[目的/意义]针对人工智能应用中表现出的偏见和歧视，本文从数据、算法和系统运维出发，分析了AI治 理中公平性的影响因素和评价方法，为AI的公平性治理提供解决思路。[方法/过程]基于人工智能治理流程和AI 的公平性治理机制两个维度，本文首先遵循“顶层设计—技术攻关—产业实践”思路，提出AI的公平性治理框架； 其次，在AI治理公平性技术攻关层面提出了GovAI-FDevOps框架并从偏见角度出发探讨了公平性影响因素和量 化评价指标；最后以信贷数据集German为例，对信用评分卡模型在性别、国籍、年龄等敏感属性上的表现进行了 公平性检测和评价。[结果/结论]从本文研究来看，AI治理中的公平性是一个具有相对性的复杂概念，不存在满足 各群体、各统计意义上的绝对公平。AI治理公平性的实现需要政策、法律和技术的多方协同努力。

[Purpose/significance] In view of the bias and discrimination in the application of artificial intelligence, this paper analyzes the influencing factors and evaluation methods of fairness in AI governance from the perspective of data, algorithms and system operation and maintenance, and provides solutions for the fairness governance of AI. [Method/process] Based on the two dimensions of AI governance process and AI fairness governance mechanism, this paper first follows the idea of "Top-level design-Technical research- Industrial practice", and proposes the fairness governance framework of AI; Secondly, at the technical level of AI governance fairness,the GovAI-FDevOps framework is proposed, the influencing factors and quantitative evaluation indicators of fairness are discussed from the perspective of bias; Finally, taking the credit data set German as an example, this paper tests and evaluates the fairness of the credit scoring card model in gender, nationality, age and other sensitive attributes. [Result/conclusion] From the research of this paper, fairness in AI governance is a complex concept with relativity, and there is no absolute fairness that satisfies all groups and statistical significance. The realization of AI governance fairness requires the concerted efforts of policy, law and technology.