Abstract: Companion technologies are increasingly used to relieve loneliness in later life, but the field lacks a unified framework for evaluating their effectiveness and pays insufficient attention to differences in older adults' social support contexts. This article defines affective companionship as the process of meeting relational, emotional, and existential needs through sustained, psychologically meaningful interaction. On this basis, we propose a three-dimensional (relational, emotional, existential) by three-level (surface, intermediate, deep) evaluation framework that distinguishes what outcome indicators such as reduced loneliness and increased satisfaction actually measure at different levels of effectiveness. A systematic review of studies published since 2016 shows that existing evidence concentrates on surface-level effects; deep-level outcomes such as a sense of connection, feeling understood, and a sense of mattering are rarely measured directly. Effect sizes also vary considerably across studies, and this heterogeneity can be largely explained by older adults' social support contexts. Accordingly, we propose a contextualized effectiveness theory, which holds that the effectiveness of companion technologies depends on the fit between technological features and older adults' social support context.

Key words: affective companionship, social robots, loneliness, contextualized effectiveness, social support