Towards Closed-Loop Chemical Interfaces: Sensing Mood and Modulating Physiology Through Wearable Systems
Abstract
The body runs much of its internal signaling on chemistry, while HCI has few well-characterized wearable channels for controlled systemic chemical delivery. This thesis advances closed-loop chemical interfaces: wearable systems that sense a user’s momentary state and modulate physiology through transdermal substance delivery. Three independently evaluated in-the-wild and in-human studies establish the sensing and actuation components such a loop requires; joining them into a working loop is left as a defined research agenda.
On the sensing side, the studies show that commercial-grade wearables can now both detect and predict mood and cognitive state. Personalized linear models trained on 7 days of hourly ecological momentary assessment decode six BRUMS (Brunel Mood Scale) mood dimensions (tension, depression, anger, vigor, fatigue, confusion), calmness and happiness, and physical states (hunger, thirst, pain) from a low-cost open-source smartwatch, with all inference running on the device. A companion study moves from detection to anticipation, estimating before a smartphone session begins whether the user will later regret it (N = 21, 1,445 sessions; within-person AUC (area under the receiver operating characteristic curve) = .740 from pre-session smartwatch-derived and phone-context features). On the actuation side, a compact, wirelessly controlled wearable sonophoresis device was evaluated with an active reservoir loaded with 100 mg of caffeine. In a double-blind, between-subjects study (N = 26 enrolled, heart-rate standard deviation (HR SD) analysis n = 13), assignment to the caffeine-reservoir condition was associated with measurable cardiovascular differences (pre-registered HR SD contrast, 𝑝 = 0.025, 𝑑 = 1.48), an early in-human signal that a wearable acoustic device can produce systemic physiological effects.
Situated within Chemical HCI, the broad area of interfaces that interact with the body through chemistry, the thesis proposes its own architecture for connecting these primitives into a closed-loop chemical interface, built from a runtime that acts on a sensed state, a personalized dosing model, and a safety program. It closes with an ethical and governance analysis of physiological actuation, mental privacy, consent, and dependency that a chemical interface must confront before deployment. Together, the studies lay a foundation for wearable systems that both read and write to the body’s chemical signaling.