Causal Explanations and XAI
This work addresses the need for reliable action-guiding explanations in XAI, with implications for fairness, but it is incremental as it builds on existing causal model insights.
The paper tackles the mismatch between standard ML models optimized for predictions and their use in action-guiding scenarios by formally defining causal notions of sufficient and counterfactual explanations, showing how these improve existing work and are action-guiding under different circumstances. It also introduces a formal definition of actual causation based on action-guiding explanations and applies it to enhance fairness in AI, such as path-specific counterfactual fairness.
Although standard Machine Learning models are optimized for making predictions about observations, more and more they are used for making predictions about the results of actions. An important goal of Explainable Artificial Intelligence (XAI) is to compensate for this mismatch by offering explanations about the predictions of an ML-model which ensure that they are reliably action-guiding. As action-guiding explanations are causal explanations, the literature on this topic is starting to embrace insights from the literature on causal models. Here I take a step further down this path by formally defining the causal notions of sufficient explanations and counterfactual explanations. I show how these notions relate to (and improve upon) existing work, and motivate their adequacy by illustrating how different explanations are action-guiding under different circumstances. Moreover, this work is the first to offer a formal definition of actual causation that is founded entirely in action-guiding explanations. Although the definitions are motivated by a focus on XAI, the analysis of causal explanation and actual causation applies in general. I also touch upon the significance of this work for fairness in AI by showing how actual causation can be used to improve the idea of path-specific counterfactual fairness.