I made a related argument recently:

A theory that doesn’t account for detailed behavior is an approximation, and even in scientific domains, you can find conflicting approximations. When that happens—and if you’re not doing science, it’s “when,” not “if”—if you want to keep using your approximation, you have to use the details of the situation to explain why your approximation is valid. Your best defense against reductio ad absurdum, against Proving Too Much, is casuistry. Expect things to be complex, expect details to matter. Don’t ascribe intention or agency to abstract concepts and institutions. Look for chains of cause and effect. Look at individual moving parts and the forces acting on them. Make empirical predictions, and look for unintended empirical predictions. Ask what the opposite principle explains, and find the details that make those explanations compatible.