As household appliances become more digital and assistive, adaptive functions may provide new opportunities to support individuals with health-related limitations in activities of daily living (ADL). Such systems, however, depend on users’ willingness to provide behavioral data that enable personalized assistance. While previous research has primarily focused on technology acceptance and use, health-related differences in willingness to share usage data for adaptive assistive systems have received little attention. Using data from a population-representative survey in Germany (n = 949), this study investigates whether and to what extent ADL limitations influence willingness to share usage data with assistive systems that enable adaptive user interfaces for household appliances such as ovens and washing machines. Logistic regression analyses show that individuals with such limitations are more likely to be willing to share usage data. This effect persists even after accounting for a broad set of potential confounders like differences in technology experience, personality traits, and sociodemographic characteristics. The findings extend current research beyond technology acceptance and use by demonstrating a previously underexplored association between health-related limitations and willingness to participate in the data-driven processes underlying personalized assistance. This greater willingness to share data suggests potential for adaptive household technologies to provide tailored assistance for this target group, while highlighting the need for further research into the mechanisms shaping data-sharing willingness and how voluntary and informed participation can best be supported across cultural contexts.
Background The pervasive integration of information and communication technologies (ICTs) in knowledge work has intensified concerns about technostress, particularly techno-overload, which is associated with cognitive resource strain and burnout. Digital detox—the intentional temporary disconnection from ICTs—is theorized to mitigate this process through both preventive and coping functions, yet whether it operates as a specific buffering mechanism or is generally associated with well-being independent of such mechanisms remains unclear. Objective This study examines an Overload model positioning cognitive overload as a mediator between techno-overload and burnout, with digital detox hypothesized to moderate this pathway at two points: between techno-overload and cognitive overload (prevention) and between cognitive overload and burnout (coping). Methods Data were collected from 413 Slovak knowledge workers (50.85% female; M = 42.89 years). Participants completed instruments measuring techno-overload, cognitive overload, digital detox, and burnout. Analyses were performed using confirmatory factor analysis, moderated mediation analysis, and network analysis. Results Techno-overload was associated with burnout, with cognitive overload partially mediating this relationship. Digital detox did not moderate the prevention or coping pathway, and the moderated mediation model was not supported. However, digital detox was associated with lower burnout levels. Network analysis revealed that disconnecting from work-related ICTs outside working hours was most centrally associated with reduced exhaustion and cognitive impairment. Discussion Cognitive overload emerged as a mechanism linking digital demands to burnout. Although digital detox did not buffer this pathway, its direct association with lower burnout suggests that it may be more closely related to employees' overall well-being than to buffering specific stressor–strain relationships. The network analysis further indicates that disconnection from work-related ICTs outside working hours may represent the most relevant aspect of digital detox in relation to burnout symptoms, highlighting the importance of psychological detachment beyond formal working time.
Integrating satellite-derived raster data (e.g., NDVI) with ground-based vector readings (e.g., air quality sensor readings) is essential for real-time smart city analytics. However, existing geospatial join systems depend on computationally expensive geometric predicates and centralized processing. They are thus considered ill-suited for resource-constrained edge–cloud architectures. We propose GeoEdgeJoin, a novel framework that integrates multi-modal geospatial data by applying a lightweight raster-vector join equijoin operator using Discrete Global Grid Systems (DGGS) such as Uber’s H3 or geohash. The system encodes spatial data into DGGS keys (e.g. Geohash or Uber’s H3) at the edge and relies on those keys to send these data to the Cloud, which ensures co-partitioning in the cloud and thus eliminates costly network shuffle and geometric computations. We have evaluated our system using real-world multi-modal geospatial data (i.e., low-cost sensor’s air quality vector data and Landsat 8 satellite raster data). Our results show that GeoEdgeJoin achieves up to 40× faster join times than exact geometric baselines (i.e., Apache Sedona, a Cloud-only specialized geospatial big data processing system) while maintaining robust statistical accuracy. Furthermore, we show that Uber’s H3 encoding outperforms Geohash by 17% in latency due to uniform spatial partitioning, which helps reduce straggler tasks in distributed environments. Our system supports both exact and error-bounded approximate analytics by using stratified DGGS-based sampling, to enable scalable, low-latency environmental monitoring. This work provides a foundational paradigm for next-generation, edge-native geospatial intelligence. Our work provides a foundational paradigm for real-time, multi-modal geospatial analytics in complex and advanced smart city applications.
Patients often describe their symptoms in broad or subjective terms (e.g., “I feel tired,” “my chest feels weird”), which can make early clinical triage difficult and delay referral to the right specialty. Many conversational AI tools still struggle in this setting because they rely heavily on complete, well-structured symptom descriptions. In this work, we introduce MedRef-KG, a framework that combines a medical knowledge graph with a multi-agent dialogue process to turn incomplete patient narratives into specialist referral recommendations. We build a Neo4j medical knowledge graph using an ensemble of five large language models, resulting in 5773 nodes and 23,752 relationships. The system follows three main steps: (1) extracting key symptoms and asking targeted follow-up questions when information is missing, (2) managing a goal-driven doctor–patient dialogue, and (3) predicting the appropriate specialty using Qwen3-30B-A3B. On our evaluation, MedRef-KG reached a weighted F1-score of 0.9051, outperforming strong baselines such as fine-tuned medical GPT-4 (＋6.01%), OpenBioLLM-Llama3-70B (＋4.41%), Me-LLaMA (＋4.51%), BioBERT (＋8.51%), and Meditron-70B (＋20.01%). Performance was particularly strong for Musculoskeletal (F1 = 0.957) and Respiratory (F1 = 0.919) referrals. Overall, the results suggest that pairing structured medical knowledge with coordinated multi-agent reasoning can improve referral quality when patient input is incomplete—especially in real-world and resource-limited triage settings.
