Research on the Framework of Bias Detection and Elimination in Artificial Intelligence Algorithms

The excessive use of artificial intelligence (AI) algorithms has caused the problem of errors in AI algorithms, which has challenged the fairness of decision-making, and has intensified people’s inequality. Therefore, it is necessary to conduct in-depth research and propose corresponding error detection and error elimination methods. This paper first proposes the root causes and threats of bias in AI algorithms, then summarizes the existing bias detection and error elimination methods, and proposes a bias processing framework in three-level dimensions of data, models, and conclusions, aiming to provide a framework for a comprehensive solution to errors in algorithms. At the same time, it also summarizes the problems and challenges in existing research and makes a prospect for future research trends. It is hoped that it will be helpful for us to build fairer AI.

Chen, W. K., Liu, H. F., & Wu, S. S. (2023). Framework for the elimination of AI algorithms based on fairness constraints. Computer Science and Exploration, 12, 102-108.

Wang, C. X., Zhang, P. F., & Liu, K. (2023). Research on algorithm bias detection and correction methods based on machine learning. Computer Applications and Software, 40(11), 48-52.

Zhao, Z. Y., & Liu, Y. X. (2024). Bias and elimination strategies in deep learning models. Artificial Intelligence and Big Data, 3, 25-30.