Autonomous conversational agents for loneliness, social isolation, depression and anxiety in older people without cognitive impairment: Systematic review and meta-analysis
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Loneliness is a major psychological challenge in older adulthood, contributing to increased risks of depression, anxiety, and mortality. Conversational agents—technologies that interact with users via natural language—have emerged as potential tools to support psychological wellbeing in later life. This systematic review and meta-analysis evaluated the effects of autonomous conversational agents, including robotic and non-robotic systems, on loneliness, as well as social isolation, depression and anxiety in older people without cognitive impairment.
Seventeen studies with pre–post intervention data were included. Nine used physically embodied robots and eight employed non-robotic agents such as personal voice assistants, chatbots, or screen-based embodied agents. Due to the limited number of high-quality comparison studies, all meta-analyses were based on within-group pre–post comparisons. Meta-analytic results showed mild to moderate improvements in loneliness (standardized mean changes using change score [SMCC] = 0.350, 95% CI: 0.180-0.520) and depression (SMCC = 0.464, 95% CI: 0.327-0.602), with no study reporting symptom worsening. Subgroup analyses suggested a somewhat greater effect for robots. No study included validated measures of social isolation, and only one assessed anxiety.
These findings indicate that conversational agents—particularly social robots—may offer scalable support for older adults’ mental health, with potential especially for reducing loneliness and depression. Nonetheless, methodological limitations, including lack of blinded outcome assessment, inconsistent reporting and heterogeneous intervention designs, underscore the need for more rigorous research. Advances in large language models may further enhance the responsiveness and relevance of these technologies for supporting psychological wellbeing in aging populations.
