Forget the epistemic demarcation problem: it's a fuzzy ill-defined boundary.

The so-called demarcation question between science and non-science is a human invention. A natural epistemology imposes on us objects of differing epistemic distance resulting in no clear cut-off  allowing us to define a sharp category distinction between objects which are the subject of  science and objects which are not.  Some objects are readily amenable to observational testing. Other objects only lend themselves to retrospective best fit explanation analysis. And in between these extremes are many shades of grey as determined by epistemic distance.

The de facto Intelligent Design web site Uncommon Descent has done a series of posts (by "News", aka Denise O'leary) informing us that at least some people in the science establishment are trying to circumvent the rigorous strictures of science by suggesting that highfalutin theoretical science need not be bound by the criterion of "falsifiablility". Falsifiability" is, in my opinion, a poor term because  it is seldom, if ever, possible to absolutely falsify a theory;  the human imagination is fertile ground for the conjuring up of quite bizarre and baroque explanations which retrospectively save a theory from absolute falsification. So, much better to my mind is to talk of the observation based  testing of theories rather than falsifiability as the main criterion defining science. With observation based testing a theory's ability to predict outcomes is assessed. For reasons given in this recently revised paper, successful prediction enhances a theory's probability of being right. So the criterion for rigorous science is not "falsifiablilty" but rather potential for "prognostication". But if a theory continues to fail in its predictions and these failures have to be constantly shored up by retrospective ad hoc narratives which attempt to explain away the failure, one's suspicions about the theory start to grow. Like any prognosticator no theory gets a blank cheque of belief. Often when people talk about "falsification" they really mean observational based testing and therefore I will proceed on understanding that this is what is meant by"falsification".

I have followed O'leary's  posts with interest. The latest post (by "News") can be see here:

https://uncommondescent.com/philosophy/does-a-fetish-for-falsification-and-observation-hold-back-science/

This post is entitled Does a "fetish for falsification and observation" hold back science? In it O'leary quotes astrophysicist Adam Becker who proposes that the falsifiabilty criterion be relaxed on the basis that falsifiability can never be absolute. Becker then goes on to give the example of the orbit of Uranus which at first appeared to falsify Newton's theory of gravity. But contrary to what Becker is leading us to expect this anomaly wasn't explained away with untestable special pleading; rather it became the basis of a prediction of the position of Neptune, a prediction which was tested by astronomical observation and proved correct. Becker's example is therefore a poor one; he really needed to show how an apparently falsified theory can be endlessly shored up with the addition of ad hoc narrative. Instead he gives us an example of rigorous science in action!

I think O'leary (and perhaps Becker as well) have missed the nuanced distinction between falsifiability and testability - they are not quite the same. However, I probably do agree with the general drift of O'leary's thought: Theories that either can't make predictions or whose predictions fail do don't have the authority of those that successfully prognosticate.

Going on what O'leary has written on this topic in past she has more than hinted that the attempt to move away from so-called "falsifiability" (i.e. observation based testing) is motivated by those working in the fields of string theory and multiverse theory, theories which are not (as yet) amenable to observation based testing (and the prospect is they may never be!). Currently these theories have more the character of retrospective sense making structures which attempt to explain a body of accepted data. Naturally enough theoreticians who have a vested interest in these theories might like to promote them as prestigious hard science; a status they cannot have while they fail to generate forward looking predictions.  

Although I probably fall on the side of O'leary on this one, there is an aspect of the attempt to broaden the epistemic remit of science that I sympathize with. As I have said so often on this blog, in the face of a spectrum of epistemic distances, we find that at increasing epistemic distance retrospective fitting of a narrative to the data in hand may be all we can do. I'm sympathetic, but only provided we carry this retrospective analysis through with self-awareness,  epistemic humility and with the appropriate disclaimers attached; especially so as the human mind is adept at inventing plausible retrospective explanations that really have the status of mythology; mythology acts as a kind of "coping" device; it provides a background story that we find amenable to our emotions and outlook.  

String theory and multiverse theories can be expressed with some mathematical precision and yet in spite of that their supporting world view narratives can be decidedly fuzzy.  Take for example the justification given by Brian Cox for his favoring of the many worlds hypothesis. Fair enough Brian, one can understand the compelling nature of the mathematical symmetry which leads one to favor the multiverse hypothesis, but in the final analysis it is the same idea which motivates Max Tegmark's extravagant mathematical universe. Some of us may not find symmetry so compelling that we feel the need to elevate symmetry to a "sense making" world view. 

Epistemic distance:The identity of Jack the
Ripper is probably beyond recovery,
Nevertheless some theorists think they
know the answer!

Nevertheless I really have no objection to people who attempt to join the data dots with a speculative  theory; it's a line of business I'm heavily into myself. As I've already said epistemic distance is often such that "best fit" analysis of data already in hand is the best we can do and it may be that like string theory the "best fits" are very intuitively compelling. I wish string theorists well in their endeavors.

The trouble arises, however, when attempts are made to foist these theories on us as definitive and authoritative, giving them the kudos of observation tested science. This should certainly be opposed. This kind of bumptious world view attitude has a track record of trying to gain an intellectual hegemony; it's almost as if their backers, in the absence of predictive evidence, feel the need for universal assent to make good that absence. We've seen it time and again; Marxism, Nazism, Maoism, fundamentalism...all headed up by intellectual control freaks who claim to know the authoritative answers.  Perhaps this is the danger that O'leary has in mind; if so I agree with her. When a world view comes with exclusive authoritarian backing you know you're in for a rough ride; I've seen far too much of that among Christian communities of one form or another.