The Role of Causality in Forecasting

Causality has come under some scrutiny around this idea that how can you forecast the world if you don't understand what causes things. One of the criticisms has been really leveled against a lot of data science practice especially things like random forest or things like neural networks is that they're really silent about this idea of causality. So I'd just love to get your view as to how important is causal variables and their role in forecasting work.