Dual-path and threshold mechanisms of AI agents in elderly emotional support: linking social engagement and cognitive load to loneliness reduction

Population ageing has intensified loneliness, reduced social engagement, and cognitive burden among older adults. Although AI agents are increasingly used to provide companionship, the socio-cognitive mechanisms underlying their emotional benefits remain unclear. This study investigates how AI interaction alleviates loneliness through two pathways: enhanced social engagement and reduced cognitive load. A three-phase empirical design was conducted. Study 1 (N = 90) showed that AI-based interactions significantly decreased loneliness while improving social engagement and lowering cognitive load. Study 2 identified social engagement and cognitive load as parallel mediators linking AI interaction to emotional outcomes. Study 3, using Necessary Condition Analysis, revealed that social engagement functions as a core necessary condition, whereas cognitive load reduction plays a supportive role. Together, the findings demonstrate a dual socio-cognitive mechanism in which social connection drives emotional improvement and cognitive relief facilitates it. Integrating Social Agency Theory and Cognitive Load Theory, this research proposes a threshold-based explanatory model and offers practical guidance for designing cognitively adaptive and socially sustainable AI systems for ageing populations.