I was using the word "parts" very loosely. In large systems, a problem might not be the fault of any specific part(s), but the result of complex interaction among parts. If so, that interaction as a whole, rather than its individual components, would be the "part" that is evil or buggy or (insert appropriately negative expression here).
Systems based on faulty architectures tend to be the most difficult to fix, since changing the basic architecture will affect nearly everything else. But even in such cases, it's not helpful to claim that the system as a whole is evil. If the architecture is bad, the architecture is bad. Anything more or less than that is hyperbole.
I don't know if I agree with your premise here. Why do destructive (or even self-destructive) systems have to have "evil parts"?
These are systems problems, not the type of problem you can solve by taking out one kind of people and putting in another (or technology, or political party, or company, or…).