Fairness Measure, Bias Mitigation Techniques and Verification Tools in Machine Learning: A Survey
Hanbit Gil, Yejin Shin, Joon Ho Kwak, Sungmin Woo · IEIE Transactions on Smart Processing and Computing · 2025
The emergence of OpenAI has served as a catalyst for public awareness of the potential of artificial intelligence technology, sparking interest not only in generative AI but also in on-device AI and general AI. Currently, AI has become indispensable not only in the field of natural language processing but also in easily accessible domains such as video and audio generation and automatic editing, leading us into an era where coexistence with such AI technologies is unavoidable. However, concerns remain. Unethical practices such as AI-generated reports and biased news articles highlight the need for responsible development. Furthermore, the ?black box? nature of neural networks presents a challenge for industries requiring high trustworthiness. To navigate this evolving landscape, we explore international standards and regulatory proposals for safe AI integration. We delve into academic research on fair AI techniques and examine bias mitigation tools offered by enterprises. Through this comprehensive analysis, we aim to empower AI developers and practitioners to build ethical systems and shape sound regulatory frameworks.