Book Title: Shaping the Future: Innovation, Sustainability, and Inclusive Growth in a Globalized Economy
Editors: Editors: Dr. Shanu Singh, and Dr. Yashmita Awasthi
Student Editor: Krishna Singh Rawat
ISBN: 978-93-7183-006-5
Chapter: 20
DOI: https://doi.org/10.59646/725/20
Authors: Maanya Singh, New Begin M, and Benita Jaison
Abstract
At the moment, the increasing influence of Artificial intelligence (AI) systems used in making high-stake decision making in various fields, such as healthcare, finance, and the criminal justice system, has resulted in an increased need for careful oversight of ethics. The present investigation is introducing EthicLens, which is a multi-layered computational framework that is designed to audit machine learning models to ensure compliance with global governance standards including the European Union AI Act, the OECD AI Principles, and the NIST bias guidelines, respectively. EthicLens operationalises abstract ethical values (fairness, transparency, accountability and privacy) to quantifiable values, therefore enabling a systematic evaluation of AI behaviour. Validation performed on the Adult Income and the COMPAS data show that fairness interventions greatly reduce bias (e.g. Statistical Parity Difference in COMPAS was reduced from -0.377 to -0.011) with minimal effect on accuracy and without sacrificing interpretability. EthicLens generates a compliance scorecard which makes the connection between technical measures and policy mandates and thereby offers a pragmatic and verifiable tool for ethical Artificial Intelligence (AI) governance.
Keywords: AI governance, EU AI act, SHAP interpretability, responsible AI, bias mitigation