Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach
The rapid advancements in artificial intelligence (AI) have revolutionized industries such as healthcare, finance, and education. However, these advancements have also intensified ethical concerns regarding bias, fairness, and accountability in AI systems. Traditional fairness evaluation methods pri...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10945860/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850214213331451904 |
|---|---|
| author | Dongsoo Moon Seongjin Ahn |
| author_facet | Dongsoo Moon Seongjin Ahn |
| author_sort | Dongsoo Moon |
| collection | DOAJ |
| description | The rapid advancements in artificial intelligence (AI) have revolutionized industries such as healthcare, finance, and education. However, these advancements have also intensified ethical concerns regarding bias, fairness, and accountability in AI systems. Traditional fairness evaluation methods primarily focus on dataset-level biases, overlooking biases arising from model decision-making processes. This study introduces a novel framework for identifying, evaluating, and mitigating biases in AI models using counterfactual fairness, a robust approach that simulates alternative outcomes to minimize discriminatory effects. The proposed methodology integrates fairness-aware data preprocessing, feature selection, and model optimization strategies, ensuring equitable treatment across demographic groups. To validate the framework, we conducted empirical experiments using random forest and eXtreme Gradient Boosting models on the xAPI-Edu-Data dataset. Our results demonstrate significant improvements in demographic parity and equal opportunity fairness metrics while maintaining high predictive performance. Furthermore, comparative analysis with existing bias mitigation techniques confirms that our approach effectively reduces bias propagation in AI decision-making processes. By incorporating counterfactual fairness into AI design, this study provides a scalable and adaptable solution for ensuring ethical AI deployments aligned with regulatory standards. |
| format | Article |
| id | doaj-art-b0f22b8dad5747f7a8ff8030417d5af3 |
| institution | OA Journals |
| issn | 2169-3536 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Access |
| spelling | doaj-art-b0f22b8dad5747f7a8ff8030417d5af32025-08-20T02:08:57ZengIEEEIEEE Access2169-35362025-01-0113591185912910.1109/ACCESS.2025.355608210945860Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness ApproachDongsoo Moon0https://orcid.org/0009-0001-5770-1786Seongjin Ahn1Department of Computer Education, Sungkyunkwan University, Seoul, Republic of KoreaDepartment of Computer Education, Sungkyunkwan University, Seoul, Republic of KoreaThe rapid advancements in artificial intelligence (AI) have revolutionized industries such as healthcare, finance, and education. However, these advancements have also intensified ethical concerns regarding bias, fairness, and accountability in AI systems. Traditional fairness evaluation methods primarily focus on dataset-level biases, overlooking biases arising from model decision-making processes. This study introduces a novel framework for identifying, evaluating, and mitigating biases in AI models using counterfactual fairness, a robust approach that simulates alternative outcomes to minimize discriminatory effects. The proposed methodology integrates fairness-aware data preprocessing, feature selection, and model optimization strategies, ensuring equitable treatment across demographic groups. To validate the framework, we conducted empirical experiments using random forest and eXtreme Gradient Boosting models on the xAPI-Edu-Data dataset. Our results demonstrate significant improvements in demographic parity and equal opportunity fairness metrics while maintaining high predictive performance. Furthermore, comparative analysis with existing bias mitigation techniques confirms that our approach effectively reduces bias propagation in AI decision-making processes. By incorporating counterfactual fairness into AI design, this study provides a scalable and adaptable solution for ensuring ethical AI deployments aligned with regulatory standards.https://ieeexplore.ieee.org/document/10945860/AI biasbias mitigationcounterfactual fairnessethical AIfairness evaluation |
| spellingShingle | Dongsoo Moon Seongjin Ahn Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach IEEE Access AI bias bias mitigation counterfactual fairness ethical AI fairness evaluation |
| title | Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach |
| title_full | Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach |
| title_fullStr | Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach |
| title_full_unstemmed | Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach |
| title_short | Metrics and Algorithms for Identifying and Mitigating Bias in AI Design: A Counterfactual Fairness Approach |
| title_sort | metrics and algorithms for identifying and mitigating bias in ai design a counterfactual fairness approach |
| topic | AI bias bias mitigation counterfactual fairness ethical AI fairness evaluation |
| url | https://ieeexplore.ieee.org/document/10945860/ |
| work_keys_str_mv | AT dongsoomoon metricsandalgorithmsforidentifyingandmitigatingbiasinaidesignacounterfactualfairnessapproach AT seongjinahn metricsandalgorithmsforidentifyingandmitigatingbiasinaidesignacounterfactualfairnessapproach |