The upcoming EU Digital Identity (EUDI) Wallet aims to reshape identification and authentication across the EU. Users will receive attestations of identity data from issuers, store them in their wallet, and present them to relying parties. To prevent illegitimate access to sensitive attestation data, the EUDI framework introduces access control policies attached to attestations by their issuers; so-called embedded disclosure policies (EDPs). However, the current EDP framework is inflexible and restricted to simple whitelists and roots-of-trust logic. We improve on the EUDI’s EDP concept, adopting ideas from the literature to build a more dynamic policy-based access control mechanism. We introduce a generic model for flexible, attribute-based access control. We then instantiate such a scheme using DCQL, a query language already established in the EUDI ecosystem. Finally, we also implement and integrate this approach into a real-world EUDI wallet, demonstrating its feasibility.
Cross-border electronic authentication in the European Union is enabled by the eIDAS framework, allowing users to access services across Member States using their national electronic identities. Many services require linking authenticated foreign identities to existing national identity records, making identity matching a critical but insufficiently studied step in practical deployments. This paper presents a systematic analysis of identity matching based on long-term operational experience in an Austrian eGovernment system. It introduces a taxonomy of challenges covering identifiers, attribute representations, semantics, register quality, operational processes, lifecycle and persistence, user interaction, and governance constraints, and evaluates how these challenges are addressed by the revised eIDAS framework. The analysis shows that while the framework strengthens governance, transparency, and accountability, it does not resolve core challenges related to identifiers, attributes, semantics, and data quality. At the same time, eIDAS2 increases the complexity of identity matching through greater reliance on attributes, fragmentation of identity data, and higher operational demands. These findings indicate that reliable identity matching cannot be ensured through regulatory measures alone but requires complementary technical and organisational approaches.
Generative Artificial Intelligence (GenAI) is increasingly seen as a transformative technology for public sector organisations, with the potential to improve administrative efficiency, enhance service delivery, and support communication-intensive work. However, emerging evidence suggests that the adoption of GenAI in public administration is often constrained by organisational and institutional dynamics that are not yet well understood. This study examines how organisational friction emerges and affects the adoption of GenAI in a public sector organisation in Kenya. Drawing on an exploratory qualitative case study based on semi-structured interviews, the paper identifies four interrelated domains of organisational friction: bureaucratic and governance friction, infrastructural and data friction, professional and operational friction, and accountability and legitimacy friction. The findings show that these frictions do not simply block adoption, but shape how GenAI initiatives are contained, adapted, delayed, or selectively implemented within the organisation. Building on these insights, the study develops the AI–Bureaucracy Friction Model, which conceptualises GenAI adoption as a process filtered through interacting organisational constraints. The model provides a process-oriented explanation of why AI adoption in public administration often results in partial, cautious, or internally focused implementation rather than full-scale deployment. By offering context-sensitive empirical insight from a Kenyan public sector setting, the study contributes to a more nuanced understanding of AI adoption in bureaucratic organisations and highlights the importance of aligning innovation with institutional logics of accountability, legitimacy, and governance.
Voting Advice Applications (VAAs) have become essential e-democracy tools, helping citizens navigate the political landscape by providing personalized voting recommendations. Despite this, their development, particularly statement formulation, remains resource-intensive and prone to bias. Statement salience, defined as the ability of a statement to distinguish between parties, is critical for generating meaningful voting advice, yet no tools currently assist developers in selecting salient statements. This study addresses this gap by leveraging machine learning and natural language processing to predict statement salience. Using a dataset of 71 German VAAs, ten classification models were trained, including Random Forest, MLP, XGBoost, LightGBM, fine-tuned BERT models, and three large language models. Results indicate that RF and BERT perform best, achieving F1 scores of up to 0.71. Explanatory analyses show models correctly identify salient topics like Climate and Immigration, distinguish non-salient ones, and highlight key words driving salience. These tools can help VAA developers choose better statements, improving VAA accuracy and usefulness.
Comparing copyright rules across jurisdictions remains a time-consuming manual task. Questions regarding protection duration, economic rights, and exceptions for AI training data demand weeks of expert work and are difficult to scale to WIPO’s 194 member states. We present a symbolic, proof-of-concept pipeline that operates on an Akoma Ntoso 3.0 (AKN-XML) corpus of 28 WIPO IP treaties and nine national copyright laws. Using keyword filtering and regular expressions, we extract structured information for three core dimensions: term of protection (Q1), types of economic rights (Q2), and personal-use exceptions (Q3). The extracted data is represented as knowledge graph triples and formalized as defeasible deontic rules. The pipeline returns deterministic outputs within a reasonable time, and every output can be traced back to the exact legal rule and version. For copyright duration, the extracted terms agree with a manually compiled reference table across all nine jurisdictions; results for economic rights and, in particular, exceptions remain more exploratory.
Rural development is underrepresented in e-Government research, where planning decisions rely on fragmented data, opaque tools, and limited explanatory support. We present a framework that unifies heterogeneous geospatial and socio-economic data to generate verifiable, human-readable explanations for rural policy analysis through an auditable pipeline combining indicator computation, spatial aggregation, and fact-grounded language generation. We introduce a formal model that constrains explanation generation through fact selection, enabling quantifiable coverage, utility, and factual precision. Evaluation across six European regions shows (i) stable composite indicators under up to 30
Hungary is switching to m-government by legislation. The growing use of smartphones instead of personal computers for accessing e-government services raises security issues in the user environment, an area rarely considered. In the paper, a simple analysis framework is described for security-level classification based on the trustfulness of the environment and the protection of the access device. The comparison of the security features of smartphones and personal computers in the presented use cases does not support the common belief that switching to a smartphone is worthwhile because it offers better security.
Data sovereignty has become a central concern in public-sector digital transformation, particularly with the rise of data-intensive technologies such as Generative Artificial Intelligence (GenAI). However, existing research has primarily focused on conceptual and regulatory dimensions, with limited empirical insight into how sovereignty shapes technology adoption in practice. This study examines how data sovereignty influences the adoption and use of GenAI in the public sector. Drawing on a qualitative case study based on 19 interviews across national governance and service delivery contexts, the findings show that data sovereignty operates as a socio-technical condition shaping organisational decision-making. It constrains implementation pathways and generates tensions related to control, compliance, infrastructure, and strategic autonomy. These findings shift the understanding of data sovereignty from a policy concern to an organisationally enacted condition that structures digital innovation. The study contributes to research on AI governance by demonstrating how sovereignty is operationalised in practice and how it shapes public-sector GenAI adoption in resource-constrained contexts.
This paper evaluates the operationalisation of explainability requirements under the European Union’s Artificial Intelligence Act (AI Act, Regulation (EU) 2024/1689). The study evaluates, under a legal informatics perspective, the consistency of current standardisation initiatives with the Act’s definitions, general objectives, and resilience to technological innovation. The research identifies critical structural gaps in explainability along the value chain by noting that linear information transfers from providers to deployers often fail to provide meaningful understanding for affected individuals. Furthermore, the paper highlights a significant challenge posed by emerging “agentic AI” systems, which challenge current standards that focus on static input-output mappings rather than autonomous, sequential decision-making. The analysis suggests that a static view of standardisation under the New Legislative Framework (NLF) risks technical obsolescence and may undermine the protection of fundamental rights. To ensure long-term regulatory effectiveness, the study advocates for modular, layered explainability architectures that can adapt to the evolving needs of diverse stakeholders across the system lifecycle.
E-government is the implementation of transforming public services through digital platforms. However, the issue of digital divide challenges the success of e-government in terms of inclusivity, particularly in developing countries where infrastructural and socioeconomic gaps are prevalent. This study investigates the influence of socioeconomic and geographical factors on e-government adoption behavior in a developing country. The data for this study were collected in Indonesia, a representative developing country. The research employed a survey method to collect the data. We have collected 457 responses across urban, semi-urban, and rural areas in Indonesia. The study analyzed several behavioral variables, including digital literacy, technology anxiety, technology affordance, social support, trust, and perceived value, to examine complex adoption behavior. Data were analyzed using nonparametric techniques, specifically descriptive statistics and the Kruskal-Wallis Test, to identify significant differences in e-government adoption behaviour across socioeconomic demographics. This study reveals that educational background and geographical characteristics play a critical role in impacting e-government adoption behavior. At the same time, we found that older users tended to have higher perceptions of digital literacy and trust than younger users. In contrast, gender had little impact on e-government adoption behaviors. The results suggest that government strategies should prioritize technical and educational support for citizens with lower digital literacy and those living in rural areas to enhance e-government inclusion and ensure the effective use of public digital investments.
This paper argues that the growing integration of artificial intelligence (AI) into e-government intensifies a structural misalignment between digital transformation and the capacities required to govern it. Artificial intelligence enables public-sector systems to scale, automate and interconnect faster than the legal, organizational, regulatory and cybersecurity networks needed to oversee them can develop. The paper introduces the concept of the AI pace gap to describe this pace asymmetry within e-government. It then identifies three reasons why AI makes pace misalignment more acute in public administration: 1) accelerated deployment without equivalent institutional redesign, 2) deeper interdependence across public systems and 3) expanded governance burden that exceeds cybersecurity in a narrow sense. The paper concludes that narrowing the AI pace gap requires anticipatory governance, including secure-by-design public infrastructures, adaptive oversight and coordinated institutional capability-building.
This work addresses the challenge of establishing systematic links between recitals and normative provisions in European Union legislation. Recitals, while non-binding, play a crucial role in legal interpretation as they express the rationale, objectives, and policy context underlying legislative norms, often reflecting soft-law influences and institutional negotiation processes. The proposed approach aims to automate the identification of relationships between recitals and articles during the legislative drafting phase within the LEOS (Legislation Editing Open Software) environment, ensuring that such links are represented in a machine-readable form compliant with the Akoma Ntoso (AKN-XML) standard. A key difficulty arises from the evolving nature of legislative texts, where recitals remain static while normative provisions may be amended over time, potentially leading to inconsistencies. To address this issue, the RECONTA (REcital CONnector To Articles) module formulates the task as a binary classification problem, determining whether a given recital–article pair is semantically related. The system employs a transformer-based architecture (ModernBERT-base), fine-tuned to capture contextual semantic relationships over extended legal texts. This enables scalable and automated support for maintaining interpretative coherence between justificatory and operative components of EU legislation.
As public administrations accelerate digital transformation, they become primary targets for state-sponsored cyber operations. The paper investigates the tension between e-government integration and the expanded attack surface it creates during geopolitical conflicts. By analyzing cyberattacks on public infrastructures, the study examines the cybersecurity perspective in developing resilient e-government systems. Using a comparative case study analysis, we examine four cases of state-sponsored cyberattacks: Estonia (2007), Georgia (2008), Albania (2022), and Ukraine (2022–Present). The analysis maps each country's UN E-Government Development Index (EGDI) at the time of the attack against its resilience capabilities. Results indicate a strategic evolution in attack patterns from early denial of service attacks to destructive wipers. The data shows that while high digital maturity increases the impact of service outages, decentralized models and sovereign cloud hosting ensure service survival under severe multi-domain attacks. The paper concludes that to advance smart governance, policymakers must prioritize resilient frameworks, integrating robust security approaches in e-government.
Artificial intelligence (AI) is transforming public service delivery and government decision making, yet countries differ significantly in their readiness for AI enabled governance. This study examines which World Development Indicators (WDI) explain variations in national AI related digital readiness using a proxy based composite framework covering multiple countries over two decades. Machine learning models were applied to identify the most influential predictors and classify readiness levels across nations. The findings highlight the central role of economic capacity, national development, urbanization, trade, health conditions, and macroeconomic stability in shaping AI readiness. Regional disparities remain evident, with more developed regions demonstrating stronger preparedness than developing regions. The study concludes that AI readiness is closely linked to broader socioeconomic development and provides policy insights for strengthening government and AI driven governance systems.
Smart city and digital twin concepts hold considerable potential for improving local governance and planning, yet practical implementations have almost exclusively focused on large metropolitan areas. Small towns and rural regions face a distinct set of challenges – limited data availability, constrained budgets and lower digital capacity – that make a direct transfer of urban approaches infeasible. This article presents results from the ongoing project “Smart Cities and Digital Twins in Lower Austria”, which adapts digital twin concepts for small-town and rural contexts following a design science research approach. Based on stakeholder workshops with municipalities in Lower Austria, three use cases were identified and implemented as prototype digital twins: traffic flow planning (emphasising data integration), population growth (emphasising simulation) and shared spaces (emphasising visualisation). We describe the design and implementation of each prototype and report on a technical evaluation of their feasibility. Our findings highlight both the potential and the specific limitations of digital twins in resource-constrained settings and offer practical guidance for municipalities, policymakers and researchers working in non-urban contexts.
This paper aims to understand how race and gender intersections shape women's experiences in smart cities, especially their perception of safety. The intersectional approach considers the embeddedness of power, privilege, differentiation, and systems of domination in urban spaces and technologies. Therefore, the approach helps comprehend inequalities in smart cities. This qualitative research was conducted based on semi-structured interviews with diverse women living in São Paulo. Based on the perspectives of women from the Global South, the study highlights the need to consider intersectionality in designing, implementing, monitoring, and evaluating smart city initiatives. The research contributes to understanding women's perspectives in the city and the role of smartness and technology.
Current e-government services for public health fail to fully address citizens’ needs due to several limitations. Aiming to augment the effectiveness of these services, this paper introduces a public health QA assistant that seamlessly integrates Knowledge Graphs, LLMs and Explainable AI techniques. The proposed approach builds on the power of LLMs and incorporates symbolic reasoning to improve citizen awareness of medical issues and enhance trust in health services. By automating public health QA and providing transparent, user-friendly explanations, the proposed assistant may foster citizen participation and understanding without requiring technical expertise, thus significantly increasing the social impact of public health services. This work also intends to reveal a series of insights about the way that modern AI-enhanced public health systems should be developed.
Democracy is based on dialogue/deliberation between equal citizens. Deliberative democracy enables citizens to participate in deliberations/discussion about issues affecting their lives. However, deliberations often face challenges related to low participation, exclusion, low-quality contributions and difficult moderation. AI potentially can address many of these challenges. For example, AI can facilitate deliberations by moderating discussions, processing deliberation data (e.g., summarizing) and enhancing participation. However, the integration of AI into deliberations also raises risks related to fairness and transparency. This paper presents a review of existing research that explores the role of AI in mass deliberations. The aim is to identify AI features that can enhance deliberations, exploring also potential benefits and risks. We anticipate that the findings can contribute to the ongoing debate on the future role of AI in political deliberations.
Online petition systems serve as essential crowd-sourcing mechanisms that enable governments to collect public suggestions on policy matters. Nonetheless, the dependence on manual petition review processes presents challenges for administrators of such systems, particularly in terms of extended response times and delayed feedback, which may discourage active public participation. This study presents an interpretable automated classification approach designed to assess the admissibility of petitions, thereby assisting administrative reviews by providing prompt and transparent feedback to petitioners. The proposed approach uses a fine-tuned BERT model, augmented by principal component analysis (PCA) for dimensionality reduction. The model is further optimized using an XGBoost classifier, to categorize petitions submitted to Taiwan’s JOIN platform into three administrative categories. To address interpretability concerns commonly associated with machine learning-based systems, two explainable AI methods are incorporated: BERT’s Self-Attention mechanism for global textual analysis and SHAP for localized word-level interpretability. Experimental results demonstrate that the developed system achieves a classification accuracy of 67
Public budget preparation is a multilayered complex process, requiring the reconciliation of diverse political, economic, and societal interests while managing large-scale financial data. In this study, we explore the potential of artificial intelligence (AI) to support and enhance public budget preparation, with a specific focus on Germany's state and municipal levels. Using an exploratory case study and a human-centered development approach, we identify key challenges in budget planning, including information overload, stakeholder communication difficulties, and last-minute adjustments. We investigate various AI-driven solutions, such as large language models, intelligent communication tools, and machine learning-based anomaly detection. Among these, anomaly detection received the most positive feedback from stakeholders, leading to the development of a prototype aimed at assisting decision-makers throughout the budgeting process. While our findings indicate that AI systems can provide valuable support in public budgeting, significant challenges remain, particularly regarding data availability and system integration. This study contributes to the ongoing discourse on AI applications in public administration, highlighting both the opportunities and limitations of AI-driven budget planning.
