Depression affects millions worldwide, yet diagnosis relies on subjective self-reports that may miss authentic behavior. This paper presents an approach linking speech acoustics to DSM-5 depressive-behavior indicators through a transparent Linkage Framework. Unlike black-box models, the framework explicitly maps acoustic features (pitch variability, pauses, speech tempo) to clinical indicators, enabling interpretable, indicator-level outputs. The system runs locally on commodity hardware (HW) to preserve privacy. Preliminary evaluation on DAIC-WOZ shows directionally consistent associations between acoustic features and DSM-5 indicators for psychomotor change and concentration difficulty, supporting the design rationale. Future work will validate on longitudinal datasets and extend multimodal integration while maintaining edge constraints.

Background: Accurate measurement of lifestyle factors is central to understanding how daily behaviors act as risk factors or protective buffers to non-communicable diseases. While wearable devices enable passive monitoring of physical activity or sleep, nutritional intake still depends on active participant input such as manual dietary logs or image-based recordings. Adherence to such logging tasks often declines rapidly, impacting data completeness and clinical utility. Theory-based reminders drawing on loss-framing or logging consistency feedback (i.e., tracking streaks) can improve adherence to lifestyle data collection, but effects of these strategies on repeated dietary logging remain unclear. Objective: To examine examines the effects of two theory-driven reminder strategies, loss-framing and logging consistency, on adherence to repeated, image-based meal-logging. Methods: We employ a micro-randomized trial (MRT) embedded within a 4-week observational lifestyle phenotyping study in Switzerland (N=200, age ≥45, BMI ≥25 kg/m²). Participants photograph their meals at each mealtime (breakfast, lunch, dinner) using a mobile app over 28 days. A decision point is scheduled prior to participant-defined habitual mealtimes. Participants are randomly assigned with equal probability to: (1) a reminder emphasizing loss of a daily financial reward for not logging ("loss-framing"), (2) a reminder providing feedback on recent logging consistency ("logging consistency"), or (3) a neutral reminder ("active control"). The proximal outcome is whether the participant logs a meal within two hours of receiving a reminder. Participants earn a daily financial reward contingent on meal logging completion. To estimate intervention effects, we will use marginal excursion effect models for binary outcomes, adjusting time-varying covariates (e.g., day in study) and baseline covariates (e.g., age, gender). Ethics approval for this study was granted by the Cantonal Ethics Committee of Eastern Switzerland (BASEC ID: 2025-00972). Results: Enrollment began in November 2025, and the study was initiated on December 8, 2025, with an anticipated completion date of January 2027. Conclusions: By clarifying the proximal effects of loss-framed and consistency-based reminders, findings will inform the design of future digital health studies to improve meal-logging adherence in daily life. This work contributes to the development of scalable, theory-driven reminder strategies for enhancing dietary data quality in observational and interventional research. Clinical Trial: ClinicalTrials.gov NCT07555262; https://clinicaltrials.gov/study/NCT07555262 ClinicalTrials.gov NCT07373418; https://clinicaltrials.gov/ct2/show/NCT07373418

If trade diplomats thought they knew one thing, it was how to cut industrial tariffs. Yet the Doha deadlock rests squarely on the inability to compromise on industrial tariff cuts. This column says that the arguments made for higher levels of ambition don't stand up to much scrutiny and should not be allowed to provide a basis for a continuing impasse.