Some interesting notes

In causal inference, biases arise when omitted variables correlate with the outcome. 

But for prediction, biases arise when omitted variables correlate with payoffs.