OPERATIONALIZING RESPONSIBLE AI PRINCIPLES THROUGH A GLOBAL ACCOUNTABILITY FRAMEWORK MODEL

Abstract

The rapid expansion of artificial intelligence (AI) has intensified concerns about its ethical, social, and governance implications. While numerous frameworks and guidelines for Responsible AI have been issued globally, a gap persists between high-level principles and their operationalization. This study addresses this gap by developing an accountability framework through a systematic literature review (SLR) and validating it with a case study of IBM. The SLR analyzed 78 scholarly and policy sources, producing five thematic domains: values, governance instruments, metrics, barriers, and pathways, with accountability emerging as the integrative theme. To test this framework, the IBM case was examined across eight key documents, including the shortcomings of Watson Health and the company’s subsequent development of governance tools such as FactSheets 360, fairness toolkits, and counterfactual testing. Findings show that while IBM’s commitments align with global principles, accountability remains most vulnerable in metrics and real-world validation. Nevertheless, IBM’s innovations and participation in global standardization illustrate how accountability can be reinforced in practice. The study contributes theoretically by reframing accountability as the linchpin of Responsible AI, and practically by offering pathways for organizations and policymakers to strengthen AI governance.