Send message to Author

Paper short abstract

In this paper, we map spatio-temporal care arrangements where AI systems use voice biomarkers to detect health risks. We show how always-on machine-listening infrastructures reshape triage and eligibility via digital front doors and intensify bias and accountability tensions.

Paper long abstract

The future of health and social care is increasingly framed as a sociopolitical and economic urgency driven by demographic ageing, multimorbidity, and workforce shortages. In response, an emerging set of communicative AI (ComAI) innovations operationalises voice as a biomarker, promising predictive, pre-emptive, and preventive health through "always-on" machine listening.

We examine and map ComAI systems that translate voice into actionable risk signals across care sites including health insurers, telehealth services, clinics, and care facilities. These systems blur boundaries between diagnosis, prediction, and surveillance by classifying vulnerability, assessing (in)eligibility, and organising triage and routing through "digital front doors".

Empirically, we draw on (1) a thematic analysis of website materials from 80 ComAI providers and (2) 20 semi-structured interviews with innovation managers in healthcare and social care. We map emerging spatio-temporal care arrangements and analyse the sociotechnical visions and promises underpinning voice-biomarker infrastructures. While our focus is primarily institutional, we attend to how these arrangements position patients as subjects of continuous listening with limited recourse against the classifications made about them.

Our findings show that promissory futures of "better care" through such systems depend on continuous monitoring while shifting responsibilities and accountabilities across clinicians, organisations, and vendors. We further highlight infrastructural frictions, invisibilised labour, and concerns about consent, privacy, and bias-particularly where voice analytics misread or exclude certain bodies and voices (e.g., accent, age-related change, disability).

Keywords: voice biomarkers; health surveillance; communicative AI; sociotechnical visions; anticipatory infrastructures