I don't really get Kuhn's approach to the demarcation problem. I can throw around terms like "paradigm shift", "normal science", "revolutionary science" and "incommensurable", but I don't really understand how Kuhn arrived at his result that science is defined sociologically. In this vein, I'd like to know what is the problem with falsificationism that Kuhn's approach solves?
In Kuhn's own words, "Sir Karl mistook what happens during scientific revolutions for the normal practice of scientists... it is normal science, in which Sir Karl's sort of testing does not occur, rather than extraordinary science which most nearly distinguishes science from other enterprises”". In other words, falsificationism overemphasized theory testing and revision, whereas during "normal science" periods it is application of the "paradigm" that dictates scientific activities, not attempts to falsify it (according to Kuhn). As a historiographer, Kuhn saw falsificationism as romanticizing science, which is why it offered a poor account of its historical development, not to mention false positives and false negatives in demarcation, especially on young and non-mathematized sciences.
Kuhn describes science during normal periods as "puzzle-solving", not only does the paradigm remain unchallenged, it has to be presupposed to even make sense of the puzzles, e.g. classical mechanics has to be presupposed in measurements of masses, forces, etc. So astronomy was a science since inception because fitting observed motions of planets to homocentric spheres or deferents/epicycles were puzzles to solve. Despite the astronomers normally not trying to test or disprove spherical or epicyclic schemes, which made their puzzles possible in the first place. But astrology never was a science because there was nothing to constructively adjust in the tradition to make more accurate horoscopes, there were no puzzles.
Needless to say, "Sir Karl" begged to differ, calling puzzles "minor problems", and Kuhn's demarcation criterion "major disaster" that replaces "rational criterion of science by a sociological one".
The central notion of Kuhn is that of 'normal science', that the goal of a paradigm was actually to reduce the practice of science to a kind of stability that allowed its primary work to be solving technical puzzles that could be stated in ways accepted by the entire community.
If this is what most scientists do most of the time, what happens between periods of such stability? Historically, what happens in those periods between convergent puzzle-solving looks much more like politics or religion than like logical argumentation. Appeals to intuition abound, and formulations gather adherents, rather than data.
So he characterized science as alternating between these two extremes. In each extreme, the notion of what failure means is completely different. In both cases it has something in common with Popper's idea that theories exist to be challenged, but in neither case do successful challenges actually cause theories to be abandoned.
During normal science, the not-yet-covered part of the domain is constantly predicted wrongly by the theory at hand, and yet the theory remains largely intact, making finer elaborations or very small adaptations. Faith exists that some minor change in perspective, some intricate refolding of existing work, or some unaccounted fact will explain the failure. So falsification is continual -- and irrelevant.
Between periods, when new sets of groundwork are being proposed, each set acknowledges its own incompleteness, so it purposely does not try to cover large chunks of the domain that it knows it is not fit for. So falsification remains unattempted -- and irrelevant.
The point at which falsification matters is late in a 'revolution' as a paradigm comes together. And its primary application is to make sure the paradigm does not so thoroughly foreclose the imagination that those proposing it could never be made to abandon it.
It is a challenge of the imagination that prevents the ascendance of overly-philosophical attempts at science like Alchemy or Freudianism, which are so flexible they cannot be challenged, and will occupy much more time in ascendency than they deserve, in a counterproductive manner that hinders overall progress.
So although Popper has identified something extremely important, from a Kuhnian point of view, it is not useable most of the time, and surely does not apply to each individual observation in the way its originator imagined it should.
Given the ascendency of statistics, with the habitual framing of he null-hypothesis, it is easier for us to imagine or pretend that we are producing science with the intention of challenging it continually. But even though we inject it at the very base of our consideration, this is actually not true at any important level.
We still take overall theories that frame our experiments as whole systems of belief, which we do not set aside for lack of support. When a researcher finds lack of support for an accepted paradigm, it is very seldom the paradigm, and much more often the researcher who is declared to have failed.