Good point. Would you agree that: even if a phenomenon (for example performance of some sort of worker) is hard to model, we should still try, especially if that phenomenon provides enormous value to society?