机器学习实验室博士生系列论坛（第十四期）—— Fairness in Machine Learning: Definition, Evaluation and Mitigation
报告人：Shiyun Lin (PKU)
地点：北大理科一号楼1513会议室 & 腾讯会议 761 4699 1810
Abstract：As AI systems become widely used in contexts that affect citizens, there is growing concern of algorithmic fairness in both academia and industry. And it is crucial to ensure that the application of machine learning technologies will not have unexpected social implications, such as bias towards gender, ethnicity. The last few years have seen an explosion of academic and popular interest in algorithmic fairness. In this talk, we will provide an overview of different schools of thought and approaches concerning fairness in the machine learning literature. We will firstly introduce several mathematical definitions of algorithmic fairness, including group fairness, individual fairness and counterfactual fairness, compare their pros and cons, and state incompatible results of different fairness notions. Then we will examine several methods to evaluate ML algorithms on fairness, both from the group level and the individual level. Furthermore, approaches aim to mitigate (social) biases and increase fairness, which categorized as pre-processing, in-processing and post-processing, will be reviewed. And we will illustrate several representative methods specifically. Finally we will shed light on possible challenges and opportunities in algorithmic fairness research.