JMIR Publications

ABSTRACT

Ambient sensing technologies increasingly enable continuous, low-burden observation of sleep, mobility, activity and other behavioral or physiological patterns among older adults living at home. Such data can support the derivation of digital biomarkers capable of identifying deviations from established patterns and signaling possible changes in health or function. Yet a technically credible signal is not necessarily clinically or personally meaningful. An accurately detected change in nocturnal activity, for example, may reflect cognitive deterioration, pain, medication effects, sleep disturbance, a change in household circumstances, an established personal routine, or an intentional choice. Existing research has addressed important components of this interpretive problem. Studies of activity-of-daily-living telemonitoring have described how professionals compare sensor-derived information with other sources before assessing risk and modifying care; health-smart-home research has demonstrated clinicians' difficulties interpreting sensor data without individual reference points and clinical information; and emerging context-aware approaches emphasize within-person change rather than population norms. In parallel, person-centered artificial intelligence scholarship emphasizes professional judgment, individual narrative and patient preferences. However, these forms of contextual knowledge remain insufficiently integrated within a single operational account of how an ambient digital signal acquires person-specific significance before informing care. This paper proposes a Contextual Interpretation Framework comprising five analytically distinct but interacting forms of knowledge: technical credibility, longitudinal significance, clinical significance, biographical significance and deliberative significance. It argues that ambient digital biomarkers should not be treated as self-interpreting evidence and that technically reliable detection can coexist with substantial uncertainty about what an observed change means for a particular person. The framework shifts attention from whether artificial intelligence should remain “human-in-the-loop” towards the more specific question of what interpretive work humans, technologies and care networks must perform within that loop. It further proposes that monitoring, seeking additional information, clinical assessment, intervention and deliberate non-intervention should all be regarded as legitimate outputs of person-centered interpretation. The framework offers a conceptual basis for future empirical research into the safe and person-centered integration of ambient digital biomarkers within aging-in-place care.