Your model will only work if it is a correct model. How can you discriminate between an error in the model and an external influence? Both will manifest themselves as deviations between reality and the model. Your premise that "there is no magic" will lead you inevitably to conclude that the model is at fault, which would quite often be the case, but following that line when an external event was the cause of the deviation will only make your model more incorrect. Now, if your model con incorporate those "external influences", they would not have been external in the first place, would they?
Cheers!
P.S.: Not advocating either for or against ID. Just following the logic in the argument.