I long held the belief that it was the mission of the scientist to describe what Nature was doing and how, exactly. This is not completely correct, since to describe an event in full detail would essentially require the complete replication of the entire Universe since the entire surroundings of the system under investigation would have to be the same.
Rather, what scientists do is construct models of Nature, with the goal that the quantitative predictions these models make be within some agreed-upon margin of error when tested against reality. These simplifications are made for mathematical convenience, since the ultimate goal of the scientist is to calculate something. “Assume a spherical Earth.” “Let .” “Assume all competitive firms are price takers.” “Assume the process is adiabatic.” “Consider atoms held together by tiny springs.” “Let the system occur near Earth’s surface.” All of these simplifications and assumptions are to allow for complications to be swept under the rug and, hopefully, the final answer won’t deviate too much from truth.
This is actually a fairly important point, and deserves some thought. The job of the scientist is to calculate something, some predictions with a certain criteria for accuracy and precision. Some allowance for error is always agreed upon, and depending on the system and what prediction is made, the error can vary widely (precision tests of QED go to many decimal places, whereas an astronomer can often be fine with being off by a factor of 10). Therefore, what is paramount is creating a model of physical processes that allow for “easy” mathematical manipulation and calculation potential. These models can differ substantially in their nature, assumptions, and the way of thinking that it forces upon the modeler. It doesn’t really matter that the model reflects what’s “really” going on, as long as it has predictive power.
This is also a big point that separates a science from a pseudoscience: The ability to predict, and therefore the ability to be falsified if those predictions don’t pan out. Every time you use GPS its an intrinsic experiment as to whether general relativity is a valid theory, whereas psi researchers can never predict anything and so never go anywhere, since they have nothing to build on. The true scientist constructs a mathematical model of some natural phenomenon and uses it to make predictions. Some amount of error is defined to be acceptable, and experiments are conducted. If the model worked in predicting the right answer, some number, it gains validity and credibility. If the experimental value falls outside the error range (and no errors were made by the experimenters), the model must be modified or discarded. By this almost Darwinian selection process over the past four or so centuries we’ve built up some pretty good models of the universe, and our technical civilization sits comfortably on the foundations of those models.
Of course, that’s not the whole story, since science is also about telling stories. The mathematical models are a necessary and sufficient condition for science, but we humans also like to tell stories about what’s going on with the Universe, and so we often use models that might be less accurate as predictive agents but are more readily understandable given the story that they tell. We often start with a simple story that has only general applicability to Nature, and then gradually add complexity as the sophistication of the student, the practitioner, increases. When you hear things like “consider a spherical cow,” while a bit facetious, is the heart of the scientific process: Start simple with simple stories, and then work your way to a more exact answer.