Molecular modeling and error sensitivity
Potential energy functions differ in what they can describe, and in what they can describe accurately. Proper choice of systems and error margins can reduce sensitivity to modeling errors in engineering applications. “All models are wrong, but some are useful.” 
Models and accuracy:
Questions and accuracy requirements:
As cited by Stephen Duffull:
 Box, G.E.P. (1979) “Robustness in the strategy of scientific model building” In: Launer RL & Wilkinson GN Robustness in Statistics Academic Press pp. 202.