In previous posts I responded (or gave other people’s responses anyhow) to the ideas that science is primarily about prediction, Reductionism, or Holism. In those ideas we found some truth, but not the whole truth.
Another common point of view is that science is really about observation. Related to this is the idea that science is primarily about empirical evidence or in other words must be falsifiable. As it turns out, these points of view are somewhat correct, but also misleading.
Science is Not Primarily Observation
I doubt science would have any meaning if we didn’t take the ideas of observation and empirical evidence seriously. Descarte is rumored to have tried to argue in favor of pure reason, but we know that this doesn’t work out in real life. The problem is that our reasoning capacity is too broad. We can think of logical possibilities that just happen to not exist.
In a past post (in the notes) I hinted at one of these: Cartesian Dualism – the idea that minds and matter are different things and that minds can exist without matter. (Seemingly contradicted by D&C 131:7). I can conceive of Cartesian Dualism, but all the evidence currently points against it. Plus Cartesian Dualism is a classic violation of Occam’s Razor. It only provides explanations by pushing the problems it purports to solve to a new location plus creating new ones. (This does not rule out the possibility that some form of dualism will turn out to be true, and in fact I believe this will turn out to be the case. But classic Cartesian Dualism seems a rational non-starter for me.)
Therefore we need empiricism and observation to test which of our ideas is correct. Observation and empiricism are therefore important parts of science.
But the simple truth is that you can have science without these. Often you have no choice because the technology hasn’t developed yet to observe predictions made by scientific theories. The Large Hadron Collider is a good example of this; it might cost billions to make an observation. Worse yet, some observations may forever be cost prohibitive.
Furthermore, sometimes the smallest of observations cascade into the largest of ground shaking conclusions. As David Deutsch points out,
Thus observations of ever smaller physical effects have been forcing ever greater changes in our world-view. It may therefore seem that we are inferring ever grander conclusions from even scantier evidence. What justifies these inferences? (The Fabric of Reality, p. 57)
But this presents a difficult with defining science through observation.
Science is Not Always Falsifiable
Karl Popper is famous for having introduced the idea that science should be falsifiable. This is often interpreted as a theory not being ‘scientific’ unless it’s possible to falsify it. Again, I think this is a powerful idea that has a lot of truth to it. Is there any doubt that we should not take Freud’s non-falsifiable theories seriously any more now that we have better falsifiable psychological theories? A falsifiable theory is automatically better than a non-falsifiable one because the most productive theories make falsifiable predictions so that we can, to a degree, check them for verisimilitude. (i.e. closeness to reality.)
But I doubt this is what really defines science either. It would seem there are too many counter examples.
One of my favorite authors, Roger Penrose, points out that Popper-style falsification is simply not enough to exclude a theory from science.
…Karl Popper provided a reasonable-looking criterion for the scientific admissibility of a proposed theory, namely that it be observationally refutable. But I fear that this is too stringent a criterion… take the example of supersymmetry in modern particle physics. …it is a central ingredient of string theory. It status among theoreticians these days is so strong that it is almost considered to be part of today’s ‘standard’ particle-physics model. Yet, it has no (serious) experimental support… The theory predicts ‘superpartners’ for all the observed fundamental particles of Nature, but none of these has so far been observed. The reason [given]… is that a symmetry-breaking mechanism (of unknown nature) causes the superpartners to be so massive that the energies needed to create them are still beyond the scope of present-day accelerators. With increased energy capabilities, the superpartners might be found… But suppose that still no superpartners are actually found. Would this disprove the supersymmetry idea? Not at all. It could (and probably would) be argued that there had simply been too much optimism about the smallness of the degree of the symmetry breaking, and even higher energies would be needed to find the missing superpartner. [So] we see that it is not so easy to dislodge a popular theoretical idea through the traditional scientific method of crucial experimentation, even if that idea happened actually to be wrong. (The Road to Reality, p. 1020-1021)
Penrose goes on to use another example. One of our current theories predicts the existence of “monopoles.” Imagine a magnet that has only a north or a south pole and not both. If even one exists in the entire universe, this theory has been vindicated. But if we never find any it’s not really that shocking because what are the odds we’re going to find that one monopole somewhere out there in the entire universe? So here is yet another non-refutable theory. But few would argue it isn’t science.
Even Popper seems to disagree with this supposedly Popper-inspired idea, or at least he never says something isn’t science just because it can’t be falsified. Rather Popper only created a demarcation between what he calls ‘empirical theories’ and non-empirical ones. He goes on to say:
This criterion of demarcation between empirical and non-empirical theories I have also called the criterion of falsifiability or the criterion of refutability. It does not imply that irrefutable theories are false. Nor does it imply that they are meaningless. But it does imply that, as long as we cannot describe what a possible refutation of a certain theory would be like, that theory may be regarded as laying outside the field of empirical science. (The Myth of the Framework, p. 88)
Therefore a theory that is not refutable is only outside a category he calls ‘empirical science’ not necessarily science in general.  In fact, Popper points out that one of our most important theories – The Theory of Evolution – is based on a non-falsifiable core tenant; namely survival of the fittest:
There is a difficulty with Darwinism… it is far from clear what we should consider a possible refutation of the theory of natural selection. If… we accept the statistical definition of fitness which defines fitness by actual survival, then the theory of survival of the fittest becomes tautological, and irrefutable. (Myth of the Framework, p. 90) 
Are we prepared to dismiss the Theory of Evolution from the realm of science on the grounds that it’s central tenet is irrefutable? I would hope not.
So we see that ‘science’ is not primarily either observation nor falsification.
 Popper admits people misuse his epistemology based on faulty understandings of it. Using Pauli’s theory of the neutrino as an example, he says:
When this theory was first proposed by Pauli, it was clearly not testable. It was even said, at one time, that the neutrino is so defined that the theory cannot be tested. About thirty years later the theory was not only found to be testable, but to pass its tests with flying colours. This should be a warning to those who are inclined to say that nontestable theories are meaningless (a view which has often but mistakenly been attributed to me) or that they have no ‘cognitive significance.’ (Myth of the Framework, p. 88-89)
 It is my understanding that Popper at least partially retracted this quote later. I haven’t gotten that far in my reading of him yet. Truth be told, I do not think the theory of evolution could be said to be in any way non-falsifiable. But I do tend to agree that ‘survival of the fittest’ is circular reasoning. But this does not make the Theory of Evolution not a scientific theory, as Popper may have (at the time) been suggesting. It just means that we have to test other aspects of it based on that explanation.