Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{199487,
        author = {Diya Prataprao Mokashi and Apurva Shaligram Kirote and Rasika Santosh Shahapure and Vitthal B. Kamble},
        title = {Bias in Artificial Intelligence: Origins, Impacts, and Mitigation Strategies},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {14400-14411},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=199487},
        abstract = {Artificial Intelligence (AI) systems are now deeply embedded in virtually every high-stakes domain including healthcare, criminal justice, financial services, hiring, and education. As their influence grows, so does the urgency of understanding and addressing the biases these systems inherit, amplify, and sometimes create. This article presents a comprehensive, multidisciplinary examination of bias in AI, tracing its origins from flawed or unrepresentative training data through structural choices made during model design, and extending to the social and institutional contexts in which systems are deployed. We analyse the principal categories of bias recognised in the literature historical bias, representation bias, measurement bias, aggregation bias, evaluation bias, and feedback loop bias and explain the mechanisms through which each propagates into automated decisions. The article surveys real-world harms across criminal justice, healthcare, financial services, employment, facial recognition, and large language models. Against this backdrop, we critically evaluate the principal mitigation strategies: pre-processing techniques such as reweighing and disparate-impact removal; in-processing approaches including adversarial debiasing and fairness-constrained optimisation; and post-processing methods such as equalized-odds adjustment. We further discuss Explainable AI (XAI) tools particularly SHAP and LIME as instruments for bias auditing, and examine Privacy-by-Design as a governance framework. We review the regulatory landscape including the EU AI Act and identify open research questions covering intersectional fairness, dynamic bias, the fairness-privacy-accuracy trilemma, and benchmark dataset diversity. Our central finding is that no single technical fix eliminates bias; sustained progress requires coordinated effort across data curation, algorithm design, organisational governance, and public policy.},
        keywords = {Algorithmic Fairness, Adversarial Debiasing, AI Bias, Bias Mitigation, Disparate Impact, Ethical AI, Explainable AI, Fairness Metrics, Machine Learning, Responsible AI, Statistical Parity.},
        month = {April},
        }