Operationalizing Ethical Principles in AI-Based Software Systems: A Systematic Mapping Study

Background: The widespread adoption of Artificial Intelligence (AI) systems has intensified concerns regarding ethical principles such as fairness, transparency, and accountability. Although these principles are well established in international guidelines and regulatory initiatives, their effective operationalization within Software Engineering practices remains fragmented and uneven across the software development life cycle (SDLC). Goal: This study aims to systematically characterize how ethical principles are addressed and operationalized in AI-based systems within the Software Engineering literature, identifying the principles most frequently discussed and the types of approaches proposed to embed them into SDLC activities. Method: We conducted a Systematic Mapping Study (SMS) following established Software Engineering guidelines, analyzing peer-reviewed studies published between 2021 and 2025. Searches were performed in four major digital libraries (ACM Digital Library, IEEE Xplore, Scopus, and Web of Science). After a multi-stage screening process, 58 primary studies were included and classified according to the ethical principles addressed and the SDLC phases in which they are operationalized. Results: The results reveal a strong concentration on Fairness, Accountability, and Transparency/Explainability, which dominate the literature. Operationalization efforts are primarily focused on early and intermediate SDLC phases, particularly Requirements, Design, and Implementation, often through frameworks, checklists, design methods, and technical tools. In contrast, later phases such as Deployment, Operation, and Maintenance, as well as crosscutting areas like DevOps/MLOps and Project Management, receive limited attention. Several ethical principles, including Sustainability and Beneficence, remain underrepresented, indicating uneven maturity in the operationalization of AI ethics within Software Engineering.