Year: 2026
Type: dissertation
Abstract: Artificial intelligence has emerged as a key factor in public administration, but its use in government to improve efficiency or performance is not as straightforward as in the private sector. Public authorities must consider legality, accountability, transparency, and democratic legitimacy when adapting AI models. This thesis examines how responsible artificial intelligence can be applied in public administration in a workable and reviewable manner. The focus lies on four key questions: which responsible AI requirements matter most in government use, how national and EU-level governance frameworks shape them, what organisational, technical, and procurement-related barriers hinder implementation, and what a lifecycle-based governance model for public-sector AI systems would consist of. The thesis draws on literature and regulatory analysis, bringing together research on responsible AI and public-sector AI governance with the European Union legal framework, especially the General Data Protection Regulation and the AI Act, as well as Finnish legal and policy perspectives on digital governance. The takeaway of the findings is that responsible AI in public administration cannot be treated simply as a set of broad ethical principles. In the public sector, those principles become meaningful only when embedded in concrete governance practices. The thesis identifies human oversight, robustness, privacy and data governance, transparency, fairness, accountability, and social and environmental well-being as the key requirements for public-sector AI. It also shows that implementation is often held back by fragmented data environments, limited institutional capacity, legacy systems, and dependence on vendors. These conditions can create a gap between formal responsibility and actual control. The main contribution is the proposal of a phased lifecycle governance framework that turns responsible AI into a series of practical measures consisting of decision points, documentation, procurement controls, pre-deployment assurance, monitoring, and change-management practices. It also argues that environmental sustainability should be addressed within the same governance lifecycle rather than separately. The overall conclusion is that responsible AI in public administration is best understood as an institutional capability that must be sustained throughout the full lifecycle.
Source: Doria (University of Helsinki)
Author: Tomperi, Nadia
Institution: Åbo Akademi University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7163784596