Model explainability is a very active area of research. For the example of an individual neural network prediction you could use SHAP values[0]. Explainability

Model explainability is a very active area of research. For the example of an individual neural network prediction you could use SHAP values[0]. Explainability is far from a solved problem but there are certainly tools available to provide at least some transparency. I guess the website could provide references but that doesn't really seem to be the resolution being aimed for. [0] https://www.kaggle.com/dansbecker/shap-values https://www.kaggle.com/dansbecker/shap-values.