JMIR Publications

ABSTRACT

Background:

Social connection is a critical determinant of well-being in later life. However, older adults often experience social isolation and increased loneliness, which are associated with mental and physical health challenges. Artificial intelligence (AI)-based interventions have emerged as scalable approaches to support social connection. A growing body of research has examined the effects of individual AI technologies in improving social connection across diverse devices and care settings. Despite growing evidence on their effects, little attention has been paid to the common pathways through which these interventions support social connection.

Objective:

This scoping review aimed to develop a typology of common categories of support across AI-based interventions and to map how these categories are configured in different intervention contexts.

Methods:

Five databases (PubMed, MEDLINE, CINAHL, PsycINFO, and Web of Science) were searched for studies published between 2016 and 2025. Studies were included if they examined AI-based interventions designed to support social connection among older adults, conceptualized as a multidimensional construct (e.g., loneliness reduction, social interaction, and behavioral engagement).

Results:

A total of 1110 articles were retrieved, of which 47 met the inclusion criteria. Qualitative and mixed-methods studies were the most prevalent (68%), with quantitative studies accounting for the remainder (32%). Four common support categories were identified: relational interaction (forming ties), emotional support (enhancing interaction quality), functional support (sustaining engagement), and risk detection (reducing constraints). Relational interaction and emotional support were present across all studies. Functional support and risk detection were observed in 53% and 17% of studies, respectively. The most prevalent configuration combined relational interaction, emotional support, and functional support, forming a companionship-plus-routine pathway that sustained engagement in everyday contexts. Functional support was associated with sustained participation in daily routines and activities, which may have contributed to opportunities for ongoing social interaction. Risk detection, while less prevalent, functioned as a constraint-reducing condition that supported continuity of engagement.

Conclusions:

This review proposes a conceptual framework that shifts attention from specific technologies to the interactional configurations that enable AI to support social connection across care settings. This framework may inform the design and tailoring of AI-based interventions to address the multidimensional nature of social connection among older adults.