Home / Current Issue / Paper 1710379
Techniques to Reduce Bias in Training Datasets: A Survey in Fairness in Artificial Intelligence
Subject area: Science,Engineering and Technology · Area of research: Artificial Inteliigence
Abstract
Artificial Intelligence (AI) and Machine Learning (ML) systems are increasingly used in sensitive areas like healthcare, recruitment, finance, and law enforcement. However, people often question the fairness of these systems due to biases in training datasets. These biases come from issues like sampling errors and historical prejudice. They can carry through algorithms and cause unfair or discriminatory outcomes. This survey reviews current methods to reduce dataset bias in ML models. The study divides these methods into pre-processing, in-processing, and post-processing approaches and compares their strengths and weaknesses. The paper notes recent developments in fairness-focused ML and offers insights into the trade-offs between model performance and fairness. It wraps up with a discussion on future research opportunities to create more fair and transparent AI systems.
Keywords
Artificial Intelligence, Bias Mitigation, Dataset Fairness, Ethical Machine Learning, Responsible AI
References
[1] Mehrabi, N., Morstatter, F., Saxena, N., Lerman, K., & Galstyan, A. (2021). A survey on bias and fairness in machine learning. ACM Computing Surveys (CSUR), 54(6), 1-35.
[2] Barocas, S., Hardt, M., & Narayanan, A. (2019). Fairness and Machine Learning. Cambridge: fairmlbook.org.
[3] Bellamy, R. K., et al. (2019). AI Fairness 360: An extensible toolkit for detecting, understanding, and mitigating unwanted algorithmic bias. IBM Journal of Research and Development, 63(4/5), 4-1
[4] Corbett-Davies, S., & Goel, S. (2018). The measure and mismeasure of fairness: A critical review of fair machine learning. arXiv preprint arXiv:1808.00023.
[5] Kleinberg, J., Mullainathan, S., & Raghavan, M. (2017). Inherent trade-offs in the fair determination of risk scores. Proceedings of Innovations in Theoretical Computer Science (ITCS).
How to cite this paper
@article{1710379,
author = {Vivek Santhosh Rai},
title = {Techniques to Reduce Bias in Training Datasets: A Survey in Fairness in Artificial Intelligence},
journal = {Iconic Research And Engineering Journals},
year = {2025},
volume = {9},
number = {3},
pages = {19-20},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1710379.pdf},
abstract = {Artificial Intelligence (AI) and Machine Learning (ML) systems are increasingly used in sensitive areas like healthcare, recruitment, finance, and law enforcement. However, people often question the fairness of these systems due to biases in training datasets. These biases come from issues like sampling errors and historical prejudice. They can carry through algorithms and cause unfair or discriminatory outcomes. This survey reviews current methods to reduce dataset bias in ML models. The study divides these methods into pre-processing, in-processing, and post-processing approaches and compares their strengths and weaknesses. The paper notes recent developments in fairness-focused ML and offers insights into the trade-offs between model performance and fairness. It wraps up with a discussion on future research opportunities to create more fair and transparent AI systems.},
keywords = {Artificial Intelligence, Bias Mitigation, Dataset Fairness, Ethical Machine Learning, Responsible AI },
month = {September},
}