Scientists follow what's now called postpositivism. The modern formulation was created by a guy named Karl Popper, and thus it's often called "Popperian postpositivism", but it dates back to Newton, who phrased it such: hypotheses non fingo -- "I do not frame hypotheses", but less literally, "I do not engage in untestable speculation".
Most formal logic you read about talks about evidence supporting theories, but in science, it's the other way around. Reality provides evidence, that we need to describe. We call these observations and phenomena.
What we call theories, scientific theories, describe the evidence. When a theory is disproved, it failed to describe the evidence accurately, and has to be replaced by a new theory. This does not change the fact that evidence exists, and has to be accounted for.
All scientific knowldege must be testable, repeatable, and disprovable.
If there is no possible situation where your theory can be disproved, then your theory cannot be used for anything useful, and cannot be improved, because you will have a retort for any criticism, every piece of evidence, no matter how valid, and no matter how many contortions you use. There must be defined criteria by which a theory is disproved, because that's the only way it can be improved.
An accepted theory is thus just one that describes the universe accurately. Human shit included. Disproving a theory doesn't alter the universe, it just means our abstractions were wrong.
Note, however, that disproven theories are often retained as easier to use approximations. Newtonian physics is an approximation of Einstein's general relativity...which turned out it turn to be incompatible with quantum mechanic's Standard Model, which is conversely incompatible with general relativity. Not to mention that the SM says gravity dosen't exist, which is plainly not true. Searching for a theory to merge them is a goal of physics.
Now, when I say that a scientific theory must be testable, I mean it. If it can be disproved, it must be done through evidence, not by pure reason, as Kant would call it. You can't learn about the universe simply by thinking really hard about it. (Ever hear of the book Critique of Pure Reason? The title might make more sense to you now.)
The adoption of Popperian-Newtonian science eventually caused Freud to become known for being a bunch of horseshit in psychology, and linguistics is increasingly in that vein as well. Many of, say, Chomsky's assumptions have been increasingly contradicted by reality, as have the derivative theories like universals, which turned out to be more "tendencies". Even things like ethics, considered the domain of philosophy, may be solved by the scientific application of game theory, evolution, and psychology.
Now, here's a thing you have to know: most scientific research findings are wrong. Any value can vary by the testing set up, possible researcher bias, and by pure chance, and a piece of research may get an inaccurate result. Thus science must also be repeatable, a theory must be tested multiple times, to ensure that the observation was accurate. This takes a lot of work. Complicating the matter is when news media fails to understand this and presents the new observations as facts, rather than another piece of data to be considered.
These criteria, testability, repeatability, and disprovability is how scence works. You may heard of Occam's razor. Well, this is a much more fine and ruthless razor -- Alder's razor, or, Newton's flaming laser sword.
Philosophers often disagree with Popperian/Newtonian philosophy, but to be frank, it produces results. We know of no other philosophy that allows us to accurately describe the world and gets us the information we want to do things, to describe the universe accurately. We might be wrong, but we don't seem to be wrong just yet.