299645 research outputs found
Sort by
Economic crises and happiness: Empirical insights from 134 countries (2008–2019)
No full textThis study examines the relationship between economic crises and subjective well-being, measured as happiness, using a balanced panel of 134 countries from 2008 to 2019. It identifies and differentiates three main types of crises—banking, currency, and sovereign debt—to capture the heterogeneous mechanisms through which macroeconomic shocks influence everyday life. Drawing on an extensive range of socio-economic, political, and demographic covariates, the analysis employs fixed-effects models and incorporates lagged variables, ensuring careful consideration of temporal dynamics. Additional robustness checks, including alternative dependent variables, two-stage least squares, and panel-corrected standard errors, confirm that the results do not hinge on modeling choices. The findings indicate that while economic crises broadly diminish happiness, their intensity, persistence, and immediacy vary by crisis type. Banking and currency crises impose substantial and relatively swift declines in subjective well-being, with banking crises leaving particularly enduring marks. By contrast, sovereign debt crises exhibit weaker and more delayed effects, suggesting that institutional factors, redistributive policies, and differing welfare regimes mediate their long-term impact. The heterogeneity analysis further reveals that developing countries experience more profound and widespread reductions in happiness, underscoring the role of structural inequalities and weaker social protections. The study’s insights inform policies aiming to enhance resilience, foster inclusive growth, and safeguard well-being amid economic volatility.</p
Global greenhouse gas reconciliation 2022
No full textIn this study, we provide an update on the methodology and data used by Deng et al. (2022) to compare the national greenhouse gas inventories (NGHGIs) and atmospheric inversion model ensembles contributed by international research teams coordinated by the Global Carbon Project. The comparison framework uses transparent processing of the net ecosystem exchange fluxes of carbon dioxide (CO2) from inversions to provide estimates of terrestrial carbon stock changes over managed land that can be used to evaluate NGHGIs. For methane (CH4), and nitrous oxide (N2O), we separate anthropogenic emissions from natural sources based directly on the inversion results to make them compatible with NGHGIs. Our global harmonized NGHGI database was updated with inventory data until February 2023 by compiling data from periodical United Nations Framework Convention on Climate Change (UNFCCC) inventories by Annex I countries and sporadic and less detailed emissions reports by non-Annex I countries given by national communications and biennial update reports. For the inversion data, we used an ensemble of 22 global inversions produced for the most recent assessments of the global budgets of CO2, CH4, and N2O coordinated by the Global Carbon Project with ancillary data. The CO2 inversion ensemble in this study goes through 2021, building on our previous report from 1990 to 2019, and includes three new satellite inversions compared to the previous study and an improved managed-land mask. As a result, although significant differences exist between the CO2 inversion estimates, both satellite and in situ inversions over managed lands indicate that Russia and Canada had a larger land carbon sink in recent years than reported in their NGHGIs, while the NGHGIs reported a significant upward trend of carbon sink in Russia but a downward trend in Canada. For CH4 and N2O, the results of the new inversion ensembles are extended to 2020. Rapid increases in anthropogenic CH4 emissions were observed in developing countries, with varying levels of agreement between NGHGIs and inversion results, while developed countries showed a slowly declining or stable trend in emissions. Much denser sampling of atmospheric CO2 and CH4 concentrations by different satellites, coordinated into a global constellation, is expected in the coming years. The methodology proposed here to compare inversion results with NGHGIs can be applied regularly for monitoring the effectiveness of mitigation policy and progress by countries to meet the objectives of their pledges. The dataset constructed for this study is publicly available at https://doi.org/10.5281/zenodo.13887128 (Deng et al., 2024).</p
AgentTrek: Agent Trajectory Synthesis via Guiding Replay with Web Tutorials
No full textGraphical User Interface (GUI) agents can automate complex tasks across digital environments, but their development is hindered by the scarcity of high-quality trajectory data for training. Existing approaches rely on expensive human annotation, making them unsustainable at scale. We propose AgentTrek, a scalable data synthesis pipeline that generates web agent trajectories by leveraging publicly available tutorials. Our three-stage method: (1) automatically harvests and filters tutorial-like texts from the internet using a specialized classification model, (2) transforms these texts into structured task specifications with step-by-step instructions, and (3) employs a visual-language model (VLM) agent to execute these instructions in real environments, while a VLM-based evaluator verifies trajectory correctness. The synthesized trajectories encompass multiple modalities, including text-based HTML observations with function-calling API actions, and vision-based screenshot observations with pixel-level actions. This multimodal data, enriched with chain-of-thought reasoning, enables agents to achieve state-of-the-art performance on both textual web browsing benchmarks (e.g., WebArena) and visual web grounding and browsing benchmarks (e.g., ScreenSpot Web and Multimodal Mind2Web). Furthermore, our fully automated approach significantly reduces data collection costs, achieving a cost of just $0.55 per high-quality trajectory without human annotators. Our work demonstrates that guided replay using web tutorials is a practical and scalable strategy for training advanced GUI agents, paving the way for more capable and autonomous digital assistants.</p
Evaluating cost-effectiveness of 9-valent HPV vaccination for men who have sex with men by HIV status in Hong Kong
No full textHuman papillomavirus (HPV) is the most common sexually transmitted infection and a leading cause of anal cancer and genital warts, particularly among men who have sex with men (MSM). In Hong Kong, HPV vaccination is currently only offered to school-aged girls, despite the high burden of HPV-related diseases among MSM, especially those living with HIV. Existing cost-effectiveness evaluations of HPV vaccination in Hong Kong primarily focus on female-only strategies and heterosexual men without accounting for differences in HPV infection risks among MSM. In contrast, countries such as the United Kingdom initially implemented free HPV vaccination for MSM through sexual health clinics, followed by an expansion of the programme to include adolescent boys. This provides a potentially useful model for Hong Kong to consider. This study aimed to evaluate the cost-effectiveness of implementing 9-valent HPV (9vHPV) vaccination strategies among HIV-positive and HIV-negative MSM in Hong Kong.</p
Predicting Urban Vitality at Regional Scales: A Deep Learning Approach to Modelling Population Density and Pedestrian Flows
No full textHighlights: What are the major findings? The UVPN model’s innovative architecture—integrating SE block and RCA bottleneck—effectively captures intricate spatial relationships and feature interdependencies, surpassing conventional deep learning models in urban vitality prediction. Static and dynamic urban vitality are shaped by distinct spatial features: macro-scale road networks influence regional residential patterns, micro-scale streetscape elements drive localized pedestrian activity, and meso-scale factors such as built density and POI distribution influence both—highlighting the multi-layered nature of urban vibrancy. What are the implications of the main findings? The model’s ability to produce fine-grained, dual-dimensional vitality maps helps uncover how different scales of urban form—from regional infrastructure to local design—affect where and how people live and move. UVPN provides urban planners and policymakers a powerful tool for evidence-based decision-making, supporting the design of targeted interventions at multiple spatial scales to create more sustainable, functional, and livable cities. Understanding and predicting urban vitality—the intensity and diversity of human activities in urban spaces—is crucial for sustainable urban development. However, existing studies often rely on discrete sampling points and single metrics, limiting their ability to capture the continuous spatial distribution of urban vibrancy. This study introduces the UVPN (urban vitality prediction network), a novel deep-learning architecture designed to generate high-resolution predictions of static and dynamic vitality at regional scales. The architecture integrates two key innovations: a SE (squeeze-and-excitation) block for adaptive feature recalibration and an RCA (residual connection with coordinate attention) bottleneck for position-aware feature learning. Applied to New York City, UVPN leverages diverse urban morphological features such as streetscape attributes and land use patterns to predict continuous vitality distributions. The model outperforms existing architectures, achieving reductions of 34.03% and 38.66% in mean squared error for population density and pedestrian flow predictions, respectively. Feature importance analysis reveals that road networks predominantly influence population density, while streetscape features strongly affect pedestrian flows, with built density and points of interest contributing to both dimensions. By advancing urban vitality prediction, UVPN provides a robust framework for evidence-based urban planning, supporting the creation of more sustainable, functional, and livable cities
Environmental Justice in the 15-Minute City: Assessing Air Pollution Exposure Inequalities Through Machine Learning and Spatial Network Analysis
No full textHighlights: What are the main findings? The study uncovers significant socioeconomic and racial disparities in air pollution exposure within 15-minute walking ranges, with lower-income and Black communities facing higher PM2.5 levels. Traditional exposure assessments often underestimate disparities by failing to account for residents’ daily mobility patterns and activity spaces. What is the implication of the main finding? The findings emphasize the need for dynamic, accessibility-based environmental justice assessments that reflect real-world mobility and exposure patterns. The study calls for tailored urban planning strategies to address localized pollution inequalities and ensure equitable outcomes in 15-minute city initiatives. The intersection of environmental justice and urban accessibility presents a critical challenge in sustainable city planning. While the “15-minute city” concept has emerged as a prominent framework for promoting walkable neighborhoods, its implications for environmental exposure inequalities remain underexplored. This study introduces an innovative methodology for assessing air pollution exposure disparities within the context of 15-minute activity zones in New York City. By integrating street-level PM2.5 predictions with spatial network analysis, this research evaluates exposure patterns that more accurately reflect residents’ daily mobility experiences. The results reveal significant socioeconomic and racial disparities in air pollution exposure, with lower-income areas and Black communities experiencing consistently higher PM2.5 levels within their 15-minute walking ranges. A borough-level analysis further underscores the influence of localized urban development patterns and demographic distributions on environmental justice outcomes. A comparative analysis demonstrates that traditional census tract-based approaches may underestimate these disparities by failing to account for actual pedestrian mobility patterns. These findings highlight the necessity of integrating high-resolution environmental justice assessments into urban planning initiatives to foster more equitable and sustainable urban development
Ritual, Community, and Democracy: Critical Reflections on Chenyang Li’s Reshaping Confucianism: A Progressive Inquiry
No full textShifting tides: recent advances in island and coastal human bioarchaeology
No full textThis talk examines the evolving landscape of human bioarchaeology, focusing on osteological and palaeopathological studies of coastal and island populations surrounding the Indian Ocean, Southeast Asia, and East Asia. Over the past fifteen years, significant advances have transformed our understanding of mobility, diet, physical activity, and health among ancient communities. By collating recent literature, key findings will illustrate: a) how ancient DNA analysis has complicated our knowledge of patterns in the region (e.g., Roopkund Lake), b) the importance of careful diagnosis regarding infectious diseases (e.g., treponemal diagnoses in Southeast Asia and signs of leprosy from Balathal), c) case studies of the bioarchaeology of care for disabled individuals (e.g., Metal Period Philippines), and d) the relationships between burial practices and embodied experiences among groups across the Pacific and Indian Oceans.Human palaeopathology (and bioarchaeology overall) often overlooks island and coastal populations, due to biases favouring continental studies. This neglect results in incomplete health narratives and underrepresentation of unique patterns of health, disease and development. Emphasising collaboration, diverse perspectives and a decolonial approach are essential to addressing these gaps. Together, we may enrich our understanding of human evolution among these marginalised yet historically significant communities</p
Quantifying the Optimization and Generalization Advantages of Graph Neural Networks Over Multilayer Perceptrons
No full textGraph neural networks (GNNs) have demonstrated remarkable capabilities in learning from graph-structured data, often outperforming traditional Multilayer Perceptrons (MLPs) in numerous graph-based tasks. Although existing works have demonstrated the benefits of graph convolution through Laplacian smoothing, expressivity or separability, there remains a lack of quantitative analysis comparing GNNs and MLPs from an optimization and generalization perspective. This study aims to address this gap by examining the role of graph convolution through feature learning theory. Using a signal-noise data model, we conduct a comparative analysis of the optimization and generalization between two-layer graph convolutional networks (GCNs) and their MLP counterparts. Our approach tracks the trajectory of signal learning and noise memorization in GNNs, characterizing their post-training generalization. We reveal that GNNs significantly prioritize signal learning, thus enhancing the regime of low test error over MLPs by Dq− 2 times, where D denotes a node’s expected degree and q is the power of ReLU activation function with q> 2. This finding highlights a substantial and quantitative discrepancy between GNNs and MLPs in terms of optimization and generalization, a conclusion further supported by our empirical simulations on both synthetic and real-world datasets.</p