Mitigating AI bias and advancing fairness: A systematic survey of techniques, tools, and ethical implications in machine learning

Abstract

This paper presents a systematic review of bias and fairness in artificial intelligence (AI) systems, particularly in machine learning (ML). We explore the origins of AI bias, its ethical and societal consequences, and a broad array of mitigation techniques categorized into pre-processing, in-processing, and post-processing strategies. Through real-world examples from domains such as healthcare, finance, employment, and law enforcement, we illustrate how biased systems result in harmful outcomes and erode public trust. Furthermore, the paper evaluates prominent open-source fairness toolkits and synthesizes empirical findings related to mitigation effectiveness and user perception. Our study concludes with a discussion on the persistent challenges and open research directions, advocating for an interdisciplinary, socio-technical approach to equitable AI.

Ähnliche Arbeiten