Bias and fairness in AI models: Evidence from existing studies

The study presents a systematic literature review on fairness and bias in AI models.  The review has primarily considered the types of bias, mitigation strategies, and evaluation metrics across domains such as recruitment, finance, and healthcare. The findings indicate that vulnerable populations are disproportionately affected by structural and technical sources of bias. However, the application of the metrics is inconsistent. Besides that, the mitigation strategies can be algorithmic regularization and data augmentation. Based on the review, the recommendation is to implement a multilevel approach that integrates governance, ethical, and technical measures. It can be instrumental in presenting transparency, accountability, and equity in AI systems.