Suspect you may be confusing the state of 'being safe', with the measurement of 'being safe'. Empirical methods are not the only way to quantify something.

Either world view relies on

"Just trust me, I know" or "The numbers never lie"

Would be potentially problematic.

This debate has been going on for centuries.

http://en.wikipedia.org/wiki/A_priori_and_a_posteriori