EXPLAINABLE ARTIFICIAL INTELLIGENCE (XAI) MODELS FOR HIGH-STAKES DECISION-MAKING SYSTEMS

Abstract

As artificial intelligence (AI) continues to permeate high-stakes domains, such as 
healthcare and finance, the demand for Explainable Artificial Intelligence (XAI) has 
become increasingly urgent. The necessity for transparency in AI-driven decision
making systems arises not only from ethical considerations but also from the inherent 
complexities associated with machine learning models, which can often render their 
outputs opaque to users. It is critical to recognize that while AI models may 
demonstrate accuracy on averaged data, they can lack reliability when applied to 
specific individuals, necessitating robust frameworks for personalized uncertainty 
quantification ((Banerji et al., 2025)). XAI aims to bridge this gap by providing 
interpretable models or post hoc explanations that enhance human understanding and 
trust in AI systems ((Finzel et al., 2025)). Furthermore, as organizations navigate 
ethical and managerial implications, augmented leadership is essential for integrating 
AI insights while fostering transparency and combating biases ((Erhan et al., 2025), 
(Thulasiram et al., 2025)). Thus, XAI models serve as a vital component of responsible 
decision-making in high-stakes environments.