In what cases real empirical experiments can be replaced by computer simulations?
Computer simulations deal with simplified abstract models; therefore, cannot reflect the reality in all details. At the same time, it is sometimes impossible to perform real empirical experiments due to unaffordable costs, huge spans of time, dangerous risks, etc. When can real empirical experiments be replaced by computer simulations? How creditable are the results generated by computer simulations? When can one refer to them as scientific evidences to support a thesis statement, for instance?
Is there a specific use case or scenario you have in mind? The question is very broad and not conducive to a specific answer.
For instance, I'm using EnergyPlus, THERM, etc. to develop and compare many high rise building prototypes to see what design strategies show better energy saving potentials in the long run; I use simulations to support my thesis. Needless to say that such measurements in reality take ages and lots of budget that make them almost impossible. Nevertheless, one may argue that to announce a thesis statement, you need to have empirical experiments, and computer simulations may include errors and inaccuracies, and therefore cannot support your thesis statement.
How do you reply to such arguments?
Given your description, I would say the answer is "when not feasible". It is just not feasible to construct and instrument several identical large buildings that differ in just some design methodologies. They could not be constructed that consistently. And no one would pay for that. It is complex enough to instrument some small test buildings or small residential buildings. In addition, empirical experiments are not a panacea. Measurement is fraught with inaccuracies due to the use of instruments and the loads (unless artificially created) are likely to vary considerably between buildings.
@JasonGlazer Many thanks for your informative answer.