The problem with explaining symptoms: The origin of biases in causal processing