Reaching for Mental Wellbeing via Reachy Mini
Proceedings of the AAAI Symposium Series · 06.11.2026 · 10 (1) · 1264-1273
DOI: 10.1609/aaaiss.v10i1.43555 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Symposium Series, online erschienen am 06.11.2026. Verfasst von Aarat Satsangi und 3 weiteren Autor:innen.
Abstract (Originalsprache)
Continuous assessment of patient psychological well-being remains a critical challenge in clinical inpatient care, where demanding workloads limit manual psychiatric evaluations and static surveys induce patient fatigue. In this paper, we present an embodied, agentic bedside framework targeting deployment on the Reachy Mini robotic companion to support the global AI-ready nursing workforce. The system integrates a real-time, tri-modal emotion recognition pipeline tracking vision (facial tracking and dynamics), acoustic prosody (voice-activity-driven speech emotion recognition), and transcribed semantic text. Complementing continuous emotion tracking, the robotic agent administers an adaptive 11-question conversational screening protocol that evaluates psychological trajectories across the Patient Health Questionnaire-9 (PHQ-9) and Generalized Anxiety Disorder-7 (GAD-7) clinical tests, incorporating targeted probing to decouple somatic illness from primary affective distress. Synchronized interaction analytics and automated safety alerts are rendered on a clinician-facing nursing dashboard to facilitate proactive triage without inflating administrative overhead. We evaluate the perception models on a standard emotion benchmark and the conversational screener on a synthetic clinical cohort spanning the full range of depression and anxiety severity. The screening study shows that a single merged interview recovers both a depression and an anxiety item from one prompt on most shared questions, that item-level scores stay close to ground truth even when patients describe symptoms indirectly, that severity-band classification under natural, indirect dialogue falls to near-chance accuracy despite this item-level agreement, and that the strictness of the blind scorer governs a clear tradeoff between self-harm recall and false alarms. Together these results position the framework as a feasibility step toward continuous, interpretable psychological telemetry at the bedside.
Zitieren
APA
Satsangi, A., Lahourpour, S., Allamraju, S., & Narayan, A. (2026). Reaching for Mental Wellbeing via Reachy Mini. Proceedings of the AAAI Symposium Series, 10(1), 1264-1273. https://doi.org/10.1609/aaaiss.v10i1.43555
Vancouver
Satsangi A, Lahourpour S, Allamraju S, Narayan A. Reaching for Mental Wellbeing via Reachy Mini. Proceedings of the AAAI Symposium Series. 2026;10(1):1264-1273. doi:10.1609/aaaiss.v10i1.43555
BibTeX
@article{Satsangi2026,
author = {Satsangi, Aarat and Lahourpour, Sara and Allamraju, Sumira and Narayan, Apurva},
title = {Reaching for Mental Wellbeing via Reachy Mini},
journal = {Proceedings of the AAAI Symposium Series},
year = {2026},
volume = {10},
number = {1},
pages = {1264--1273},
doi = {10.1609/aaaiss.v10i1.43555}
}
EUVIMED