New preprint released: “Causal Concept Embedding Models: Beyond Causal Opacity in Deep Learning”! Abstract Causal opacity denotes the difficulty in understanding the “hidden” causal structure underlying a deep neural network’s (DNN) reasoning.