Explaining Classifiers with Causal Concept Effect (CaCE)
Yash Goyal Amir Feder Uri Shalit Been Kim
Abstract
How can we understand classification decisions made by deep neural networks? Many existing explainability methods rely solely on correlations and fail to account for confounding, which may result in potentially misleading explanations. To overcome this problem, we define the Causal Concept Effect (CaCE) as the causal effect of (the presence or absence of) a human-interpretable concept on a deep neural net’s predictions. We show that the CaCE measure can avoid errors stemming from confounding. Estimating CaCE is difficult in situations where we cannot easily simulate the do-operator. To mitigate this problem, we use a generative model, specifically a Variational AutoEncoder (VAE), to measure VAE-CaCE. In an extensive experimental analysis, we show that the VAE-CaCE is able to estimate the true concept causal effect, compared to baselines for a number of datasets including high dimensional images.
中文速览
深度神经网络在做图像分类时，现有可解释性方法大多只看特征与预测结果之间的相关性，一旦数据中存在混淆变量（confounding），解释结果就会出错——比如"自行车"图片里频繁出现"汽车"，相关性方法就会错误地把"汽车"当作分类自行车的原因。为此，作者提出了"因果概念效应"（Causal Concept Effect，CaCE），用因果推断中的干预算子（do-operator）来衡量某个人类可理解概念的存在与否对分类器输出的真实因果影响。由于在真实图像上直接施加干预（例如把图中的男性医生换成女性医生）极为困难，作者进一步引入条件变分自编码器（Conditional VAE）来近似这一干预过程，通过改变概念标签输入生成反事实图像，从而估计CaCE。在多个数据集（包括高维自然图像）上的实验表明，VAE-CaCE能够准确逼近真实因果效应，明显优于仅依赖相关性的基线方法，为构建不被混淆变量误导的全局可解释性工具提供了一条可行路径。
原文 arXiv:1907.07165;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1907.07165v2