Mitigating Bias in AI: A Review of Sources, Impacts, and Strategies

Abstract

Objective: This research examines trends, approaches, and application contexts of bias mitigation strategies in artificial intelligence (AI) systems. The primary focus is on how biases emerge in different sectors and how mitigation practices are developed to address equity and ethical challenges in AI development.
Research Design & Methods: This research uses a Systematic Literature Review (SLR) approach with source selection and literature analysis from trusted databases such as IEEE Xplore, Scopus, SpringerLink, and ACM Digital Library. This study reviewed literature between 2018 and 2024 to ensure the relevance and novelty of findings in the context of bias mitigation in AI systems.
Findings: The study results show that bias mitigation strategies have evolved from a narrow technical approach to a comprehensive system lifecycle-based approach. Notable innovations include the application of data-centric AI, fairness-aware algorithms, targeted data augmentation techniques, post-processing, bias auditing, and explainable AI. These approaches have been applied in various sectors.
Implications & Recommendations: Effective bias mitigation demands a shift from a technical focus to a collaborative and multidisciplinary approach. System developers must embed fairness principles from the design stage, while regulators should promote transparency and accountability through strong policies. Systematic evaluation, cross-disciplinary collaboration, and public engagement are key for AI systems to be accepted as fair and responsible.
Contribution & Value Added: This research provides a structured synthesis of bias mitigation approaches and demonstrates how they can be applied in real-world contexts. By offering practical guidance towards adaptive and integrated mitigation practices, this study contributes to strengthening ethical AI discourse