知识卡片

普通读书笔记卡 · 6570.d · 生成模型:GPT-6

反事实公平依赖规范与模型

理解等级:中级

反事实公平询问在相关背景保持适当一致时,改变某个受保护属性是否会改变预测。需要定义允许保持与变化的后代、结构模型和规范标准;只改敏感字段或若干代理的数值,不等于完整公平验证,也不自动解决所有公平问题。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      counterfactual fairness analysis would ask counterfactual questions. In this case, the counterfactual question is “Would this person have been approved if they were of a different ethnicity?” The analyst would find features that are proxies for ethnicity and then see if a change to those proxies corresponding to a change in ethnicity would result in a classification of “approve.” Some techniques attempt to use this

    参考来源

    - 位置:《Causal AI》第08章《反事实与平行世界》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-8.xhtml) - 结论依据:摘录中关于反事实公平依赖规范与模型的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:counterfactual fairness analysis would ask counterfactual questions. In this case, the counterfactual question is “Would this person have been approved if they were of a different ethnicity?” The analyst would find features that are proxies for ethnicity and then see if a change to those proxies corresponding to a change in ethnicity would result in a classification of “approve.” Some techniques attempt to use this