What I was trying to say is that it is intractable to get the right data in many common cases, and in others unsolvable because of different premises. In the latter case of course it is good to be objective about what the premises are, but you can't derive a solution there often - it's possible but may not be worth the time needed.

Handwavy: often what you need is more rough-and-ready approximate statistics than any real logic. You are still approaching it with an objective mindset, but you are being realistic about the limitations of your data and modeling capabilities.

And at the end of the day a lot of what is important is also subjective.