-level evaluations of intervention efficacy or ethical discourse . Consequently, the interactive mechanisms within the "context-behaviour-outcome" chain lack systematic integration. Significant heterogeneity in defining behaviours and assessment criteria across studies further impedes theoretical synthesis and the formation of design consensus. Therefore, this scoping review aims to systematically synthesize and deconstruct the evidence on non-verbal interaction in companion robots for the elderly. By elucidating its context-dependencies and behavioural patterns, this review seeks to provide a robust evidence-based foundation for developing more effective and elderly-centred robotic interaction paradigms.