Trustworthy AI in public administration: The PRO-Trust pipeline for ethical fraud detection in public procurement

Applying Artificial Intelligence (AI) to fraud detection in Public Procurement (PP) has become increasingly relevant due to its ability to process large volumes of data and identify suspicious patterns more efficiently than traditional methods. However, its implementation raises critical ethical concerns regarding respect for human autonomy, prevention of harm, fairness, and explainability. This study investigates these concerns through a four-phase methodology: (i) selection of ethical principles for AI in PP fraud detection, (ii) a Systematic Literature Review (SLR) on current approaches, (iii) cross-analysis and expert validation with senior auditors, and (iv) the construction of PRO-Trust, an ethics-oriented pipeline for constructing trustworthy AI-based fraud detection systems for PPs, based on the findings of the three former stages. PRO-Trust addresses five recurrent ethical challenges — system opacity, lack of user transparency, limitations in explainability tools, lack of operational oversight, and lack of strategic control — connected between tasks we specified for the construction process of AI-based technologies to fraud detection in PP, the responsible actors, and technological strategies to operationalize ethical principles. By doing so, this study contributes to the development of more trustworthy, transparent, and accountable AI-based fraud detection systems in public administration.