Research work · 2025

Bias Reduction Techniques in Machine Learning Models: Investigating Strategies to Detect and Minimize Bias in AI and Machine Learning Algorithms

Arimondo Scrivano

Institute of Electrical and Electronics Engineers (IEEE)

Abstract

The increasing deployment of machine learning models across diverse applications necessitates a critical assessment of fairness and bias minimization. Machine learning algorithms, often trained on historical data, can inadvertently perpetuate or even exacerbate existing biases, resulting in unfair outcomes. This article reviews contemporary bias detection and mitigation methodologies within artificial intelligence and machine learning contexts. We synthesize current research on quantifying algorithmic bias, with a focus on pre-processing, in-processing, and post-processing strategies to mitigate bias. Furthermore, we discuss the implications of bias in machine learning applications and propose a framework for ongoing bias monitoring and reduction. The paper aims to bridge gaps in the existing literature by compiling effective methodologies while also addressing emerging challenges, such as intersectionality and transferability across different domains. Addressing these biases is crucial for ensuring equitable and trustworthy AI systems, which are increasingly becoming integral to decision-making processes.

DOI: 10.36227/techrxiv.175355735.50084414/v1