I've travelled different countries
Travelled to the furthest lands
Couldn't find nobody could tell me
What is the soul of a man
Blind Willie Johnson
The conventional idea of science is that scientists have hypotheses, design experiments to test these hypotheses and then, if the experiments do not show that the hypotheses are wrong they say that they have a "scientific theory".
To some extent this is true but what often happens in the lab is that the boss buys a new piece of equipment that few people have used before and the scientists point this bit of kit at anything that takes their fancy. The observations are new so the lab can publish loads of papers, occasionally linking their observations to current theories. This doesn't just happen with new instruments, for instance, in a drug company the chemists are making thousands of new substances and the biological scientists screen these. If one of the chemicals cures a disease they publish the story of their work.
Scientific progress is often about observation with a bit of theory at the back of things. Even the theories are usually empirical in the sense that they do not reach back into the fundamentals of quantum theory and relativity but instead describe how sets of observations fit together. For example, the theories about the human immune system are incredibly complex but actually summarise thousands of observations. They are Empirical Theories.
Usually it is the other sort of theory that grabs the headlines, the non-empirical Platonic Theories. Platonic Theories start from simple assumptions about space, time and matter and attempt to explain the world in terms of the mathematics of these simple building blocks. 90% of working scientists know little about the inner workings of these Platonic Theories and don't need to know about them to do their job. An ecologist is interested in how the changing fly population affects the local frogs, not the quantum mechanical basis of fly capture!
Empirical theories differ from Platonic theories because the former are really just descriptions of the world whereas the latter are a set of absolute rules. Empirical theories are recipes for retrieving observations or repeating a set of events. There is normally an implication that the empirical theory will apply if the conditions of the original observations that gave rise to the empirical theory are present. This limited scope means that empirical theories can only be invalidated by demonstrating that the original observations were dishonest. In contrast, in science, Platonic theories are invalidated if there are any circumstances in which they are contradicted by empirical observation.
Science is largely an empirical art, it is the art of observation and description. The inferences that go between the observations are interesting but frequently of secondary importance.
The Platonic and empirical collide where there are deeply held convictions about Platonic theories. The most obvious example of this is the "mind". The empirical description of the mind is that it is like a set of things laid out around an observation point (Descartes' "res cogitans", the unextended viewing point). Look around, imagine things or have a day dream, see, the mind is like this. Nineteenth century Platonic theories of physics cannot explain such an observation so those Platonists who use nineteenth century ideas deny that the mind exists (and the Dualists say that the mind obviously exists so "science" is wrong).
We know that we have minds, just look, but if we believe strongly enough in nineteenth century Platonic theories of physics we can persuade ourselves that these minds do not exist. The fact that theories can be created is wonderful but if they contradict or forbid observation they are of limited applicability or wrong and are certainly not scientific.