Bias And Unfairness In Machine Learning Models: A Systematic Review - Info and Reading Options

The Internet Archive:

One of the difficulties of artificial intelligence is to ensure that model decisions are fair and free of bias. In research, datasets, metrics, techniques, and tools are applied to detect and mitigate algorithmic unfairness and bias. This study aims to examine the latest existing knowledge about bias and unfairness in machine learning (ML) models with the RSL methodology and a bibliometric analysis. A Systematic Review was carried out between 2021 and early 2022 found 128 articles published between 2017 and 2022 in the Scopus, IEEE Xplore, Web of Science, and Google Scholar knowledge bases, 45 were chosen based on their Abstract and the optimization of search strings. The results indicate that articles mainly focus on bias and unfairness identification and mitigation techniques and types, presenting tools, statistical methods, key metrics and datasets commonly used for experimentation on bias issues. In terms of the main types of bias, we emphasize data, algorithm, and user interaction, with relation to the following mitigation methods: pre-processing, in-processing, and post-processing. The use of Equalized Odds (EO), Equality of Opportunity (EOO), and Demographic (DP) Parity as the primary measures of justice further highlights the significance of sensitive attributes in mitigating biases. The 25 datasets that were found include a range of areas, including criminal justice image enhancement, financial, education, product pricing, and health, and the bulk of them contain sensitive features. In relation to the tools, we highlight that not all of them are used in any practical studies, and the Aequitas was the most mentioned of them. However, a drawback of these works is the scarcity of multi-class and multi-metric investigations, which are found in just a few papers and restrict the study to binary-focused methodologies. %limitations Furthermore, the results indicate that different fairness metrics do not present uniform results for a given use case, and that more research with varied model architectures is necessary to standardize which ones are more appropriate for a given context. It is also crucial to emphasize the relevance of the fact that all researches have addressed the algorithm's transparency, or its capacity to explain how decisions are made.

Search for “Bias And Unfairness In Machine Learning Models: A Systematic Review” downloads:

Visit our Downloads Search page to see if downloads are available.

Find “Bias And Unfairness In Machine Learning Models: A Systematic Review” in Libraries Near You:

Read or borrow “Bias And Unfairness In Machine Learning Models: A Systematic Review” from your local library.

• The WorldCat Libraries Catalog: Find a copy of “Bias And Unfairness In Machine Learning Models: A Systematic Review” at a library near you.