Being Sensed, Scored, and ‘Connected’: A Social-Gerontological Critique of AI-Driven Loneliness Detection for Older Adults in Taiwan

Abstract Enthusiasm for artificial intelligence (AI) that continuously infers emotion promises to change how care systems identify older adults at risk of loneliness. We use AI-driven loneliness detection to mean systems that infer loneliness, or a proxy for it, from sensed data rather than from a person’s answer, rendering the inference as a score or category used in care; they range from facial and vocal emotion recognition to behavioral risk prediction from everyday activity data. Engineering feasibility, more than any social-gerontological understanding of loneliness, has driven this agenda, and little attention has gone to how older adults in different cultural contexts wish to be known and supported. Because such detection is not yet in routine use in Taiwan, this Forum article adopts an anticipatory-governance stance, examining before deployment a trajectory that policy makes plausible. Anchored in Taiwan’s ageing society and long-term care system, it argues that culturally responsive design should begin with older adults’ own meanings, relational obligations, and expectations of privacy and autonomy, not with the sensors. It identifies three tensions for social gerontology: the reframing of a relational experience as an individual biometric deficit; the risk that algorithms read culturally patterned emotional restraint as noise; and the slippage from care toward surveillance once risk indices enter routine care workflows. Taking seriously the case for technological augmentation of scarce care resources, the article proposes principles for a socially accountable approach and a research agenda on how older adults negotiate being sensed, scored, and ‘connected.’