There is a rich variety of approaches for generating scientific knowledge. However, deductive hypothesis testing is often considered a defining characteristic of the scientific method. It has long been shown that individuals' perceptions of how science works often do not align with the actual nature of science, which can adversely impact scientific progress and scientific literacy. Here we present an emergent study exploring how postdoctoral scholars whose work involves materials and chemistry describe the role of hypothesis testing in their work. Through a reflexive thematic analysis of 16 interviews, we highlight how participants broadly report valuing hypothesis-based approaches and view them to be a defining characteristic of "good" science. However, simultaneously, a portion of participants find that deductive hypothesis testing is not always the most practical or useful approach in their work. This dissonance between their value of deductive hypothesis testing and their own scientific practices has implications for how participants view themselves and their work. We argue that the cultural master narrative of science emphasizes the use of deductive hypothesis-based reasoning, and that confronting incompatibilities with this narrative requires participants to formulate alternative narratives to rationalize their work in light of the cultural master narrative. This process influences participants' self-perception, their scientific practices, and/or their scientific communication. These findings illustrate that explicit consideration of the nature of science continues to be important in higher education, and motivate further study in this area. These findings also reinforce existing calls for educators and scientists to explicitly normalize the existence of multiple legitimate approaches for generating scientific knowledge.

Accurately mapping environmental and social inequalities at fine spatial scales is critical for urban policy, yet the data required are often costly and infrequently updated. Vision foundation models can extract information directly from satellite imagery, offering a rapid and scalable alternative. We compare three modelling pipelines for predicting two contrasting indicators of urban inequality - air pollution and house prices - across England on a fine hexagonal grid: regression models trained on structured features of form and function (census, land cover and urban morphology), models trained on 128-dimensional image embeddings from a geospatial foundation model and a hybrid of the two, each evaluated with and without coarse regional context. Structured features achieve the best overall accuracy, but image embeddings become competitive once regional context is added - most clearly for air pollution, where image-only models reach a median R2 of 0.78 (0.85 with regional context), indicating that the embeddings capture genuine image signal. For house prices the picture is more cautious: image-only models achieve a median R2 of around 0.58; however, much of the embeddings' apparent gain reflects coarse spatial location rather than image content. Off-the-shelf satellite embeddings, while not yet surpassing data-intensive approaches, are a promising low-cost complement, particularly for rapid, large-scale, or repeated analyses and in settings where traditional data are limited.

Phage therapy has become an important strategy against the crisis of antibiotic resistance for its potential to specifically target pathogenic bacteria. However, the narrow host range of phage makes the screening of precise matches of clinical strains inefficient, while existing computational tools are difficult to capture the dynamics of phage-host interactions due to their reliance on single modal features (genome or proteins). In this paper, we propose FusionPHI, a phage-host interaction prediction model that adopts a fully connected neural network architecture and takes multi-modal features as input, including the k-mer statistics of genome sequences, the physicochemical properties of proteins, and the embedded representations of evolutionarily conserved gene motifs for a comprehensive understanding of phage-host interactions. Moreover, FusionPHI contains a dual-stage attention-drive feature fusion module that integrates a self-attention mechanism to optimize the correlation among the features within a single modality followed by a cross-attention mechanism to dynamically fuse genetic distribution patterns with the protein function information in global or local regions of sequences. The experiments of phage-host interaction prediction show that FusionPHI achieves 91% ROC AUC in cross-validation, demonstrating competitive performance compared to the evaluated baseline methods on our dataset, and ablation experiments further validate the necessity of multi-modal features and attention mechanism. The case study of E. coli infected by the M13K07 phage further validate the prediction ability of the proposed FusionPHI model.

Cerebral Visual Impairment (CVI) of early onset has primarily been studied in children, but little is known about its progression into adulthood, despite CVI being recognized as a lifelong condition. This study investigates higher visual function deficits (HVFDs) in adults with early-onset CVI through a prospective controlled design. We first assessed the feasibility of an adapted HVF Question Inventory (HVFQI-59Q) in detecting HVFDs in adults with CVI compared to neurotypical (NT) adults (ANTs, n = 67). We then compared HVFD patterns in adults with early-onset CVI (A-Early Onset, n = 29), adults with late-onset brain injury (A-Late Onset, n = 22), and a previously interviewed cohort of children with CVI (n = 92) and child neurotypicals (n = 120). The HVFQI-59Q was used to evaluate the spectrum and severity of HVFDs across the adult groups. Results showed HVFDs are present in A-Early Onset, with similarities and distinct differences compared to children. The A-Late Onset group exhibited a broader range of deficits and lower median values, suggesting more variability in visual impairments. These findings confirm that HVFDs persist into adulthood and that late-onset brain injuries may result in HVFDs similar to CVI, challenging previous assumptions about localized effects and the need for a review of terminology for A-Late Onset, highlighting the need for further research in the adult group. The study underscores the importance of continuous CVI evaluation across the lifespan and supports the HVFQI-59Q as an effective clinical tool for detecting HVFDs.

Dengue remains a major public health problem in endemic regions, including Bangladesh. Forecasting algorithms relying on climatic variables may not capture epidemiological information such as dengue serotype patterns. This study proposes a horizon-dependent dengue forecasting framework and applies it to Bangladesh. This pipeline included temporal, meteorological, demographic, and epidemiological covariates in a district-panel pipeline and compares classical, machine-learning, and deep-learning algorithms using aggregate, horizon-wise, regime-wise, outbreak-detection, and uncertainty metrics. Coefficient-based interpretation, SHAP, permutation importance, and nonlinear causal-dependence analysis were used to examine predictor impacts across horizons. SARIMAX was used as the baseline model. TFT produced quantile forecasts but showed poor interval calibration during outbreak periods. Results show that no single model performed best across all forecasting horizons: SARIMAX ranked highest for one-step outbreak alerting (precision = 0.886, recall = 0.824, F1 score = 0.854, and ROC-AUC = 0.950), whereas Prophet performed best for pooled magnitude forecasting at horizons 2-6. MLR showed competitive performance in regime-wise evaluation. These findings show that dengue forecasting algorithms should be selected according to the intended decision objective and evaluated using task-relevant protocols.

Brain-computer interface (BCI) systems have advanced with deep learning, but they are still limited by designs tied to specific applications, poor scalability, weak portability, the need for user-specific adaptation, and privacy concerns. We present BELT, a modular Bayesian Edge-Cloud architecture based on three principles: (i) Bayesian priors and posteriors to balance generalization and subject-specific learning, (ii) lightweight classifiers suitable for embedded devices, and (iii) task-aware compression to reduce bandwidth and improve privacy in edge-cloud communication. To show feasibility, we implement BELT-lite as an instantiation of BELT, a lightweight version built only from linear time-invariant operations, making it directly compatible with digital signal processing hardware. Using the BCI Competition IV-2a and IV-2b motor imagery datasets (18 subjects total, ten-fold cross-validation), BELT-lite achieved strong posterior performance after subject-specific fine-tuning: mean accuracy of 87.9%±6.8% on Dataset B and 80.6%±8.6% on Dataset A with data augmentation. After adaptation, four subjects from Dataset B and two from Dataset A exceeded 90% accuracy. On ARM Cortex-A7 hardware, BELT-lite achieved a mean latency of 6.75 ms per sample, significantly faster than EEGNet's 8.36 ms (p < 10-17)-a 21% speed improvement-at the cost of a modest but statistically significant accuracy reduction of approximately 2.7 percentage points compared to EEGNet. Network Tuning Blocks allowed partial parameter freezing: classifier-only fine-tuning incurred a modest 2-5% accuracy drop while substantially reducing training cost. Compression via the task-unaware autoencoder reduced data size by 3.3× while maintaining high accuracy: prior-model performance stayed within ≈1% of the uncompressed baseline (with slight improvements in some configurations), full posterior fine-tuning showed a ≈1% drop, and classifier-only fine-tuning incurred a ≈3% drop-an acceptable trade-off for privacy-preserving edge-cloud communication, where only a compressed latent representation is transmitted instead of raw EEG. Notably, this task-unaware autoencoder (trained solely to reconstruct the input) consistently outperformed autoencoders that also incorporated classification objectives (task-aware or task-only), providing the best accuracy-compression trade-off across all fine-tuning scenarios. These findings show that BELT provides a principled design for modular and scalable BCIs, while BELT-lite demonstrates that the approach supports accurate, efficient, and portable implementations. Together, they point toward BCI systems that are more practical, mass-producible, and privacy-aware, enabling wider use in real-world settings.

In 2023, the Children's Museum, a new permanent exhibition for children, opened at the National Museum in Prague. Primarily designed for children aged 5-10, the exhibition aims to provide an educational alternative to traditional (pre-)school education and classical museums. Through a combination of interactive and traditional presentation methods, the exhibition offers imaginative and interdisciplinary perspectives on various phenomena. It encourages visitors to explore and learn in their own ways. This approach requires a significant interpretative effort from both children and their accompanying adults. To evaluate how the exhibition's concept resonates with visitors, we observed visitor interactions over a period of 2.5 months, analysing 720 interactions involving more than 2000 visitors. Our analysis revealed that older children (aged 6-10, the primary target group) showed higher levels of interaction and engagement compared to younger children. However, the primary target group constituted only 23% of the child visitors, with 72% being younger children. This indicates that the exhibition is not attracting its intended audience effectively, and the majority of the young visitors may face challenges due to the complexity of the exhibition's concept. We discuss the implications of these findings and suggest strategies to align the exhibition's content better with its target audience.

Non-stationary time series are common in many real-world domains, including infectious disease spread, where the underlying relationships between variables evolve over time. However, most existing forecasting methods assume stationarity and fail to capture changing causal dynamics. To address this challenge, we propose the Causal Regime Bayesian (CaReBayes) forecasting framework, which integrates regime detection, causal discovery, and Bayesian forecasting within a unified approach. CaReBayes segments time series into regimes using temporal causal discovery, fits a Bayesian structural autoregressive model for each regime, classifies the current regime, then performs regime-specific forecasting with uncertainty quantification. The framework introduces methodological advances: a grid-search procedure for automated regime-dependent causal discovery, a classification method that assigns future observations to regimes based on learned Bayesian structures, and regime-conditioned Bayesian structural forecasting. Across both synthetic and Ontario COVID-19 time series data, CaReBayes outperforms benchmark models for time series forecasting. In addition to improved forecasting performance, it produces regime-dependent causal graphs that summarize candidate structural relationships in the system, enhancing interpretability.

Ecosystem health is a measure that is used to describe the condition of the ecosystem. The stability of the ecosystem and sustainable growth are represented by ecosystem health which covers the overall condition of the environment. The loss of habitat and deterioration of ecosystem services are the outcomes of coastal urbanization and ecosystem health damage. Potential natural disasters and biodiversity loss are happening more frequently and intensely which makes an emerging issue of environmental resilience. That's why ecosystem health assessment is must. The 1,47,570 km2 of Bangladesh has 47,201 km2 of coastal zone consisting of 19 districts which is 32% of the country and about 43.8 million people live here. This study aims to assess the ecosystem health of 2013 and 2023 and to investigate the change in the coastal ecosystem health in evaluating with past and present conditions of the ecosystem. For assessment, a random forest machine learning model has been used in conjunction with remote sensing and GIS due to its high accuracy and efficiencies. The PSR (Pressure-State-Response) model has been used in this investigation. This study reveals that the ecosystem health of 2023 deteriorated mostly than in 2013. About 19.76% of the lower ecosystem health area of coastal zone has been increased over the decades. In addition, 6.92% and 12.84% of higher and medium ecosystem health areas have been degraded accordingly. The Satkhira district had higher ecosystem health and Chittagong had lower ecosystem health in 2013; whereas in 2023, it became Narail and Jhalokathi accordingly. The point is that the ecosystem is down falling from the time on. This concerning factor can assist the body in making decisions on how should interact with the environment with adequate controlling crucial measures to heal the environment's health.

Teaching is widely documented as one of the most psychologically and physiologically demanding knowledge-intensive occupations, with chronic exposure to combined intellectual, emotional, and regulatory workloads placing teachers at elevated risk for autonomic dysregulation, cardiovascular strain, cognitive decline, and occupational burnout. A question that occupational health research has not adequately resolved is whether the functional consequences of career-long teaching are driven primarily by biological ageing or by the accumulation of occupationally specific stressor exposure, two processes that are conceptually distinct yet rarely examined separately. We assessed psychophysiological, emotional, and cognitive indicators in 103 female secondary school teachers in Karaganda, Kazakhstan, stratified by age and length of service. The assessment battery included heart rate variability recording, systolic and diastolic blood pressure measurement, the Ruffier physical performance index, the Well-being, Activity, and Mood questionnaire, the Spielberger-Khanin State Anxiety Inventory, the Landolt ring correction test, a digital selective attention test, and three sensorimotor reaction time measures. To allow direct comparison of the explanatory power and predictor composition of the two variables, separate multiple linear regression models were constructed with chronological age and length of professional service as dependent variables, with the full psychophysiological battery serving as independent variables in both. Chronological age and length of service were strongly intercorrelated (r = 0.915; p < 0.001), and partial correlation analysis showed that neither variable independently predicted digital test performance once the other was controlled. The age-as-dependent model was statistically significant and explained 42.2% of the variance (R2 = 0.422; adjusted R2 = 0.392; F = 14.166; p < 0.001), with significant predictors confined to physiological and cognitive domains: perceived activity (β = 0.417), digital test response time (β = 0.415), subjective well-being (β = -0.365), systolic blood pressure (β = 0.190), and correction test execution time (β = 0.171). The service-length model also showed strong explanatory coverage (R2 = 0.510; adjusted R2 = 0.479; F = 16.657; p < 0.001) and additionally captured the Ruffier index (β = -0.209), reactive anxiety (β = -0.189), and digital test response time (β = 0.345) as significant predictors alongside activity, well-being, and systolic blood pressure. These differences in predictor composition suggest that length of professional service is associated with a broader functional profile than chronological age, additionally capturing physical performance capacity and emotional reactivity. Taken together, the differential predictor structures indicate that biological ageing and cumulative occupational exposure operate through partially distinct functional pathways, and that treating the two variables as analytically interchangeable may obscure information relevant to the design of targeted occupational health monitoring programmes. These findings support the analytical differentiation of age and service length in occupational health research and provide a basis for developing career-stage-stratified preventive strategies for teachers.

Forest restoration reshapes nutrient cycling and soil stoichiometric regulation in fragile plateau ecosystems. To clarify variations in soil C:N:P stoichiometry and its environmental controls during ecosystem recovery on the southwestern China plateau, this study investigated three representative forest types along a restoration gradient, including Pinus yunnanensis forest, mixed conifer broadleaf forest, and secondary evergreen broadleaved forest. Soil samples were collected from three depths of 0-20 cm, 20-50 cm, and 50-80 cm, together with dominant tree leaf samples, and a total of 27 composite soil samples and 81 undisturbed soil cores were analyzed. Heatmap analysis, Mantel tests, redundancy analysis, and random forest modeling were used to evaluate stoichiometric characteristics and regulatory mechanisms. The results showed that soil organic carbon or total organic carbon and total nitrogen were the dominant factors controlling soil stoichiometry across forest types, whereas the effects of soil layer, iron and aluminum oxides, pH, and nutrient fractions varied with restoration stage. In the Pinus yunnanensis forest, available K and available Al were the primary predictors of C:P, and total nitrogen was the primary predictor of N:P. In the mixed conifer broadleaf forest, total organic carbon became the primary predictor of C:P and C:N ratios, while nitrogen fractions strongly regulated N:P variation. In the evergreen broadleaved forest, C:P was co-regulated by total P, available Al, TOC, and nitrate nitrogen, while N:P was predominantly governed by total P and total N. The explanatory power of environmental variables increased markedly along the restoration gradient, and nutrient heterogeneity in the 0-20 cm soil layer intensified during later restoration stages.Mixed‑effects models confirmed that forest type × depth interactions significantly modulated Al, Fe, and P stoichiometry, corroborating the successional shift from physical to biogeochemical controls. These results indicate that forest restoration progressively strengthens plant-soil nutrient associations and corresponds to a transition from available K/Al and bulk density constraints in the early-stage plantation toward absolute nutrient pools and organo-mineral interactions in the mature forest.

Concerns regarding the potential for artificial intelligence to cause harm to humans have intensified. Since approximately 2020, various organizations-including governments, think tanks, and big tech companies-have sequentially enacted artificial intelligence ethical principles. However, prior studies on these principles exhibit several limitations: First, previous studies failed to provide an integrated framework of artificial intelligence ethical principles; second, they lack a consensus on the prioritization of these principles; third, five years have passed since 2020 when the artificial intelligence ethical principles were enacted, but they have not been updated; and fourth, there has been a lack of strategic consideration regarding the relative importance and performance of these principles. To address this gap, this study extracted 17 artificial intelligence ethical principles by synthesizing 37 representative previous studies. Universal priorities were determined based on a comparative analysis of these principles' reference rankings. By evaluating the shifts in these ranking before and after 2020, this study identifies three principles-transparency, robustness, prohibition of infringement-that should be continuously adopted in artificial intelligence system development, and ive principles- responsibility, wellbeing, human rights, controllability, and sustainability-require greater emphasis and implementary development. the existing 10 Korean AI ethical principles are proposed to be expanded to 14. The results of the Importance-Performance Analysis (IPA) indicate that social value principles, such as respect of diversity, human rights, transparency, common good, and solidity, which demonstrated high performance levels, should be continuously integrated into development of artificial intelligence systems. Moreover, it is urgent to establish a governance framework to ensure accountability for responsible artificial intelligence and to develop systems for safe artificial intelligence usage, focusing on privacy, safety, prohibition of infringement, and data management. This study contributes to the foundation for building responsible, trustworthy, and safe AI systems.