This article introduces the global Women’s Safety Index (WSI), outlining its rationale, purpose, and potential applications. The Index consist of three dimensions: Equity, Protection, and Resources, identified as foundational to women’s safety. Key indicators within each dimension are selected based on theoretical relevance and empirical evidence. We detail the statistical methodology and framework used to construct the Index and present validation analyses demonstrating its ability to capture changes in women’s safety, particularly in response to external disruptions. The WSI is available on an interactive digital platform, enabling users to explore, visualize, and compare women’s safety data across regions and over time.

Gauging the extent of public acceptability of reforms is an important concern for policymakers. Timely insights into public perceptions can illuminate how reforms are received and how attitudes evolve over time. In this study, we build on the OECD’s Public Acceptability Tool, a framework encompassing four key dimensions of reform acceptability—Economic, Fairness, Behavioural, and Process—to evaluate the public acceptability of policy reforms. We take the 2023 French pension reform as a relevant case study, using online media articles and parliamentary speeches as indicators of discourse surrounding the reform. Using word embeddings, we classify these texts according to the four dimensions and apply matrix factorisation topic algorithms to uncover the latent themes within each. Our analysis shows that the Process dimension dominated media coverage during the discussion and legislative phases of the reform, consistent with previous literature on pension reforms. In contrast, no particular dimension was predominant in parliamentary speeches, suggesting a mismatch between policy and public debates. Finally, we identify the main topics driving public discussion within each dimension, highlighting notable differences between media narratives and parliamentary discourse that offer further insight into the dynamics of public acceptability.

Long-term unemployment (LTU) is a challenge for both jobseekers and public employment services. Statistical profiling tools are increasingly used to predict LTU risk. Some profiling tools are opaque, black-box machine learning (ML) models, which raise issues of transparency and fairness. The present paper investigates whether interpretable models could serve as an alternative, using administrative data from Switzerland. Traditional statistical, interpretable, and black-box models are compared in terms of predictive performance, interpretability, and fairness. It is shown that explainable boosting machines, a recent interpretable model, perform nearly as well as the best black-box models. It is also shown how model sparsity, feature smoothing, and fairness mitigation can enhance transparency and fairness with only minor losses in performance. These findings suggest that interpretable profiling provides an accountable and trustworthy alternative to black-box models without compromising performance.

The growing availability of information sources has offered central banks new opportunities to enhance their statistical, analytical, and policy functions. By linking—or integrating—various data sets, they have been able to produce more granular, timely, and diverse statistics in a cost-efficient way. These advancements have also enabled a better use of information available in society, such as administrative records, to improve statistical agility in responding to user needs. Yet integrating alternative data—often generated as a by-product of other processes—also raises challenges, including concerns over accuracy, representativeness, and reliability. This paper aims to review systematically the opportunities and limitations of data integration in central banks, taking stock of their experience thus far. Results underscore the need for strengthening the global statistical infrastructure through adequate data governance, management, and public resources.

The rapid proliferation of AI systems has outpaced regulatory and insurance frameworks, leaving risks from unpredictable rogue AI behaviors unaddressed. While academic debates prioritize existential threats, this article shifts focus to governing present-day AI through AI Risk Bonds: market-driven instruments inspired by catastrophe bonds. These bonds securitize AI-related liabilities, using investor scrutiny to price risks based on a system’s expected impact and behavioral predictability. By dynamically adjusting bond yields, higher risks escalate capital costs for developers, incentivizing proactive risk mitigation. The mechanism addresses regulatory blind spots via market oversight, disperses liability through capital markets, and reduces moral hazard by linking financing to risk profiles. Complementing initiatives like the EU AI Act, this framework balances innovation with precaution, tethering profitability to risk minimization for responsible AI development.

Gender data gaps in the United Kingdom are political choices that reflect deeper assumptions about whose labour, lives, and experiences are deemed worth measuring. This paper examines the strengths and limitations of the UK census for bridging these gaps, drawing on feminist participatory action research with 170 participants across 12 workshops in England, Wales, Scotland, and Northern Ireland. Participants comprised representatives from local government, women’s and non-profit organisations, academia, and unaffiliated individuals with varying statistical skills. Through co-produced inquiry, participants identified critical gender data gaps and questioned the political and epistemological assumptions embedded in census design. We find that while the UK census provide comprehensive disaggregated multivariate data at the local level, they systematically omit critical dimensions of gendered life, including income, unpaid childcare, and occupational segregation. Drawing on Fricker’s (2007) concept of epistemic injustice, we argue that these omissions are not incidental but structural: the census lacks the conceptual framework to recognise them as dimensions worth counting. Realising the census’s potential for feminist analysis, therefore, requires deliberate reshaping of data systems and practices.Crucially, we demonstrate that involving those affected by gender data gaps as co-analysts, rather than consultation subjects, surfaces blind spots and analytical priorities that academic or statistical institutions are unlikely to identify from within, making the case for embedding participatory mechanisms in how official statistics are designed, produced, and used. The paper concludes with four user-informed recommendations for census reform and the wider reimagining of UK gender data systems.

Blockchain technology has gained immense popularity in enhancing the security and privacy of Information Systems (IS), reflected in exponential increases in published research articles. The rapid proliferation of published research presents significant challenges for manual analysis and synthesis due to the vast volume of information. To this end, we adopted the Computational Literature Review (CLR) and Latent Dirichlet Allocation (LDA) topic modeling techniques to analyze the impact of the pertinent literature. This study is among the first studies in the IS field to apply the CLR technique in the Legal-Tech context, focusing on security, privacy, and governance in blockchain systems. Using CLR and LDA topic modeling, we identified 10 topics related to security and privacy, each accompanied by an in-depth description aligned with relevant policies and regulations. Our findings identify both the strengths and current gaps in the literature, providing valuable insights into ongoing scholarly discourse. Through our analysis, we outline future research directions that align with the evolving dynamics of blockchain technology, data management, and policy implications, aiming to guide further academic and practical advancements in the IS field.

This study focuses on a multidimensional evaluation of China’s artificial intelligence (AI) governance policies by employing the Policy Modeling Consistency (PMC) index model to systematically analyze the performance and coherence of these policies in terms of goal setting, technological attributes, risk governance, and socio-cultural institutional embedding. The study conceptualizes AI as a novel form of cultural and social technology rather than merely an intelligent agent. This perspective highlights AI’s role in reorganizing and restructuring information flow, cultural expression, and institutional operations. By constructing a PMC evaluation framework that integrates technological, risk, and socio-cultural dimensions, this research reveals both strengths and weaknesses in the current policy system, particularly significant gaps in embedding socio-cultural institutions. Combining theoretical and empirical analyses, the PMC model offers an effective framework for understanding the institutional dynamics of AI governance in China. The findings enrich theoretical perspectives on AI governance and provide empirical support for policy formulation.

Indonesia’s Tax Court faces a rapidly growing caseload and limited judicial capacity, leading to prolonged dispute resolution and eroding public trust. Manual review processes create inconsistencies that undermine legal certainty, particularly within Indonesia’s civil law system, where jurisprudence is nonbinding and a judge’s personal knowledge may serve as admissible evidence. To address these challenges, we introduce the AI-driven Decision and Integrity for Tax Court Law (ADIL) pipeline, which automates the analysis of 4377 Indonesian Tax Court decisions from 2008 to 2023. ADIL applies advanced natural language processing and legal informatics to extract structured insights from unstructured judicial texts through a trilayered architecture: (i) rule-based pattern matching for high-precision extraction, (ii) transformer-based named-entity recognition for robust entity detection, and (iii) local large-language-model prompting for semantic summarisation and classification. Using exclusively pre-decision metadata, 49 structured metadata fields, including dispute number, tax year, judge identity, and verdict, were used to train a model that resulted in several machine-learning models to predict case outcomes and resolution duration. ADIL achieved 97.5% precision and strong macro-F1 performance in outcome prediction, and an R2 of 92.9% (RMSE = 63.8%) for duration estimation, equivalent to 7–8 months of predictive accuracy. Notably, incorporating encoded judge identity improved accuracy by 11–16%, revealing systematic variation in judicial behaviour. Feature-importance analysis suggests that ADIL can substantially enhance efficiency, consistency, and transparency in tax adjudication. ADIL offers a data-driven decision support framework towards a more agile and accountable judicial system in Indonesia’s legal and institutional context.

Rapid developments in Artificial Intelligence (AI) tools and their ever-expanding use in societal contexts raises questions regarding their design, social acceptance, and regulation. We develop a randomized vignette experiment, detailing six contemporary high-stakes contexts employing algorithmic tools and uncover distinct attitudes toward the source of the AI provisioning (public vs private sector). Participants are drawn from technology and law schools in India, representing key future stakeholders in the AI policy ecosystem in the country. Our main result indicates an overall preference for public sector provisioning of AI tools, with important implications on the social acceptability of public sector expansion of AI systems. Although the Global South has witnessed widespread expansion of AI tools with limited scrutiny, the nascent scholarship on AI attitudes disproportionately emanates from the Global North. Our study fills this gap, as well as the gap in empirical evidence on private vs public provisioning of AI tools in social contexts.

This article explores the integration of large language models (LLMs) and AI research agents into global benchmarking frameworks, with a focus on data for the public good. Against a backdrop of shrinking funding and rising demand for scalable and reproducible assessments, we ask whether AI can assume core roles in indicator development, evidence discovery, and policy evaluation without compromising contextual nuance or democratic legitimacy. Building on pilot experiments conducted within the Global Data Barometer (GDB), we employed a phased, adaptive methodology that tested workflow-based platforms and deep research agents across tasks ranging from legal interpretation to multisource policy analysis. The preliminary findings suggest that while AI systems show strong potential for automating structured assessments, they falter on complex, fragmented, or normatively loaded indicators, raising concerns about opacity, overinterpretation, and inclusivity. To navigate these tensions, we propose a hybrid human-AI architecture that combines standardized workflows, adaptive agent capabilities, and critical human oversight. Central to this model is the concept of a dynamic evidence infrastructure, designed to embed participatory validation and enhance transparency. By reframing automation as augmentation, the study contributes both an empirical, domain-specific assessment of the opportunities and limits of AI-assisted benchmarking and a theoretical framework for sustainable, context-aware evaluation in the age of AI. We argue that the success of AI-assisted benchmarking should be measured not only in efficiency gains but also in its ability to strengthen legitimacy, accountability, and inclusiveness in data ecosystems worldwide.

As large language models (LLMs) increasingly shape decision-making, public discourse, education, healthcare, and governance, a critical question emerges: Whose values are these systems truly reflecting? While today’s generative AI systems demonstrate extraordinary capabilities, they also inherit hidden biases, ethical blind spots, and implicit assumptions embedded within their training data. Existing alignment approaches often remain opaque, resource-intensive, or insufficiently adaptive to diverse societal expectations.

This paper introduces a novel value-driven LLM framework designed to systematically uncover, quantify, and realign the implicit values embedded within LLMs toward socially desirable outcomes. Built upon the AI for Social Good (AIfSG) framework, our proposed methodology operationalizes ethical alignment across six critical domains: reasoning and interpretability, bias removal, transparency and accountability, security and privacy, moral and ethical observations, and public understanding. Using advanced embedding techniques, cosine-based value-difference metrics, and topic-weighted iterative fine-tuning, the framework transforms ethical alignment from an abstract aspiration into a measurable and actionable computational process.

To demonstrate adaptability across moral and regulatory paradigms, the framework evaluates two distinct reference value systems: the Ten Commandments and the General Data Protection Regulation (GDPR). Experimental results using open-source LLMs, including Llama 3.2 and Gemma 2, reveal substantial reductions in value misalignment, ranging from approximately 25% to 70% across ethical domains, while preserving model flexibility and scalability. Visualizations in value-embedding space further confirm significant convergence between original model outputs and socially aligned reference values after iterative realignment.

Beyond technical innovation, this work positions value alignment as a foundational challenge for the future governance of generative AI. The proposed framework functions both as a diagnostic instrument for identifying ethical gaps and as an intervention mechanism for adaptive value realignment, offering policymakers, developers, and institutions a scalable pathway toward transparent, accountable, and socially responsible AI systems. By bridging computational methods with moral, legal, and societal principles, this research advances a new paradigm for embedding human-centered values directly into the next generation of intelligent systems.

Democratic resilience is as much about the narratives of our nation we affirm, as the institutions that enshrine our values and laws, a fact re-affirmed by scholarship across many branches of social science in recent decades. For policymakers and quantitative social scientists, analysing or tracking public discourse through the lens of narrative and framing has historically involved the annotation of texts by hand, placing severe limitations on the scale and modality of discourse under inquiry. Yet, a revolution is at hand—a transformer revolution: first arising in computer science, and now enabling a new kind of automated narrative analysis at scale, transformers are opening up new horizons for the tracking of public narratives of democratic resilience. Here, we: formulate a conceptual framework linking computational language methods to democratic resilience analysis; introduce transformer-based artificial intelligence (AI) methods as a third wave in natural language processing technology; and demonstrate two practical applications of transformer methods to democratic narrative analysis. Finally, we conclude by contributing data and research recommendations which flow naturally from the opportunities unlocked by transformer methods for public stakeholders who wish to see these opportunities realised. Together, we suggest that, perhaps for the first time, the “holy grail” of the quantitative social scientist is near: the ability to identify, accurately, and efficiently, nuanced narratives in text, at scale.

Indonesia’s rapid digital transformation generates massive economic opportunities while posing new regulatory challenges. This study examines how Indonesia’s digital economy is governed through key regulatory domains: electronic commerce, data governance, and digital taxation, and assesses their impact on inclusive growth. The research utilizes an integrated theoretical lens, Institutional Theory and Digital Divide Theory, to analyze how formal rules and on-the-ground access to technology together shape outcomes. The research employs a multimethod approach, to understand how different regulatory models influence market participation, business compliance, and digital inclusion. The findings reveal significant regional disparities in internet access and digital business adoption, with rural and remote areas lagging behind urban centers. Micro, small, and medium-sized enterprises face disproportionate compliance burdens under current regulations, a gap partly addressed through cooperative initiatives like the Indonesia–Japan Track 1.5 Public–Private Partnership. A key insight is the complex impact of data localization policies: while strengthening digital sovereignty and user privacy, strict localization requirements can raise operational costs and deter foreign investment. The paper concludes with policy recommendations for improving regulatory implementation, expanding digital infrastructure and literacy, and fostering international cooperation to ensure Indonesia’s digital economy regulations promote equitable growth across its diverse archipelago.

Citizen-generated data (CGD) is increasingly embraced as a strategy for filling data gaps to achieve the United Nations Sustainable Development Goals (SDGs), including SDG 5, Gender Equality and the Empowerment of Women and Girls. Existing frameworks to guide the design and use of CGD, however, do not reflect the unique considerations of CGD projects addressing issues of gender inequality. Answering recent calls for study of “data practices,” this article analyzes common CGD principles through findings from an action research project to address gender-based violence at the Colombia–Venezuela border. We suggest that while existing frameworks provide generative pathways forward, several principles around which consensus appears to be emerging are too rigid or insufficiently nuanced to account for the dynamics that many gender-focused CGD projects confront. These include dynamics such as physical security, the role of emotion in shaping project implementation, unequal access to resources, and political will. We suggest that this rigidity truncates the utility of these frameworks for CGD actors navigating highly sensitive issues in risky environments where serious violations of women’s human rights are taking place—and where the generation of gender data is one of several motivating factors for the work being done. This article reads gender into these frameworks to broaden the range of sustainable development issues for which CGD can help catalyze progress.

Web-enabled large language models (LLMs) frequently answer queries without crediting the web pages they consume, creating an “attribution gap” in responsible artificial intelligence (AI) usage—defined as the difference between relevant URLs read and those actually cited. Drawing on approximately 14,000 real-world LMArena conversation logs with search-enabled LLM systems, we document three exploitation patterns: (1) no search: 34% of Google Gemini and 24% of OpenAI GPT-4o responses are generated without explicitly fetching any online content; (2) no citation: Gemini provides no clickable citation source in 92% of answers; (3) high-volume, low-credit: Perplexity’s Sonar visits approximately 10 relevant pages per query but cites only three to four. A negative binomial hurdle model shows that the average query answered by Gemini or Sonar leaves about three relevant websites uncited, whereas GPT-4o’s tiny uncited gap is best explained by its selective log disclosures rather than by better attribution. Citation efficiency—extra citations provided per additional relevant web page visited—varies widely across models, from 0.19 to 0.45 on identical queries, underscoring that retrieval design, not technical limits, shapes ecosystem impact. To advance auditing and monitoring of AI systems, we recommend a transparent LLM search architecture based on standardized telemetry and full disclosure of search traces and citation logs.

Greenwashing poses a significant challenge to the fight against climate change by undermining trust in corporate sustainability claims. This study introduced the greenwashing tendency score (GTS), an automatable method designed to detect greenwashing tendencies in corporate sustainability reports. By leveraging textual sentiment and alignment analysis techniques in conjunction with environmental, social, and governance ratings, the GTS quantifies discrepancies between communicated and actual sustainability performance. We applied our methodology to 36 German stock index companies during the years from 2020 to 2022. Our key findings reveal substantial variations in greenwashing tendencies among these companies, emphasizing the need for more transparent and reliable sustainability reporting. The GTS emerged as a scalable, reproducible, and objective tool that can aid, for example, investors, regulators, and Non-government organizations in identifying greenwashing practices. This research contributed to the sustainable finance literature by introducing a neutral and open measure to assess firms’ greenwashing tendency, summarizing implications for policymaking and regulatory authorities and discussing its potential for long-term accountability and integrity in corporate sustainability communications.

As a major metropolitan city, London faces persistent road congestion and severe air pollution. To address these issues, static electronic road pricing (ERP) models have been implemented. While effective, these are inherently limited in flexibility. This paper explores dynamic ERP models to improve upon static pricing by minimizing air pollution and traffic congestion within the Congestion Charge Zone. The problem is formulated as a multi-stakeholder multi-objective optimization problem, incorporating the perspectives of three stakeholders—the government, vehicle owners, and environmental organizations—and three objectives: air pollution, traffic congestion, and price. The NSGA-II optimization algorithm was applied on a representative day and demonstrated substantial improvements. The concentration of PM$ {}_{2.5} $—the more harmful pollutant—was reduced by up to 23%, while NO2 levels fell by 2–3%. Traffic flow, used as a proxy for congestion, decreased by approximately 3–4% during peak hours. These improvements were achieved with only a modest increase in the mean price to £12.51 (from a baseline of £11.50), with a standard deviation of £1.59 and a variance of £2.43 across hourly prices. These results suggest that targeted dynamic pricing—when aligned with environmental and behavioural incentives—can deliver measurable gains in urban air quality and congestion without imposing a significant cost burden on drivers. A core novelty of this work lies in its practical, stakeholder-inclusive problem formulation. While the approach assumes infrastructure for automated price deduction and routing, this limitation can be addressed in future work through advances in vehicle–infrastructure communication systems.

The role of data and automated (non-artificial intelligence [AI]) algorithmic targeting in adaptive social cash systems is gaining increasing significance, but few governments have yet leveraged on AI technologies to reap its benefits. Hence, there is mounting pressure on social cash policymakers and practitioners to rapidly embrace the opportunities arising from AI applications, especially in times of crisis. While data and algorithmic targeting (non-AI and AI) are efficient in enrolling beneficiaries in emergency social cash systems, it may also pose serious challenges. Through a qualitative case study of an adaptive social cash programme in Pakistan, the research critically examines the data/algorithmic targeting process, and unveils the shortcomings prevalent in design, data and algorithmic decision-making that lead to certain exclusionary outcomes. The study makes several contributions to the data and policy literature. Drawing on the limitations, it first offers a set of practical recommendations for greater enrolment, and hence inclusion of beneficiaries. Second, it discusses novel opportunities that AI technologies may present in adaptive social cash systems, whilst carefully assessing the risks. Third, the study proposes an organisational AI governance framework to guide the development of responsible and ethical AI practices. The study affords policy and practical implications for governments, social cash policymakers, and practitioners in providing invaluable insights into how changing targeting practices, via AI technologies, under a governance framework can direct ethical practices that positively impacts on beneficiaries, social cash organisations, and stakeholders.