Abstract:Abstract With the increasing influence of machine learning algorithms in decision-making processes, concerns about fairness have gained significant attention. This area now offers significant literature that is complex and hard to penetrate for newcomers to the domain. Thus, a mapping study of articles exploring fairness issues is a valuable tool to provide a general introduction to this field. Our paper presents a systematic approach for exploring existing literature by aligning their discoveries with predetermined inquiries and a comprehensive overview of diverse bias dimensions, encompassing training data bias, model bias, conflicting fairness concepts, and the absence of prediction transparency, as observed across several influential articles. To establish connections between fairness issues and various issue mitigation approaches, we propose a taxonomy of machine learning fairness issues and map the diverse range of approaches scholars developed to address issues. We briefly explain the responsible critical factors behind these issues in a graphical view with a discussion and also highlight the limitations of each approach analyzed in the reviewed articles. Our study leads to a discussion regarding the potential future direction in ML and AI fairness.
What problem does this paper attempt to address?
This paper mainly explores the issue of fairness in machine learning models, current solutions, and challenges. With the increasing influence of machine learning in decision-making processes, fairness has attracted widespread attention. Researchers have found that due to bias in training data, bias in the models themselves, conflicts in different fairness concepts, and lack of transparency in predictions, machine learning models may produce discriminatory outcomes. The paper presents a classification of fairness issues in machine learning and maps various methods developed by scholars to address these problems. Additionally, the paper graphically explains the key factors behind these issues, discusses the limitations of various methods, and points out possible future research directions.
The paper first introduces the significant progress of machine learning models in multiple fields but also points out the potential biases that may exist in applications such as university admissions and job screening. The paper discusses in detail issues such as biased datasets, biased models, conflicts in fairness definitions, and lack of prediction transparency, and proposes a classification system for fairness problems. Furthermore, the paper emphasizes the importance of addressing these issues to ensure fairness in machine learning predictions.
Through a systematic analysis of existing literature, the paper aims to provide a comprehensive introduction for researchers entering this field, helping them understand various fairness-related terms, methods, and their limitations. The paper also identifies research gaps and challenges, providing guidance for future work. Despite the existing literature reviews on fairness assurance methods, this paper believes that a deeper exploration of the relationship between problems and methods is needed and proposes a more detailed classification that links fairness issues to corresponding solutions.
Through this approach, the paper aims to provide a lasting and organized framework for fairness research in machine learning and artificial intelligence, promoting the development of more robust, transparent, and fair machine learning models.