Both scholars and public officials have voiced concerns on the legitimacy of AI use within the public sector, as the introduction of AI within the public sector has led to a number of examples where citizens were negatively impacted and harmed. There is a growing body of literature aimed at getting a better understanding of legitimacy of public sector AI and algorithm use. In particular, literature identified that the use of externally developed AI applications for government decision-making can pose several risks to democratic legitimacy of government. However, the debate in the literature on this topic is still at a rather conceptual level and it is unclear what the extent is of these risks, and which type of AI applications tend to be developed externally. Empirical studies on this are lacking. This study aims to fill this gap in the literature and presents the results of an exploratory empirical study into AI applications in the Dutch public sector and make connections with literature on risks of externally developed AI applications to legitimacy of government in an exploratory way. In particular this study has mapped where the technical development of AI applications has taken place: in-house, by an external party, in collaboration with an external party or in a collaborative partnership between multiple parties. Moreover it provides insight into which type of applications are developed externally, which government organisations use externally developed AI applications and in which domains these applications are applied. This exploratory mapping study found 265 AI applications in the Dutch public sector, of 207 applications it was clear how the technical development has taken place. We find that the majority of applications have been developed with the involvement of an external party or where multiple parties are involved. This indicates that the use of externally developed AI applications by governments is widespread, in particular municipalities make use of externally developed AI applications. The findings also give insight into a number of risks to legitimacy of government such as a greater dependency and intertwinement with external parties and risks to accountability, transparency and oversight of public sector AI applications.

This paper explores the intricate dynamics of increased workload and e-governance in the context of amalgamation reforms in Armenia. Amid a backdrop of rapid decentralisation and local government consolidation, the study delves into the implications of these reforms on the daily lives of public servants on different levels of government and on ordinary citizens. It meticulously examines how the distribution of work and responsibilities shifted as local communities amalgamated, providing insights into the efficiencies and challenges that emerged. Additionally, the paper scrutinises the role of digital governance tools in mitigating the workload pressures experienced by local administrators.

By investigating the synergy between amalgamation-induced increased workloads and the introduction of e-governance tools, the paper aims to shed light on the underlying tensions and potential solutions that emerge in local governance systems undergoing significant transformation. It is within this dynamic space that we uncover the intricate interplay between administrative efficiency and public service delivery, shedding light on the broader implications for democracy, accountability, and citizen engagement.

Ultimately, this study contributes to the ongoing discourse surrounding decentralisation, local democracy, and e-governance, offering valuable insights into the evolving landscape of local governance in the era of consolidation and digitalisation.

The usage of AI applications in providing public administrations' services has specific opportunities and risks which highlight the necessity to carefully consider how AI is used and implemented. To this, research requires content-valid conceptualization of AI utilization and impacts and an understanding of how AI affects the quality of administration. We build on the general notion of task-technology-fit (TTF; Goodhue and Thompson 1995) that, in order to positively impact performance, technology must match the tasks characteristics well that it supports when being utilized. Sturm & Peters (2020) describe an extended and contextualized theoretical model of TTF in the context of AI which details characteristics of data, AI and task as well as task-data fit and task-AI fit. This task-AI-fit finally determines the utilization of AI and its performance impacts. This study seeks to explore the applicability and validity of the proposed model for AI implementations in public administrations by examining current use cases of AI-systems, based on Large Language Models (LLMs) in two German municipal administrations. For each AI implementation, the conceptualization of the proposed characteristics will be analyzed. A comparison of the two cases allows to explore differences in AI-task fit and their consequences.

The data collection is conducted in a three-step mixed-methods process for each case. First, focus groups are used to gather tasks that were previously delegated to the AI-system providing information on the characteristics of these tasks. Furthermore, interviewees reported the perceived usability of the AI generated output which indirectly indicates high or low fit of the used AI application and the particular task. Secondly, expert interviews are conducted to gain insight into the characteristics of data and AI implementation. Thirdly, in-depth interviews validate use cases and provide insights into the traits of users and their perception of AI. Finally, a comparison of the two cases furthermore provides insight into the applicability of the model to different AI characteristics and allows exploring the consequences of differences in the certain AI attribute on TTF.

Preliminary results give information for the further development of a model for evaluating TTF for AI systems in public administrations as well as insights into risks of their deficits.

Hence, the results of our research will have implications for research on AI usage in public administration by contributing to content-valid conceptualization of relevant variables and provide some insight into status, chances, and hindrances of current AI implementation in German municipal administration. Limitations of the study lie in the nature of the case study design which limits the generalizability of our findings. Moreover, by focusing on TTF as predictor of AI impact on performance a number of important factors such as the individual public servant’s AI competence are neglected.

The adoption of artificial intelligence (AI) in public administration (PA) represents a significant shift towards a more transparent, efficient, and responsive governance that creates remarkable public value. It enables smarter resources management, the provision of customised public services and improves citizen-government interaction. However, the rapid progress of AI, coupled with strict institutional regulations, poses an obstacle for the PA, which often finds it difficult to keep up with the pace of AI innovation. This problem stems from the unclear conceptualisations of AI in the context of PA. Although AI is gaining momentum, its integration into PA strategies and operations is a complex task with unique challenges that have not yet been fully explored. Therefore, this paper aims to fill the existing research gap in comprehensive analysis, especially with regard to the practical implementation of AI in PA. Although there are many theoretical reviews conducted, there is a clear lack of empirical evidence and dissemination of knowledge about systematic practical implementation. Hence, this paper represents an extension of existing papers (e.g. Aristovnik et al., 2024; Reis et al., 2019; Savaget et al., 2019; Sharma et al., 2020). The objectives of this paper are: to identify the current state of scientific AI research in PA, to identify overlaps and differences of concepts related to AI in PA, and to provide a systematisation of the practical implementation of AI in PA.

In order to achieve these objectives, this paper uses a combination of bibliometric and content analysis methods. First, a bibliometric analysis of 3,149 documents in the Scopus database published up to and including 2023 is conducted to gain an understanding of the current state and research trends in scientific research. In addition to a descriptive overview, advanced bibliometric approaches such as regression-based text mining are applied to understand the overlap between subsets, types, and practical applications of AI. Subsequently, an advanced content analysis using conceptual, thematic, and relational approaches is performed on prominent reports and analyses from institutions such as the World Bank, OECD, and European Commission, which are additionally analysed with quantitative cluster analysis to verify the results. The presented systematisation of the practical implementation of AI in PA covers three levels of governance: internal processes, service delivery and policy making.

The preliminary results indicate that AI is becoming increasingly important in PA research and is finding concrete applications in various sectors. The expected outcome of this study is a refined conceptualisation of AI and its subsets, types, and practical applications within PA. This improved understanding is crucial for addressing the theoretical and practical challenges related to AI in PA. This paper aims to provide researchers with deep theoretical insights and identify potential areas for further research on the use of AI in PA. Furthermore, it aims to support policy makers and practitioners by providing a comprehensive systematic overview of the practical applications of AI in PA. By linking theory with practical applications, this paper aims to facilitate the successful integration of AI and thus improve PA services.

The national strategy for artificial intelligence (AI) holds great practical significance for a country's AI technology application and AI industry development. From the perspective of multi-streams theory, a framework for analysis is constructed based on four aspects: problem stream, policy stream, political stream, and policy entrepreneurship. A fuzzy set qualitative comparative analysis is conducted on 68 cases of national AI strategy formulation. The study reveals that no single variable becomes a necessary condition for opening the policy window of national AI strategy. Further regional analysis indicates that opening the policy window of national AI strategy in developing countries is primarily driven by problem stream or political stream, while in developed countries, it is primarily driven by policy entrepreneurship or multi- streams. These findings help clarify the driving factors and constraints for opening the policy window of national AI strategy, providing insights for understanding the differentiated paths of opening the policy window of national AI strategy and for tailoring AI national strategy formulation according to specific contexts.

Digital platforms are a strategic behavior adopted by the government in response to complex task situations, and it has been widely used as a technology intermediary to improve government efficiency. Existing research focuses on the changes in organizational power relations caused by digital platforms and the mechanism of their synergistic functions. However, there is still room for front-end discussions on digital platforms: platform technology adoption and the process of building them. The shortcomings of this discussion lead to the inability of existing studies to explain the success of digital platforms and how they compensate for deficiencies in organizational power structure.

Based on the case study method, the study selects the application system of hazardous rock disaster risk management and control in City C. This digital platform is a government-tailored management application system based on the scenario of risk management in dangerous rock formations. It not only enables data sharing but also facilitates cross-departmental collaboration in risk control based on shared data, achieving efficient risk management processes and controllable governance outcomes.

Based on the case study, this paper will answers the following questions: how do organizations make decisions about technology adoption, how do they design digital platforms, and how does the application of digital platforms fill the power gap between organizations. Based on the theory of technology-organization relationship, this paper constructs an analytical framework of "two-way empowerment of technology and governance". It proposes that the adoption, design, and application of digital platforms are the result of the mutual construction of technological structure and organizational structure, which is essentially to fill the power gap between organizations through technology and realize the consensus that technology empowers organizations. In the stage of technology adoption and technology design, the organization empowers technology, which provides an institutional basis for the successful embedding of technology in the organization, and in the technology application stage, the technology empowers the organization which promotes the improvement of organizational governance efficiency. The platform's full-process construction has achieved intergovernmental coordination, data sharing, and resource integration, thereby promoting the mutual transformation and empowerment between technology application systems and organizational governance, filling organizational power gaps, and achieving the governance goal of using technological innovation to promote high-quality development of emergency management work.