Title: Artificial Intelligence Across Disciplines: Research, Innovation, and Intelligent Solutions
Editors: Dr. Subita Bhagat, Dr. A. Balamurugan, Dr. P. Krishna Kumar, and Mrs. S. Nandhini Devi
ISBN: 978-81-69857-83-3
Chapter: 10
DOI: https://doi.org/10.59646/815/10
Authors: Dr. I. Karthiga, Mrs. P. Joy Kiruba, Mrs. A. Sulthana Rashya Begam, and Mrs. A. Snegaa
Abstract
Autonomous machine learning models deployed in high-stakes environments—such as clinical risk prediction, algorithmic lending, and public safety—frequently function as opaque systems, obscuring critical internal reasoning mechanisms from human operators. This chapter addresses the critical imperative for Explainable and Trustworthy Artificial Intelligence (XTAI) within responsible decision-making pipelines. We formulate a mathematically unified, dual-stage interpretability framework that integrates post-hoc cooperative game-theoretic attribution (SHAP) with intrinsic, sparsity-constrained counterfactual optimization under fairness-regularized constraints. Evaluated on an empirical cohort of 25,000 algorithmic credit decision profiles, our proposed framework mitigates disparate predictive outcomes across protected demographic attributes while isolating localized feature-level risks. The methodology ensures non-negative marginal contributions, eliminates local perturbation instability, and maintains predictive fidelity alongside explainability. Experimental results indicate a 4.1% gain in model fairness alongside a significant reduction in counterfactual generation latency, establishing a mathematically sound and auditable blueprint for regulatory-compliant, human-in-the-loop algorithmic governance across mission-critical domains.
Keywords: Explainable Artificial Intelligence, Algorithmic Fairness, Trustworthy AI, Counterfactual Explanations, Shapley Additive Explanations, Algorithmic Governance