25 December 2024 in IEEE Access
In recent years, school attendance issues among university students have been increasing, which can lead to repeating courses, dropping out of school, or even social withdrawal. Despite the existence of counseling services, students often delay help-seeking, which can cause symptoms to worsen and make support more difficult. Thus, it is essential to identify at-risk students early and encourage them to seek help. A realistic approach must minimize the burden on students, rely only on devices they already own, and operate correctly even for students who are less engaged or prone to social withdrawal. While several techniques have been proposed to estimate individual indicators, they fail to address one of these requirements due to requiring additional devices or requiring user attention and interaction. In this paper, we propose an unobtrusive screening method for detecting subtle signs of school attendance issues in university students. We develop a smartphone app to collect sensor data and collect ground truth information using questionnaires for (1) sleep problems and (2) decreased student engagement. We collect data from 58 university students for about 10 months, and build estimation models for the above indicators. Our evaluation shows that the estimation models are sufficiently accurate in flagging problematic cases. The indicators can then be used to notify at-risk students and medical practitioners, enabling timely intervention. This screening is not intended to replace traditional face-to-face medical examinations, but rather to selectively flag at-risk students and connect them with medical experts.
6 April 2026 in ACM Transactions on Computing for Healthcare
Prolonged loneliness in older adults increases the risk of dementia, cardiovascular diseases, and premature death. However, timely detection and behavioral pattern analysis are challenging due to the slow progression of loneliness and diverse individual lifestyles. To address these barriers, we propose SMILE, an end-to-end framework for sensor-driven detection and counterfactual pattern analysis of loneliness. Our approach uses non-intrusive sensors to passively observe behavior and build a personalized profile over time. When SMILE detects meaningful changes in activity patterns, it employs generative models to produce counterfactual sensor profiles—synthetic representations of behavior predicted to result in lower loneliness scores. Suggested pattern changes are identified as the difference between observed and counterfactual profiles, highlighting the behaviors most associated with loneliness. We conducted a 6-month feasibility study with 18 participants from the United States and 15 from Japan to develop and evaluate SMILE. Our framework integrates a Time-series Transformer (TST) model for loneliness detection, which achieved 96.25% accuracy, and a diffusion model for generating behavioral explanations, which outperformed a baseline by up to 76.69%. These findings highlight the potential of SMILE to support sensor-driven, personalized loneliness detection and behavioral explanation in aging populations.