:: Article

String theory and post-empiricism

Richard Dawid interviewed by Richard Marshall.

Richard Dawid is always wondering about philosophical issues arising from physics and string theory, in particular the problem that string theory hasn’t been empirically tested, that it looks like it won’t be in the near future and that fundamental physics is entering a phase when empirical testing is increasingly difficult. He thinks about why physicists trust their theories, why some think this is no better than theology, why he doesn’t, why nuance in understanding underdetermination is required, about how a theory can be scientific without empirical testing, about whether such theories are strictly true, about why this doesn’t result in a constructivist, anti-realist position, about the status of string theory, about how physicists think about what they’re doing, about reliability, about the relevance of the discovery of the Higgs-boson, about how we’re entering a Kuhnian paradigm shift but only in physics and why reliance on non-empirical theory assessment is not a deficiency of soft sciences but integral to all scientific reasoning. Bazinga!

3:AM: What made you become a philosopher?

Richard Dawid: I started out as a theoretical physicist but always had an interest in philosophy. After my PhD, I spent two years as a high energy physicist at Berkeley, which was when I became increasingly fascinated by the epistemic and ontological questions raised by string theory in particular. Those questions seemed to me to have considerable philosophical relevance but were not seriously addressed by anyone in philosophy at the time. I ended up having some ideas how to deal with those issues and found those ideas more promising than the ideas I had in physics. I decided that one should follow one’s ideas, so I switched from physics to philosophy.

3:AM: A key question in philosophy of science is about how scientific theories work, what they are and what they can claim to show. Before we discuss your innovative theory about scientific methodology that addresses these issues can we start by giving us a brief intro to string theory, as that is the big theory that helps illustrate your ideas. Is it right, by way of a cartoon introduction, to say that string theory is the theory physics is using to bind up quantum mechanics (Schrödinger’s cat etc) with general relativity (Einstein)?

RD: That’s the basic idea, right. Finding a theory that consistently accounts for the quantum nature of microphysics as well as for the laws of general relativity that guide physics at cosmic scales may be the most burning question of fundamental physics today. To understand the full scope of string theory, however, one must know that our understanding of microphysics has moved far beyond the equations of quantum mechanics. We have come to understand that the particle content of the world as well as the nature of all nuclear interactions are in an intricate way related to internal symmetry principles. Gauge field theory describes those relations and currently constitutes our best conceptual understanding of observable microphysics. String theory unifies that level of our understanding of microphysics with general relativity.

The basic idea of string theory is to replace the point-like particles we assume in traditional elementary particle physics by extended objects, the so called strings. These strings are so small that our limited observational capacities based on present day high energy experiments perceive them as effectively point-like. All properties we attribute to the point-like particles are explained in string theory in terms of oscillation modes or topological properties of the string. That is, the dynamics of our observed world is at the most fundamental level explained by a purely geometrical theory of strings in space-time. All interactions, nuclear interaction as well as gravity, can be extracted from the dynamics of those strings.

3:AM: Ok, so the problem with the theory is that scientists can’t test it empirically. Is that right? If so, why can’t they, and is this situation confronting a scientific theory a new kind of difficulty not experienced by scientists before?

RD: We have no empirical confirmation of string theory at this point and the perspectives for getting empirical confirmation in the foreseeable future are dim. String theory does make predictions, though. It predicts that elementary objects are extended rather than point-like. Based on consistency arguments, it also predicts the existence of additional spatial dimensions which are curled up like tiny cylinder surfaces, so that moving along those dimensions leads back to the point of departure after an extremely short distance. These predictions do have characteristic empirical signatures one could observe in principle. The main problem is that the size of strings and extra dimensions most likely is so small that they will never be observable by a collider experiment, which is the type of experiment used today in the search for elementary particles. There is a second problem: string theory is not a fully understood and calculable theory at this point. This means that more specific predictions string theory might make with respect to empirically testable energy scales cannot be calculated today.

Of course it has happened many times in the history of physics that certain predictions of a physical theory could not be tested for a long time. There are also cases of theories or models that were not empirically testable at all for very long periods of time, sometimes until today. Still, two points are very different the case of string theory than in most historic cases. First, historic examples of theories that could not be tested empirically for a long time mostly were specific scientific models developed within a well-established general conceptual framework. The fundamental physical frameworks themselves, like Newtonian physics, electrodynamics, special and general relativity, quantum mechanics and non-abelian gauge field theory were either developed in order to explain specific previously inexplicable phenomena or empirically tested shortly after they had been developed. String theory is a theory that is placed at the most fundamental level of theory building and still remains detached from all empirical testing. Second, string theorists have a surprising degree of trust in their theory despite the lack of empirical evidence. Arguably, the last time a fundamental physical theory was so strongly believed in the absence of empirical confirmation was atomism in the 17th and 18th century.

So in certain respects string theory does represent a novel phenomenon in modern physics. That novel phenomenon transcends the case of string physics, however. Fundamental physics has generally entered a phase where empirical testing is increasingly difficult to achieve and often remains absent for many decades after a theory has been formulated. And there are quite some cases apart from string theory where physicists have considerable trust in their theories even in the absence of empirical confirmation. So looking at string theory can tell us something general about the recent evolution of fundamental physics.

3:AM: So how do string theorists justify any of their claims?

RD: String theorists justify their trust in the theory by the theory’s conceptual qualities and by characteristics of the research process. More specifically, one may isolate three elementary arguments and one more far-reaching consideration.

The three arguments are the following. First, string theorists point out that their theory is the only one that offers a concrete and promising idea for a consistent description of microphysics and general relativity. This argument hinges on a point I mentioned before: string theory is not just a unification of quantum mechanics and general relativity. There are other approaches, like loop quantum gravity, which address that problem as well. But string theory is the only theory that integrates into one overall theory our topical understanding of high energy physics based on gauge field theory and our understanding of cosmology based on general relativity.

Second, string theorists emphasize that their theory has given them far more insights and explanation than they could expect when the basic posit of string theory was formulated. The basic posit of string theory is very simple: elementary objects are extended and behave according to the laws of relativistic quantum mechanics. That posit was introduced as the basis for a fundamental theory of matter and all interactions because physicists came to understand that it automatically implied gravity and seemed to offer a solution to a deep technical problem faced by quantum mechanical descriptions of gravity. When investigating the theory that resulted from the basic posit of strings, it turned out that many other open questions of physics appeared in a clearer light. String physicists took that as a sign that they were on the right track.

Third, string theory is developed within the conceptual context of high energy physics, which has a strong record of predictive success. More specifically, high energy physics is a field where a lack of alternatives to a given theory in conjunction with the observation that the theory explains more than what it was initially built to explain up to now has been a good indicator of that theory’s empirical viability. This suggests that trusting a theory on that basis may make sense again this time.

The three described arguments for trusting string theory are of a quite general nature. Their applicability is by no means confined to string physics. Beyond those arguments, string theorists also resort to a more far-reaching consideration that crucially relies on the specific conceptual content of string theory. The internal structure of string theory gives reasons to believe that, once it has found a fully consistent formulation, it might be a final theory: no further, more fundamental physical theory shall ever be required in order to account for new empirical data. This final theory claim alters the understanding of string theory’s position within the research process and, in the eyes of many of its exponents, increases the probability that it is a viable theory.

3:AM: Lee Smolin thinks this is just self-deceptive ‘groupthink’. He doesn’t think string theorists are justified in making theor claims does he? Why aren’t the justifications string theorists use condemned as mere wishful thinking, or worse, a case of unverifiable or unfalsifiable religious piety, no better or worse than theology?

RD: Smolin and a number of other critics of string theory have quite vigorously argued that the string physicists’ trust in the viability of their theory is unfounded and constitutes an unfortunate deviation from the path of legitimate scientific reasoning. I think that those critics make two mistakes. First, they implicitly presume that there is an unchanging conception of theory confirmation that can serve as an eternal criterion for sound scientific reasoning. If this were the case, showing that a certain group violates that criterion would per se refute that group’s line of reasoning. But we have no god-given principles of theory confirmation. The principles we have are themselves a product of the scientific process. They vary from context to context and they change with time based on scientific progress. This means that, in order to criticize a strategy of theory assessment, it’s not enough to point out that the strategy doesn’t agree with a particular more traditional notion.

Second, the fundamental critics of string theory misunderstand the nature of the arguments which support the theory. Those arguments are neither arbitrarily chosen nor uncritical. And they are not decoupled from observation. String theory is indirectly based on the empirical data that drove the development of those theories string theory aims to unify. But more importantly for our discussion, the arguments for the viability of string theory are based on meta-level observations about the research process. As described before, one argument uses the observation that no-one has found a good alternative to string theory. Another one uses the observation that theories without alternatives tended to be viable in the past.

From the canonical perspective on theory confirmation, the problem with those observations is that they are not predicted by string theory. String theory itself did not predict that we wouldn’t find an alternative consistent description of gauge field theory and gravitation. And obviously string theory cannot predict the success rate of earlier theory building in high energy physics. The canonical understanding of theory confirmation holds, however, that only observations which can be predicted by the theory in question can constitute empirical evidence for that theory. And only such empirical evidence can confirm a theory.

I reject the very last point of this canonical understanding. I argue that observations which do not constitute empirical evidence for a given scientific theory can still make it more likely that this theory will eventually turn out empirically viable. This can be the case if the observations in question change our understanding of the overall research context in a specific way. The observations on which string theorists base their trust in their theory are of exactly that kind. There is an influential strand of philosophy of science, called Bayesian epistemology, that identifies theory confirmation directly with an increase of the probability that a theory is true or viable. Based on this understanding, we may say that the string theorist’s arguments do constitute a specific form of theory confirmation, which I call non-empirical theory confirmation.

The kinds of observation which provide non-empirical theory confirmation can constitute a serious and critical test of a theory for one reason: the same types of observation which can confirm the theory can end up disconfirming the theory as well. If alternatives to string theory were found or if other theories in high energy physics which did not have alternatives and were believed on that ground got empirically disconfirmed, observing those events would decrease our trust in string theory.

3:AM: There is always a problem with narrowing down to a single set of axioms that say that this is the true theory – and pessimists react to this by saying that theories are therefore always underdetermined and optimists resist by saying that underdetermination is too rare to worry about or Occam’s Razor sorts out the issue satisfactorily. You think there are problems with all these views don’t you?

RD: I think that a nuanced treatment of the question of the underdetermination of scientific theory building is of crucial importance for understanding the way we assess and use scientific theories. The basic question is: if I have developed a scientific theory based on a given set of empirical data, should I assume that many other scientific theories which are consistent with the given data could be built as well? We can distinguish two sub-questions here. First, should I expect many alternative theories that are empirically fully equivalent to the one I have found? Second, should I expect alternative theories that can reproduce the known data but give different predictions than my theory with respect to future empirical testing? We will only be interested in the second sub-question, because this is the one crucial for understanding scientific progress. There is a natural intuitive response to that question: of course we should expect many possible alternatives. After all, we view scientific progress as an infinite sequence of theory successions.

Scientists proceed by collecting new data that eventually force them to replace the old theory by an empirically more adequate one. The entire sequence of future new theories can then be understood as a sequence of alternatives to the initial old theory with respect to the initial set of data.

But things are not that simple. Think about the following scenario. Imagine, we have developed a theory based on known data that gives specific predictions with respect to an upcoming new experiment. Imagine further that the spectrum of possible alternative scientific theories were so wide that each possible outcome of that new experiment could be reproduced by a scientific theory that is consistent with the known data. In such a scenario, we would have no reason whatsoever to trust the predictions of our own theory. After all, why should we believe that our theory is viable but all those other theories, which in conjunction cover the entire spectrum of possible empirical outcomes of the future experiment, are not? The fact that scientists often do trust the predictions of their theories implies that they implicitly assume that there aren’t that many possible alternatives to it, or, as I call it, that the underdetermination of theory building is limited. How strongly limited it is, is difficult to assess. But, it can be assessed.

The crucial point in my conception of non-empirical theory confirmation is that all three arguments of non-empirical theory confirmation that I’ve described before rely on assessments of limitations to underdetermination. In effect, scientists infer the strength of limitations to underdetermination from observing a lack of known alternatives, the surprising explanatory extra value of their theory or a tendency of predictive success in the research field. Eventually, these observations amount to theory confirmation because strong limitations to underdetermination increase the probability that the known theory is viable. The connection between the number of possible alternatives and the chances of predictive success is intuitively most plausible when looking at the extreme cases: if there are infinitely many alternatives to choose from and just one of them is empirically viable, the chances to pick the correct one are zero. If there is just one possible consistent theory – and if I assume that there is a viable scientific theory at all -, the chance that the consistent theory I found will be predictively successful is 100 percent.

3:AM: So how can a theory still be scientific if we can’t empirically test it?

RD: I think that elements of non-empirical theory assessment play a much underappreciated role in all of physics, even in empirical theory confirmation itself. Assessments of underdetermination even enter our basic understanding of what it means to discover a microphysical object. This can be seen nicely when looking at the history of the concept of the atom. In the late 19th century, physics as well as chemistry provided a wide range of empirical support for the atomist hypothesis. On that basis, many physicists and chemists strongly believed that the atomist hypothesis was conducive to scientific progress. However, the understanding of theory confirmation at the time required direct observation of the objects posited by the theory. Atoms are not directly observable, which meant that even supporters of atomism mostly conceded that atomism could not be confirmed as a matter of principle. Atoms thus seemed condemned to remain in a kind of eternal limbo between the realm of pure speculation and the realm of confirmed theories.

One crucial reason why scientists and philosophers of science connected confirmation to direct observation at the time was the threat of underdetermination: if one could not directly observe the objects posited by a given theory (let us say an atomist theory), even strong predictive and explanatory success of that theory could not rule out the possibility of other theories that explained and predicted the same phenomena without those objects. And why should one assume that the known theory rather than one of those other theories was true?

In the early 20th century, something interesting happened. The growing spectrum of consistent predictions of atomism in conjunction with the growing and increasingly subtle explanatory power of atoms made the idea that a non-atomist theory could work just as well as atomism look more and more far-fetched. At some stage – around 1910 – scientists decided to decouple confirmation from direct observation. From that time onwards, if the web of arguments supporting the hypothesis was sufficiently tight, strong empirical evidence that was causally related to the posited object was taken to amount to a discovery of that object. And the corresponding theory was taken to be empirically confirmed.

Note that the arguments which brought about that shift were conspicuously similar to the three arguments for string theory today. No equally satisfactory alternatives to atomism had been found or were considered likely to be found. Atoms explained much more than what those who had introduced them into physics had aimed at explaining. And there was a long series of predictive successes (though, in this case, directly connected to atoms rather than just to the same research program as in the case of string theory). To be sure, there is an obvious difference between atoms in 1910 and strings today: atomism at that time was empirically confirmed according to today’s standards while string theory is not empirically confirmed today. But in order to call atomism empirically confirmed, one had to rely on arguments which were very similar to those which establish trust in string theory today: arguments of non-empirical theory assessment.

In other words, our modern notion of the empirical discovery of a microphysical object relies on the same kind of non-empirical theory assessment as the string theorists’ trust in their theory. The difference just is that in the case of the empirical discovery of microphysical objects non-empirical theory assessment works in conjunction with empirical evidence while in the case of string theory or other, comparable cases it stands on its own as non-empirical theory confirmation. The empirical discovery of a microphysical object obviously has a stronger basis than the trust in string theory. But one cannot reject the scientific legitimacy of non-empirical theory assessment without rejecting the scientific legitimacy of the empirical discovery of microphysical objects as well. And if one acknowledges the scientific legitimacy of non-empirical theory assessment in principle, there is no good reason for denying the legitimacy of non-empirical theory confirmation either. A coherent conception of the scientific process thus should acknowledge non-empirical theory confirmation as a valid form of scientific reasoning.

3:AM: Is the result of this a theory that is strictly true?

RD: The three described arguments of non-empirical theory confirmation don’t establish the truth of the assessed theory. One can understand this point most clearly by having a closer look at the process of the assessment of underdetermination. What is really assessed there is the number of possible scientific theories which give different predictions with respect to the next generation of empirical tests. (In the case of string theory, that would be an imagined first generation of empirical tests that can test the energy scale that corresponds to testing the string length.)

All three arguments of non-empirical theory assessment work on that basis. If scientists have found no alternatives to our theory, this just means that they have no alternative solution to a specific conceptual problem that is related to a given empirical context. If we notice a record of predictive success in the research field, we just talk about predictive success at the next steps of empirical testing. None of these arguments is capable of addressing the question how many alternative theories exist that can be empirically distinguished in principle by some so far unimaginable experiments in the unforeseeable distant future. I call the limited question addressed by the three arguments the question of local underdetermination as opposed to the question of global underdetermination that would address an infinite time-horizon of empirical testing.

The fact that the three arguments of non-empirical theory assessment do not address global underdetermination means that those arguments don’t address the question of the truth of the considered theory. After all, even if a conceptually completely different theory would supersede the present one after 1000 years, that would still mean that our present theory would have turned out false. The three arguments have a more modest goal than asserting truth. They just assess the theory’s chances of being successful at the next stages of empirical testing.

3:AM: Why isn’t this response a constructivist, anti-realist position? Aren’t you just saying that we can say the theory is true about the next theoretical layer to be tested empirically?

RD: Non-empirical theory confirmation does not imply scientific realism but it is nevertheless incompatible with a constructivist or empiricist position. Empiricism asserts that our scientific testing of theories relies exclusively on empirical evidence. If empiricism is right, we cannot know more about a theory’s epistemic status than what we learn from comparing the theory’s predictions with empirical data. Non-empirical theory confirmation, to the contrary, implies that there is another ways to learn about a theory’s epistemic status: by looking at non-empirical evidence.

This does have repercussions for the debate on realism. The empiricist argues that a realist position on scientific objects cannot have scientific backing for the very reason that science offers no strategies for assessing the status of a theory beyond the limits of what has been measured already. Kyle Stanford has framed that point in terms of underdetermination: science is good at building theories in agreement with the known data but it is not good at assessing the probability of unconceived alternatives to the theory that has been developed. Scientific realism, however, only seems plausible based on the assumption that few if any such alternatives exist. Otherwise, we should suspect that one of those alternatives is the true theory. If science is not in the business of justifying the assumption of no unconceived alternatives, scientific realism cannot be justified by scientific means.

Now, if non-empirical theory assessment plays an important role in the scientific process, this means that science is in the business of assessing the number of unconceived alternatives after all. Therefore, a fundamental objection to scientific realism looks far less convincing.

So the verdict on the impact of non-empirical theory confirmation on the realism debate is threefold. Non-empirical theory confirmation is incompatible with empiricism. It weakens a core argument against scientific realism. But it does not reach towards scientific realism because it addresses only local underdetermination.

3:AM: Like many people, I get my physics from watching ‘The Big Bang Theory’ so Sheldon Cooper is clearly to blame for any stupidity in my questions.(Think of me as Penny here without the looks!) But one thing I get from Sheldon is that string theorists argue about different geometries and topologies – the number of dimensions and all that. So how does your approach help sort out what we’re supposed to take as real if the theorists claim different numbers of dimensions?

RD: There is no doubt that Sheldon Cooper is a physically well informed character. Your question brings us to a more string-specific kind of discussion – which makes it a little tricky to explain the line of reasoning without entering the intricacies of the theory.

If we ask for the plausibility of a realist understanding of conceptual features of string theory, the theory sends us mixed messages. On the one hand, string theory contains an interesting element that seems to drag it closer to actual scientific realism. I have briefly mentioned that point already, it is the final theory claim suggested by string physics. The structure of string theory implies that we cannot learn anything new by considering physics at distances smaller than the string length. String theory tells us that all information about distances smaller than the string length is in an intricate way encoded already in information about length scales larger than the string length. This means that the string length in effect serves as a minimal length scale, which leaves no room for new physics that becomes relevant only when describing length scales smaller than the string length. In other words, string theory seems to suggest that it constitutes a final theory. Now to make a final theory claim based on a specific theory is tricky because it seems to presume already the absolute viability of the theory based on which the final theory claim is argued. I cannot go into any details, but I hold the position that there are deep connections between the question of limitations to underdetermination and the question of finality. Eventually, I argue, final theory claims can make sense in conjunction with the three described arguments of limitations to underdetermination.

If a final theory claim makes philosophical sense, it may be taken to imply the truth of the final theory and therefore to suggest scientific realism. Classical realists want to be more specific, however: they want to assert that the objects posited by a theory exist in the external world. And here we face a serious problem in string physics. String theory implies that certain different formulations of string theory, which posit completely different sets of fundamental objects and structures – in some cases even one string theory and one theory that doesn’t posit strings at all – are dual to each other. This means that they are empirically completely equivalent. An astonishing abundance of duality relations has been discovered in string theory, which arguably is one of the deepest physical messages from the theory. This, however, seems inconsistent with canonical forms of realism. It just does not seem to make any sense to single out a specific set of real elementary objects or structures in string theory because, whenever we have done so, we may switch to a different empirically equivalent picture where we find very different objects and structures. So string theory seems at variance with fully fledged realism. But it arguably suggests an understanding of physics where it makes sense to attribute truth to the overall physical theory.

3:AM: Do you think string theorists actually conceive of what they’re doing in terms you explain?

RD: The three arguments of non-empirical theory confirmation as well as the argument for a final theory claim are frequently applied by string theorists. Often they are used implicitly but sometimes they are explicitly spelled out. The philosophical reconstruction of those arguments in terms of limitation to underdetermination is not done by string theorists themselves. But I may say that some of them take it to be a plausible reconstruction.

3:AM: How reliable is this post-empirical method?

RD: Clearly, non-empirical theory confirmation is less reliable and more intricate than empirical confirmation. There is a significant risk of overextending the application of the involved arguments, which makes it important to reconsider those arguments in each individual research context. It will always remain the ultimate goal of science to achieve solid empirical evidence for a theory. But the question is: what can we do if empirical testing is not forthcoming for a long period of time? The answer I propose is that non-empirical theory confirmation is a scientifically viable second-best option for assessing the status of a theory. Understanding the role of non-empirical theory confirmation becomes particularly urgent in the face of the present situation in fundamental physics.

The canonical understanding of theory confirmation that takes all empirically unconfirmed theories to be mere speculations seems an increasingly artificial and inadequate characterization of the actual situation in fundamental physics. For most experts in the field, theories like string theory or cosmic inflation (an important cosmological theory that got strong empirical confirmation this year for the first time, 33 years after it was proposed) have told us something about our world for quite some time already despite the fact that they were supported by little or no empirical evidence. These theories have long become an essential part of the way we understand fundamental physics. I don’t suggest that one shouldn’t second guess the physicists’ trust in specific theories and check its plausibility in each individual case. But if one takes contemporary high energy physics and cosmology seriously, it does not seem plausible to reject the role of non-empirical theory confirmation tout court. A viable philosophical understanding of science in my eyes must be able to account for the physicists’ own perspective on their theories.

3:AM: Is the Higgs-Boson discovery helpful in showing how your approach to scientific methodology works?

RD: The Higgs-Boson is an excellent example of the high degree of trust physicists can have in a hypothesis in the absence of empirical testing. Even though the particle was only discovered in 2012, physicists were quite sure that a Higgs-Boson of some kind existed since the standard model of particle physics got empirically well confirmed in the 1980s. Once again, the reasons for that confidence fit nicely within the pattern of non-empirical theory confirmation. Without the Higgs-hypothesis, the standard model could not explain the masses of elementary particles. No other theory than the Higgs hypothesis was in sight that could satisfactorily explain those masses. In addition, the standard model itself provided a very strong example of a theory that was first understood to be the only available solution to a technical problem (that was the problem of renormalizability in the 1960s and early 70s) and then turned out to be empirically viable in many respects. So the trust in the Higgs Boson before its discovery provides a very nice example of arguments of non-empirical theory confirmation at work. Moreover, the fact that the Higgs Boson was then actually discovered in 2012 demonstrates that those arguments indeed worked very well in the given case: they led us to trust a hypothesis that eventually turned out trustworthy. The next time a new theory in high energy physics will be assessed based on the method of non-empirical theory confirmation, the Higgs-Boson case will itself serve as an indicator that the method is reliable.

Note that the Higgs Boson case also exemplifies nicely that non-empirical theory confirmation is a critical method. Let us assume for a moment that the LHC experiments had demonstrated that no Higgs Boson existed. This would not just have refuted the Higgs hypothesis but would also have considerably weakened the trust in the reliability of non-empirical theory confirmation in high energy physics. Therefore, it would eventually also have weakened the trust in string theory. People would plausibly have asked: how can you still believe in your arguments for the viability of remote string theory after your strong belief in the existence of the much less speculative Higgs Boson has turned out unjustified?

3:AM: Is your post-empirical approach just applicable to physics or are you arguing that it covers all science? And does this reflect that scientific territory is changing and moving away from understandings that can be empirically tested? Does this situation, and what your approach characterizes, characterize a Khunian paradigm shift?

RD: Let me start with the Kuhnian part of the question. Indeed, what is going on in high energy physics in recent years has many characteristics of a Kuhnian paradigm shift. What we can see very clearly is a split between two groups of scientists that endorse very different paradigms of theory assessment. On one side of the divide are most string theorists and physicists working in related fields of high energy physics and cosmology. They believe, based on their own scientific experience, that non-empirical theory assessment can play an important role and can justify strong trust in a theory’s viability in the absence of empirical testing. On the other side of the divide we find many physicists who work in more empirical branches of physics and many philosophers of physics. They feel uncomfortable with the high degree of trust in empirically unconfirmed theories because they don’t know that phenomenon from their own research context. The conflict between the two groups has become publicly highly visible due to the popular books by string critics like Lee Smolin and Peter Woit.

It makes a lot of sense to understand the rift between the two sides in terms of a paradigm shift that has occurred in a part of the physics community. What makes the Kuhnian perspective particularly instructive is the fact that the conflict is marred by misunderstandings that are caused by the different paradigms endorsed by the two groups. The two paradigms are based on conflicting understandings of what counts as a scientific success. Since each side argues based on its own paradigm, the debate on the successes of string theory must remain unproductive. The case of theory assessment in string theory differs from a classical Kuhnian scenario of a paradigm shift in one important respect, however: there was no revolution, no bold epistemic conjecture that led to the new paradigm of theory assessment. Rather, the new paradigm slowly emerged from the evolving research contexts of high energy physics through recent decades.

So the specific developments in recent high energy physics and cosmology can be characterized quite well in Kuhnian terms. But the described substantial shift in theory assessment remains confined to fundamental physics. Thorough empirical testing is still possible within reasonable time-frames in applied physics, chemistry and most other field of natural science. And the power of consistency arguments for limiting underdetermination has nowhere else become as strong as in high energy physics. So if we talk about a paradigm shift that establishes new scientific strategies of theory assessment, that paradigm shift is happening in one particular scientific field.

Nevertheless, being inspired by the case of string physics, the philosophical observer might consider a paradigm shift of a far more general kind with regard to the philosophical understanding of theory confirmation. I have already pointed out that non-empirical theory assessment has played a significant role in physics long before the advent of string theory and was unduly neglected by philosophers of science. If we look beyond the confines of physics, we find numerous examples of non-empirical theory confirmation as well.

In fact, while non-empirical theory confirmation is not considered “canonical” in physics, it is perfectly reputable in historical sciences like paleontology or archeology. Those fields are far less formalized than physics, which means that assessments of underdetermination work quite differently. Rigid consistency arguments, which play a crucial role in physics, are absent in those fields, which makes the exclusion of alternatives much vaguer than in the context of physics. But fields like paleontology of archeology systematically deal with scarce and insufficient data. Most of the evidence that could have supported or refuted a hypothesis has long decayed. Any appraisal of the viability of a hypothesis thus must involve an assessment of underdetermination of theory building by the scarce empirical evidence that happens to be available. If no good case for limitations to underdetermination can be made, a paleontological theory is pointless even if it reproduces the available empirical data: there just wouldn’t be any reason to believe in it. Historical sciences are merely the most obvious examples of this element of scientific reasoning. Many other applied scientific disciplines involve similar kinds of arguments. The canonical understanding of theory confirmation would suggest that the reliance on non-empirical evidence corresponds to the lack of formalization and rigor in those fields. The particularly strong role of non-empirical evidence in the conceptually most advanced parts of physics demonstrates that this perception is mistaken. The reliance on non-empirical theory assessment is not a deficiency of “soft” science but an integral part of all scientific reasoning.

3:AM: And for those of us wanting to delve further, are there five books other than your own that you could recommend we read?

RD: Henri Poincare 1902: Science and Hypothesis, Larry Laudan, Univ. of California Press 1977: “Progress and its Problems”, Brian Greene, Norton 1999: “The Elegant Universe” (not a philosophy book but a nice popular introduction to string theory), Craig Callender and Nick Huggett, Cambridge University Press 2001: “Physics meets Philosophy at the Planck Scale”, Kyle Stanford, Oxford University Press 2006: “Exceeding Our Grasp”.


ABOUT THE INTERVIEWER
Richard Marshall is still biding his time.

Buy the book here to keep him biding!

First published in 3:AM Magazine: Friday, July 4th, 2014.