AI Fairness Decision Enhancement Algorithm Based On Adversarial Bias Elimination
JIAO Wanni
LIU Haofeng
Abstract：Artificial intelligence decision models often face ethical and legal challenges due to biases in training data,choice of algorithms,or their implementation,which not only may contravene social justice and legal norms but also may limit the univer-sality and quality assurance of the models.The paper presents a multidimensional adversarial debiasing method that employs adver-sarial learning mechanisms to enhance fairness and reduce biases in the models.Experimental results demonstrates that the multidi-mensional adversarial debiasing model achieves significant improvements of 7%～10%in fairness metrics,with reductions of 8%～11%in both equal opportunity difference and average odds parity difference.This paper applies adversarial learning to eliminate un-fairness in algorithmic decision-making,effectively balancing the model's predictive performance with fairness,and provides a sol-id theoretical foundation and practical pathway for developing more refined and practical fairness metrics and standardized fairness algorithms in the future.
Keywords：adversarial debiasing techniquesalgorithmic fairnessethical and legal challengesAI applications
Publication Date：2025-07-20
Online Publishing Date：2025-09-25(First online date of this platform, not the publication date of the document)
Pages：5( 1812-1816 )