Data do not materialize out of thin air. Rather, data are generated from real-world processes with complex causal structures we do not observe directly. G nor can we observe both interventions for the same subject. For example, in the context of CEBaB, we might ask 1. Each of the above questions requires the estimation of a different theoretical quantity.

Explaining model behavior has emerged as a central goal within ML.

Causal concept effect estimation is gaining increasing interest in the field of interpretable machine learning. This general approach explains the behaviors of machine learning models by estimating the causal effect of human-understandable concepts, which represent high-level knowledge more comprehensibly than raw inputs like tokens. However, existing causal concept effect explanation methods assume complete observation of all concepts involved within the dataset, which can fail in practice due to incomplete annotations or missing concept data. We theoretically demonstrate that unobserved concepts can bias the estimation of the causal effects of observed concepts. To address this limitation, we introduce the Missingness-aware Causal Concept Explainer (MCCE), a novel framework specifically designed to estimate causal concept effects when not all concepts are observable. Our framework learns to account for residual bias resulting from missing concepts and utilizes a linear predictor to model the relationships between these concepts and the outputs of black-box machine learning models. It can offer explanations on both local and global levels. We conduct validations using a real-world dataset, demonstrating that MCCE achieves promising performance compared to state-of-the-art explanation methods in causal concept effect estimation.

The increasing size and complexity of modern ML systems has improved their predictive capabilities but made their behavior harder to explain. Many techniques for model explanation have been developed in response, but we lack clear criteria for assessing these techniques. In this paper, we cast model explanation as the causal inference problem of estimating causal effects of real-world concepts on the output behavior of ML models given actual input data. We introduce CEBaB, a new benchmark dataset for assessing concept-based explanation methods in Natural Language Processing (NLP). CEBaB consists of short restaurant reviews with human-generated counterfactual reviews in which an aspect (food, noise, ambiance, service) of the dining experience was modified. Original and counterfactual reviews are annotated with multiply-validated sentiment ratings at the aspect-level and review-level. The rich structure of CEBaB allows us to go beyond input features to study the effects of abstract, real-world concepts on model behavior. We use CEBaB to compare the quality of a range of concept-based explanation methods covering different assumptions and conceptions of the problem, and we seek to establish natural metrics for comparative assessments of these methods.

How can we understand classification decisions made by deep neural nets? We propose answering this question by using ideas from causal inference. We define the ``Causal Concept Effect'' (CaCE) as the causal effect that the presence or absence of a concept has on the prediction of a given deep neural net. We then use this measure as a mean to understand what drives the network's prediction and what does not. Yet many existing interpretability methods rely solely on correlations, resulting in potentially misleading explanations. We show how CaCE can avoid such mistakes. In high-risk domains such as medicine, knowing the root cause of the prediction is crucial. If we knew that the network's prediction was caused by arbitrary concepts such as the lighting conditions in an X-ray room instead of medically meaningful concept, this would prevent us from disastrous deployment of such models. Estimating CaCE is difficult in situations where we cannot easily simulate the do-operator. As a simple solution, we propose learning a generative model, specifically a Variational AutoEncoder (VAE) on image pixels or image embeddings extracted from the classifier to measure VAE-CaCE. We show that VAE-CaCE is able to correctly estimate the true causal effect as compared to other baselines in controlled settings with synthetic and semi-natural high dimensional images.