154092 research outputs found
Sort by
Causal Cartographer:From Mapping to Reasoning Over Counterfactual Worlds
Get PDFCausal world models are systems that can answer counterfactual questions about an environment of interest, i.e. predict how it would have evolved if an arbitrary subset of events had been realized differently. It requires understanding the underlying causes behind chains of events and conducting causal inference for arbitrary unseen distributions. So far, this task eludes foundation models, notably large language models (LLMs), which do not have demonstrated causal reasoning capabilities beyond the memorization of existing causal relationships. Furthermore, evaluating counterfactuals in real-world applications is challenging since only the factual world is observed, limiting evaluation to synthetic datasets. We address these problems by explicitly extracting and modeling causal relationships and propose the Causal Cartographer framework. First, we introduce a graph retrieval-augmented generation agent tasked to retrieve causal relationships from data. This approach allows us to construct a large network of real-world causal relationships that can serve as a repository of causal knowledge and build real-world counterfactuals. In addition, we create a counterfactual reasoning agent constrained by causal relationships to perform reliable step-by-step causal inference. We show that our approach can extract causal knowledge and improve the robustness of LLMs for causal reasoning tasks while reducing inference costs and spurious correlations
Co-Designing a Multidomain Digital Toolkit to Support Cognitive Health in the Aging Dutch Population:Schouderklopje
Get PDFThe increasing prevalence of cognitive decline among older adults, coupled with the limited availability of accessible treatments, underscores the need for self-managed digital interventions to help mitigate the risk factors associated with cognitive decline. In that regard, this study explores the perceived challenges, motivations, needs, and key design features to inform the development of Schouderklopje, a user-centered, self-managed multidomain digital toolkit, from the perspective of domain experts. Challenges such as evolving and dynamic personal circumstances, limited awareness of brain health, and a lack of resources or support after the evaluation phase can affect the adoption of Schouderklopje. The motivation to adopt such lifestyle interventions was from a desire to improve overall health and social contribution. Furthermore, to ensure the toolkit is both self-managed and user-centered, importance of incorporating features supporting autonomy, personalization, education, skill development, and trustworthiness was noted.</p
In the Eye of the Beholder: Explaining the Effects of Local Participatory Budgeting on General Public Political Support
Get PDFGovernments increasingly combine deliberative and direct-democratic instruments to address democratic dissatisfaction and declining political trust. However, how the general public evaluates these hybrid democratic innovations in relation to their political support remains largely unexplored. This study examines the mechanisms through which a hybrid form of participatory budgeting (PB) in Amsterdam influences public political support. In this hybrid PB model, large-scale public voting complements traditional small-scale deliberation in allocating public funds. Drawing on systems theory, this study first theorises how public evaluations of hybrid PB input, throughput, and output components relate to shifts in their levels of political support. The analysis of 23 semi-structured interviews with citizens reveals that increased political support frequently depends on established mechanisms of feeling heard and perceiving direct policy influence. These increases also depend on a set of newly identified mechanisms, including issue salience, budget adequacy, throughput effectiveness, democratic representativeness, equitable participation, and transparent communication. Furthermore, pre-existing perspectives on political engagement and government constrain these potential positive effects. Finally, ambiguity in citizens’ evaluations of hybrid PB throughput and output introduces another underexplored pathway. The central finding is that the public interprets the same hybrid PB differently, resulting in varied impacts on their political support
Multi-temporal mineral mapping in two torrential basins using PRISMA hyperspectral imagery
Get PDFThe Sierra Minera de Cartagena-La Unión, located in southeast of the Iberian Peninsula, has been significantly impacted by historical mining activities, which resulted in environmental degradation, including acid mine drainage (AMD) and heavy metal contamination. This study evaluates the potential of PRISMA hyperspectral imagery for multi-temporal mapping of AMD-related minerals in two mining-affected drainage basins: Beal and Gorguel. Key minerals indicative of AMD—iron oxides and hydroxides (hematite, jarosite, goethite), gypsum, and aluminium-bearing clays—were identified and mapped using band ratios applied to PRISMA data acquired over five dates between 2020 and 2024. Additionally, Sentinel-2 data were incorporated in the analysis due to their higher temporal resolution to complement iron oxide and hydroxide evolution from PRISMA. Results reveal distinct temporal and spatial patterns in mineral distribution, influenced by seasonal precipitation and climatic factors. Jarosite was predominant after torrential precipitation events, reflecting recent AMD deposition, while gypsum exhibited seasonal variability linked to evaporation cycles. Goethite and hematite increased in drier conditions, indicating transitions in oxidation states. Validation using X-ray diffraction (XRD), laboratory spectral curves, and a larger time-series of Sentinel-2 imagery demonstrated strong correlations, confirming PRISMA’s effectiveness for iron oxides and hydroxides and gypsum identification and monitoring. However, challenges such as noise, striping effects, and limited image availability affected the accuracy of aluminium-bearing clay mapping and limited long-term trend analysis
A physically consistent dataset of water-energy-carbon fluxes across the Soil-Plant-Atmosphere Continuum
Get PDFHight-quality and Long-term measurements of land-atmosphere fluxes are vital for climate monitoring and Land Surface models (LSMs) benchmarking. Eddy covariance provides key in-situ data for theory and LSMs evaluation, but most flux towers lack continuous soil-plant-atmosphere measurements. Here, we present a long-term global dataset of water, energy and carbon fluxes, along with the corresponding above and below-ground hydrological, photosynthetic, and radiative data derived from the STEMMUS-SCOPE model simulations at 170 sites. In-situ observed fluxes data from PLUMBER2 and soil moisture (SM) data from FLUXNET2015 are employed to validate the effectiveness of the STEMMUS-SCOPE dataset. Results demonstrate that, without site-specific model tuning or calibration, and driven solely by global parameters and forcing datasets, simulated net radiation, latent heat flux, sensible heat flux, gross primary production, net ecosystem exchange, and SM datasets consistently agree with available in-situ measurements (median KGE: −0.03 to 0.80; median R2: 0.46 to 0.97; median rRMSE: 4.09% to 29.11%). This dataset supplements the existing ecosystem flux and SM network, enhancing our understanding of ecosystem functioning
Why we should stress about stress scores: issues and directions for wearable stress-tracking technology
Get PDFStress-tracking technology is becoming widely available and accessible, offering non-invasive ways to monitor and regulate stress. Yet, current stress-tracking technology presents stress scores that lack transparency, informativeness, and nuance. This provocation paper discusses five key issues that the current stress-tracking technology should overcome, involving conceptualization, measurements, transparency, interpretation, and responsibility. Next, we provide three directions to overcome these issues and inform better stress-tracking practices and future research. Future stress technology should improve stress response measurements, use better terminology and data visualizations, and increase user involvement and transparency. We conclude that there is a pressing need for designers and researchers to take greater responsibility in creating stress-tracking technologies that are useful, fair and just, and centered around individuals’ needs
Challenges and approaches to design serial production of hydrogen electrolyzers
Get PDFThe evolution from fossil fuels to hydrogen energy plays a vital role in the reduction of greenhouse gas emissions for a sustainable future. The increasing use of hydrogen will require scaling up the production of large electrolyzer. However, significant challenges in serial production of large electrolyzers exist that are capable of generating power in the megawatt range. This is due to factors such as, large size of components, constraints in weight and uncertain manufacturing and assembly processes. Moreover, lack of established production concepts affects the development of future manufacturing systems and limits potential automation possibilities for serial production based on the challenges in the current scenario. Thus, a strategic approach must be explored to design a production configuration for manufacturing electrolyzer components, cell and stack assembly with automation strategies aimed to scale-up the electrolyzer production. This paper proposes a generic methodology for Digital Factory Planning and Automation Strategies applicable across various manufacturing domains. This investigates viable methodologies utilizing tools like Value Stream Mapping (VSM) to map production flows for electrolyzer components, taking into account feasible manufacturing processes, technologies involved, key performance indicators (KPIs), which influence the cycle times. It also explores potential automation approaches, considering assembly processes, part-handling solutions, and tooling options. Finally, future steps are proposed, including virtual factory simulations and automated assembly solutions for a specific use case.</p
Environmental Assessment and Improvement of Factory Building Designs based on Generative Artificial Intelligence
Get PDFThe paper explores an innovative approach to evaluate the environmental impact of factory buildings at early design stages. Generative design, a cutting-edge computational technique, is employed to generate multiple factory building design alternatives based on user and case specific boundary conditions, e.g. related to material flow and space restrictions. This paper aims to integrate generative design principles with environmental assessment metrics to improve factory buildings for minimal environmental footprint, e.g. driven through energy demand. Thus, a framework that combines the generative factory design approach with key environmental assessment parameters is introduced. The effectiveness of generative design in enhancing the environmental performance of factory buildings is demonstrated with a case study. A comparative analysis of different designs highlights main influencing factors, as well as trade-offs and synergies between different manufacturing system performances and environmental oriented objectives. With that, the paper underlines the value of generative design as a transformative tool in sustainable factory design and provides actionable insights for architects, engineers, and policymakers aiming to develop greener industrial facilities.</p
High-Resolution prediction of soil pH in European temperate forests using Sentinel-2 and ancillary environmental data
Get PDFSoil pH is a key indicator for understanding soil health status in forested ecosystems, yet high-resolution mapping of this variable, especially at a 30-m spatial resolution, remains limited. This study uses Sentinel-2 spectral data, in-situ soil pH measurements, topsoil physical properties from the Land Use/Cover Area Frame Survey (LUCAS) database, and elevation data to estimate soil pH across temperate forests in Europe using a Random Forest model. Despite challenges in signal penetration due to forest canopy cover, the model achieved high prediction accuracy (R² = 0.62) at 30 m resolution. Bulk density, available water capacity, and clay content were the most influential physical predictors, while Sentinel-2 bands, particularly SWIR (1.610 and 2.190 μm), NIR (0.842 μm), and red-edge (0.705 and 0.783 μm), captured key vegetation responses related to soil acidity. Spatial analysis showed higher model accuracy in central and southern Europe, with reduced performance in Scandinavia, likely due to more acidic soils and extreme seasonal variation. The model also revealed significant pH differences among forest types, with deciduous forests showing the highest values and coniferous the lowest. These findings demonstrate the potential of high-resolution remote sensing data for monitoring soil pH, supporting forest management, biodiversity conservation, and climate adaptation strategies
