Perceived Algorithmic Care Intensity and Older Adults' Autonomy in Community Smart Care: Mixed Methods Study.

Abstract

Background: As smart older adult care shifts from basic information support to more continuous algorithm-driven care, concerns have grown about whether such systems support older adults' independence or weaken their sense of autonomy. Although previous research has focused mainly on technology acceptance, less is known about how algorithmic care shapes perceived autonomy and through which psychological pathways this occurs.

Objective: This study examined whether perceived algorithmic care intensity reduces older adults' perceived autonomy in community smart care settings, whether decisional substitution and perceived surveillance mediate this relationship, and whether digital literacy moderates these effects.

Methods: A 3-stage design was adopted. First, a qualitative prestudy with 15 older adults was conducted to refine the focal constructs and measures. Second, a vignette-based experiment with 233 valid participants tested the effects of algorithmic care intensity on decisional substitution, perceived surveillance, and perceived autonomy. Third, a community-based survey with 298 valid participants examined the moderated mediation model in real-world smart older adult care settings.

Results: In study 1, compared with the low-intensity condition, the high-intensity condition significantly increased decisional substitution and perceived surveillance, while reducing perceived autonomy. Indirect effects through decisional substitution (effect=-0.072, 95% CI -0.144 to -0.016) and perceived surveillance (effect=-0.084, 95% CI -0.161 to -0.019) were both significant. In study 2, perceived algorithmic care intensity positively predicted decisional substitution (β=.400, P<.001) and perceived surveillance (β=.440, P<.001), and negatively predicted perceived autonomy (β=-.356, P<.001). The indirect effects through decisional substitution (effect=-0.112, 95% CI -0.172 to -0.065) and perceived surveillance (effect=-0.149, 95% CI -0.215 to -0.093) were significant. Digital literacy significantly weakened the effect of algorithmic care intensity on decisional substitution (β=-.174, P=.002), but not on perceived surveillance (β=-.071, P=.18).

Conclusions: Algorithmic care may undermine older adults' perceived autonomy both directly and indirectly through decisional substitution and perceived surveillance. Digital literacy appears to be a selective rather than universal buffer. These findings extend smart older adult care research beyond technology acceptance and clarify the tension between empowerment and control in algorithmic care.