Improving Older Adults’ Social Interaction Quality with the Virtual Agent: Response Surface Model Based on Self-Disclosure Valence and Interpersonal Distance

Virtual reality compensates for older adults’ lack of social interaction. However, how to jointly design self-disclosure valence and interpersonal distance between virtual social agents and older adults to improve older adults’ social interaction quality remains unclear. Sixty-five older adults were randomly assigned to one of nine conditions across three self-disclosure valence levels (negative, neutral, or positive) and three interpersonal distances (too close, moderate, or too far). Ten metrics of social interaction quality were combined into a comprehensive quality index, and a response surface model was employed to examine the relationship. Results showed a significant positive U-shaped effect of valence on quality, no significant effect of distance, but a marginally significant interaction: when distance was too close, the U-shaped valence effect became more pronounced. These findings reveal that valence and distance are interrelated and require coordinated dynamic adjustments. Accordingly, practical design recommendations for older adults’ human–agent interaction were proposed.