Abstract

The growing use of machine learning (ML) in healthcare has generated substantial interest in predictive performance. However, relatively less attention has been given to how algorithmic insights are shaped by institutional contexts. Consequently, explainable artificial intelligence (XAI) has emerged as a means to improve transparency, trust, and accountability in high-stakes settings such as healthcare. Despite its potential, existing research has largely treated explainability as a stable, model-level property and evaluated XAI techniques within isolated organizational or national contexts, which limits our understanding of whether insights travel across contexts. This study, therefore, adopts an information systems (IS) perspective to examine how institutional contexts condition the patterns, interpretability, and fairness of insights produced by explainable AI models. We leverage publicly available healthcare data from the United States and the United Kingdom to propose a cross-national comparative approach that treats public data as institutionally situated artifacts rather than neutral inputs. We also draw on institutional theory and socio-technical systems theory to conceptualize explainability as an emergent capability shaped by data infrastructures, governance arrangements, and population representation. Our research develops a theoretically grounded framework for understanding how institutional differences influence the behavior and interpretation of explainable AI models. Finally, we highlight public data as a valuable resource for transparent, replicable, and cumulative inquiry into algorithmic decision making. By reframing explainable AI as a context-dependent socio-technical phenomenon using the healthcare industry, we contribute to the IS healthcare literature by offering a foundation for future empirical research.

Download

Share

COinS