Conference Agenda

Despite the recognized potential of artificial intelligence (AI) in public administration (PA), its practical implementation often remains limited. This study posits that organisational culture is a crucial determinant of digital maturity and thus successful AI adoption. Focusing on Italian municipalities - a context shaped by an aging workforce, territorial disparities, and massive investments in digital transition - this research investigates the hows and the whys of digital transition and AI adoption in PA. A two-stage, mixed-methods design was employed. First, a fuzzy-set Qualitative Comparative Analysis (fsQCA) on 2020 survey data from 17 municipalities identified two pathways to digital maturity: (1) a strong information culture as a standalone catalyst, and (2) the combination of large organisational size and mature data management practices. This was followed by in-depth interviews in 2025 with eight digital transition managers to explore the underlying cultural dynamics. The qualitative interviews revealed a central tension between a traditional, risk-averse bureaucratic-juridical culture and an emergent innovation-driven culture. Key challenges identified include staff resistance, skills shortages, over-reliance on external suppliers leading to a loss of strategic control, and concerns over the long-term financial sustainability. The study concludes that digital transformation and AI adoption in PA is mainly a cultural challenge, not merely a technological one. Findings call for policies that foster agile experimentation and collaborative governance while ensuring long-term financial and operational sustainability beyond temporary funding cycles.

Purpose of the Paper

This paper presents an innovative application of AI—declarative constraint modelling—to analyze deep contradictions in contemporary public policy. Two of the most pressing challenges facing the study of public administration in the modern era are the rapid rise of artificial intelligence and the erosion of the postwar free trade consensus, particularly in the United States. These twin shocks are fragmenting the policy-making process, creating conflicting institutional responsibilities, and undermining the discipline's dominant governance paradigms. We show how AI-based declarative modelling provides a new tool to represent, simulate, and evaluate policy processes in such fractured environments.

Research Approach and Methods

We model the ideological and policy conflict between two dominant but incompatible policy paradigms in recent U.S. trade governance: one grounded in liberal internationalism and open-market integration, the other rooted in economic nationalism and strategic protectionism. Drawing from speeches, legislation, and strategic policy documents (2016–2024), we encode each paradigm’s assumptions, goals (e.g. efficiency vs. sovereignty), institutional mandates (e.g. WTO rules vs. national security exceptions), and accountability structures. Using declarative modelling and Answer Set Programming (ASP), we simulate which combinations of assumptions, institutional roles, and policy actions produce internally consistent traces, and which lead to contradiction, deadlock, or drift.

Main Findings and Implications

Our model reveals that the opposing policy orientations are not merely ideological but are structurally valid within different, yet mutually exclusive, constraint systems. Much like two architects designing buildings under different building codes, each approach yields a coherent design, but the rules they follow are incommensurable. Declarative AI modelling enables public administration scholars to trace not only what policies are possible, but also under which assumptions and institutional conditions they make sense. This opens new space for evidence-based adjudication between rival policy logics—by showing, for example, which constraint systems better support long-term coherence, institutional accountability, or democratic legitimacy. More broadly, the paper advocates for AI-based declarative modelling as a vital method for navigating the non-linear, contradictory, and multi-agent challenges that now define effective public governance.

References

Simon, H. A. (1996). The Sciences of the Artificial. MIT Press.

Zuboff, S. (2019). The Age of Surveillance Capitalism. PublicAffairs.

Rodrik, D. (2018). Straight Talk on Trade: Ideas for a Sane World Economy. Princeton University Press.

Kuttner, R. (2018). Can Democracy Survive Global Capitalism? W. W. Norton.

This paper presents the proposal for a Transversal Literacy Scale for the Public Service, conceived as a strategic tool for diagnosis, self-positioning and training planning in a context of accelerated transformation of the Public Administration (PA). The scale is based on four meta-dimensions — cognitive, socio-emotional, technological and transformative — identified by a panel of experts in training and skills development in the Portuguese PA, using a qualitative and inductive methodology. The aim of this proposal is to contribute to the development of training policies that are more informed, responsive and aligned with the contemporary imperatives of public governance.

Artificial intelligence (AI) is one of the defining themes of our time, capable of driving significant changes in both the economy and society. In the public sector, AI offers promising prospects for improving public functions and public services. AI, therefore, holds cross-cutting relevance, impacting various domains of the public sphere, such as healthcare, public safety, education, and transportation. Over time, first through analogical applications and general principles, and later through legislative interventions at both the European level and within individual EU member states, a framework of "rules" governing AI usage has been developed - some of which apply specifically to the public sector. In this context, the principles of “algorithmic legality” related to AI are expressed through transparency, the non-exclusivity of automation, and non-discrimination. However, a crucial aspect that is often overlooked lies in the methods through which public administrations acquire AI systems. Since governments and public entities cannot typically independently produce and manage these systems, they must procure them from private sector operators. As a result, public procurement play a central role in the transformative change of the State. AI public contracts, however, present challenges far more complex than those of traditional public procurement. The key issues extend beyond innovative procurement (mechanisms) or the heightened significance of intellectual property rights. Additional complexities arise from the need to structure, manage, and utilize a vast amounts of data effectively. More importantly, however, AI public procurement reshapes the relationship between public entities and private operators in delivering public functions and services. This legal transformation must be analyzed within the broader social change driven by AI, as it may also lead to shifts in the legal boundaries and power dynamics between the "public" and "private" sectors. The primary challenge, therefore, is not only to enhance the efficiency of public functions and services through AI but also to define AI’s legal framework, ensuring that it meets the necessary legal standards for its legitimate use. In this context, public procurement takes on a new and crucial role, potentially becoming a gatekeeper for the transformative change of the State. This evolving role is complex, not only due to its intrinsic nature but also because it is subject to multiple regulatory frameworks, including public contracts law, AI governance, data protection, cybersecurity, and sector-specific regulations - all of which must be harmonized into a functionally and coherent system. Above all, this role must address the fundamental reality that, while public functions and services remain under government authority, they are increasingly underpinned by “private powers” of global relevance, further complicating the governance of AI in the public sector.