Discretion is a central concept in street-level bureaucracy, yet its role is evolving in the context of digital transformation in public administration. This paper examines how discretion is reflected in policy discourse and organizational practices within Dutch public service organizations. The study draws on focus group discussions with public employees from municipalities and national government agencies involved in policy development and digital transformation initiatives. The data were analyzed using qualitative text analysis supported by large language models. The findings show that discretion is shaped by the interaction between external mandates, organizational visions, and implementation practices. Legal frameworks, accountability mechanisms, and political priorities continue to define the boundaries within which discretion can be exercised. At the same time, organizations increasingly emphasize citizen-centered service delivery and the importance of professional judgment. However, translating these ambitions into practice remains challenging due to digital infrastructures, standardized procedures, and organizational silos. The study highlights how discretion in digitally transforming public administrations emerges from the interaction between legal, technological, and organizational factors.
Many OGD-based applications play a key role in making public data usable, yet most are short-lived because their sustainability hinges on the continued engagement and connections of independent developers. This study investigates how developer motivation and networking shape the long-term sustainability of such applications. Drawing on an exploratory case study of two mobile OGD-based applications that have been maintained for over a decade, we analyze the temporal evolution of motivational drivers and networking dynamics. Our findings show that sustainability depends not only on strong initial intrinsic motivation, but also on later obligation-based drivers, predictable maintenance demands, data-access conditions, and the changing strength of external ties. We derive six propositions that extend prior research by highlighting the temporal fragility of developer engagement and the conditional role of networking as a multiplier. Practically, the study offers insights for independent developers seeking to manage maintenance burdens and sustain long-term involvement in OGD-based initiatives.
Public ombudsman offices play a central role in digital government by providing institutional channels through which citizens submit complaints, suggestions, and service requests to public authorities. The growing volume of textual submissions creates operational challenges for the triage process, which requires accurate identification and routing of citizen demands. In many public organizations, this task remains manual, making it time-consuming and difficult to scale, particularly when labeled data for training automated systems are scarce. This study examines the use of Natural Language Processing (NLP) and Machine Learning (ML) techniques to support the automated classification of ombudsman submissions under conditions of limited data availability. To mitigate data scarcity, a data augmentation strategy based on synthetic text generation using a large language model was adopted, enabling the construction of synthetic and hybrid datasets to complement real submissions. Five classification algorithms were evaluated across three experimental settings: real-only, synthetic-only, and hybrid datasets. Results suggest that hybrid configurations outperform real-only data, although category differences and possible synthetic-data bias limit this comparison. These findings indicate that synthetic data may support ombudsman triage when combined with validation and human review.
Generative AI is rapidly diffusing across digital government operations, reshaping how administrative judgments are formed and validated. While existing AI governance research has focused on algorithmic transparency, accountability, and regulatory design, far less attention has been paid to how reliance on AI alters the judgment processes of public sector users and produces new forms of vulnerability. This study examines vulnerability in AI-mediated administrative environments not as the average effect of a single factor, but as patterns associated with the alignment of multiple conditions. We conceptualize Human Judgment Risk as the likelihood of inaccurate or insufficiently verified judgment under AI-mediated conditions, particularly when self-assessed capability diverges from observed task performance. The study focuses on three conditions: AI usage intensity, the Self–Actual Gap between self-assessed and observed performance, and institutional control environments. Using experimental data from 177 participants, we explore whether specific condition alignments are associated with concentrated risk patterns. Approximately 28
Public procurement is a central mechanism through which digital government is realized, yet research on how public organizations procure information systems remains fragmented within digital government scholarship. This scoping review maps research on public procurement of information systems within digital government and examines how prior studies conceptualize and analyze user needs, their translation into requirements, and their relation to later outcomes in use. Based on 14 studies identified through searches in Scopus, Web of Science Core Collection, and the Digital Government Reference Library, complemented by citation searching, the review shows that the literature provides comparatively strong insight into procurement processes, requirements work, and public–private relationships, but much weaker insight into how early procurement choices shape later implementation, use, and value-in-use. The review contributes a synthesis of a fragmented body of research on this topic and identifies a key research gap, deserving of more research interest: the limited traceability from needs and requirements to later outcomes in use.
