We assume theories need experimental proof to be credible. But philosopher of science Richard Dawid argues that string theory has earned trust through a different route: meta-empirical assessment. When decades of searching turn up no viable alternatives, and unexpected predictions keep emerging, a theory's viability can become probable even without testing. Empirical confirmation isn't the only path to scientific credence.

For many decades, theoretical physicists have aimed at developing a theory that jointly describes quantum phenomena and gravity. String theory, which is widely considered to be the most promising candidate, takes gauge field theory, the conceptual framework for contemporary particle physics, as its starting point and develops a theory that includes gravity from there. Other approaches, pursued by smaller numbers of physicists, choose a more modest starting point and deploy the basic principles of quantum mechanics for developing a theory of gravity.      

Theories of quantum gravity face two serious problems that set them apart from well-established theories like gauge field theories or general relativity.  First, their characteristic empirical predictions are so difficult to test by experiment that no such empirical tests have been done or can be expected in the foreseeable future. Second, the theories are conceptually and mathematically so difficult that decades of intense work have not even come close to formulating a complete theory. 

___

Arguments in support of string theory’s viability don’t look weaker today than 25 years ago.

___

In light of these problems, the question is: what can be the status of the current theories of quantum gravity, in particular of its leading exponent, string theory? Views on this question have varied a lot during recent decades. The 1990s saw a high level of confidence among string theorists. They presented their theory as a decisive breakthrough in the human understanding of fundamental physics, had high confidence in the theory’s viability, and were optimistic about imminent substantial steps towards a full understanding of the theory. The early 2000s witnessed a highly visible pushback by physicists from rivaling research programs who doubted the scientific basis for string theorists’ optimism and promoted the idea that, in the absence of empirical testing, all theories of quantum gravity should equally be considered mere speculations.

The last 20 years saw substantial changes in the understanding of what string theory amounted to, but also a significant slowdown of conspicuous conceptual progress towards a full understanding of the theory. Among many string theorists, what was in retrospect viewed as previous overconfidence with respect to completing the theory has led to a more hedged approach towards characterizing the theory’s status in public. Today, the dominant attitude is to avoid any impression that string theorists judge their theory by standards that differ from the canonical principles of empirical confirmation. Working on the theory, therefore, is often motivated by pointing at its pursuit-worthiness rather than by emphasizing the theory’s prospects of being correct.

It seems plausible to say, however, that, while expectations in the late 1990s regarding the prospects of approaching the completion of string theory have turned out overly optimistic, arguments in support of string theory’s viability don’t look weaker today than 25 years ago.

In my book String Theory and the Scientific Method (Cambridge University Press, 2013) and in more recent work, I have proposed and developed further the concept of meta-empirical theory assessment (MEA). MEA is a mechanism that, under specific circumstances, can generate significant credence in a scientific theory even if the theory’s characteristic predictions have not been empirically confirmed. The claim is not that, based on MEA alone, a theory can reach a status comparable to that of an empirically well-confirmed theory.  But an empirically unconfirmed theory can, based on MEA, be considered viable with fairly high probability. Calling such a theory a mere speculation would then be a serious misrepresentation of its epistemic status. String theory, I claim, is the best example of such a theory. And string theorists, the claim continues, do implicitly apply MEA-type reasoning when assessing the status of their theory.

SUGGESTED READING In defence of string theory By Juan Maldacena

The relevance of MEA is by no means confined to empirically unconfirmed theories like string theory. Quite to the contrary, it is a universal feature of scientific reasoning. Let us think about an empirically well-confirmed theory, general relativity. The theory’s predictions have turned out to agree with empirical data in a wide range of experimental contexts. But why should that mean that the theory’s predictions will again agree with the data we collect in the next experiment? Note that this is not a question that can be answered by just looking at general relativity itself. To answer the question, we need to consider possible alternative theories that could represent all the old data just like general relativity but make different predictions regarding the new research context. If we knew any such theory, we wouldn’t be confident about the new predictions of general relativity. So, the first answer to our question is: we trust the predictions of general relativity in the new research context because we do not know any theory that would make different predictions. But the question remains: why don’t we assume or at least suspect that there are such theories, and we just don’t know them? This is the question MEA aims to answer.

MEA offers three types of arguments to support the hypothesis that there are no such unconceived alternatives. Those arguments are based on observations, but not of the kind that could be predicted by our scientific theory. Rather, they are based on meta-level observations about the research process that have implications for our expectations regarding the number of unconceived alternative theories.

Argument 1 is based on an observation quite similar to what we have stated above already: even though scientists have tried hard to find convincing alternative theories to general relativity (in the regime where quantum effects are not important), they haven’t found any. This observation can be the basis for suspecting that, maybe, there are no such theories. But one might also suspect that scientists might just have overlooked them. So, one needs some further arguments. MEA provides a couple of those.

Argument 2 is based on the observation that general relativity made correct predictions of data that had not been relied upon when developing the theory. If there had been lots of unconceived alternatives making different predictions in that respect, physicists would have been amazingly lucky to develop exactly the one theory that made the correct predictions.

Argument 3 is broader but therefore also less focused: one observes the general tendency in physics that theories that seem to be without alternatives, when tested, are predictively successful. So, it seems plausible to infer that not finding promising alternatives to a theory is a good indicator of a scarcity of unconceived alternative theories. All three MEA arguments in conjunction can be very powerful, and can and do generate very high trust in predictions of general relativity in a wide range of new research contexts.

___

The full set of MEA-type arguments can be deployed in the context of string theory, which makes the theory far more trustworthy than a mere speculation.

___

The crucial point now is that the described lines of reasoning do not necessarily depend on the empirical testing of the theory in question. String theory has not been empirically tested, but it aims to cover general relativity and gauge field theory (or, more specifically, the standard model of particle physics), two theories that were thoroughly tested. So, string theorists can have trust in their theory if they see reasons to assume that there are no alternative theories to string theory that can do that job convincingly. They can, in principle, deploy argument 1 of MEA in full analogy to the case of an empirically confirmed theory. They can also reason along the lines of argument 3 to support their claim. Argument 2 cannot be directly applied because it would require previous predictive success of string theory, which has not been achieved. This arguably is the main reason why MEA can never be as powerful in the absence of empirical confirmation of the given theory as with regard to an empirically well-confirmed theory. It turns out, however, that a more abstract variant of argument 2, which replaces novel confirming empirical data by unexpected explanation at a more conceptual level, can be used and plays a very important role in the context of string theory. A wide range of unexpected explanations and interconnections have been found in string theory over decades and provide the basis for string theorists’ trust in their theory. These findings are less powerful in supporting the theory’s viability than precise and correct empirical predictions would be, but they provide a basis for the same type of reasoning and therefore are highly significant.

We thus see that the full set of MEA-type arguments can be deployed in the context of string theory, and makes the theory far more trustworthy than a mere speculation. How much trust in the theory is justified on their basis is a matter of careful analysis and needs to be continuously assessed by the physicists involved. But a plausible scenario may be the following: string theory is a theory that is far from completion and will remain so for the foreseeable future. The understanding of what the theory amounts to has changed dramatically during recent decades and most probably will do so in the future as well. But even though no one today fully understands string theory, what is known about string physics today are elements and aspects of a theory that, if transformed into a consistent whole, would amount to a theory about the world. Based on the arguments sketched above, that theory can be expected with high but limited probability to be viable.

Two main characteristics distinguish the assessment of the status of string theory based on MEA from the standard view of conclusive empirical confirmation: first, MEA is based on an assessment of the chances for the existence of alternative theories that have not been found; and second, it treats a high but limited probability of a theory’s viability in a given regime as an important and potentially longstanding judgement about the theory’s status rather than a fleeting and somewhat insignificant transition period on the way to conclusive empirical confirmation. Both points may look non-standard in the context of physics, where aiming at precise empirical testing is the norm.

SUGGESTED VIEWING The trouble with string theory With Katie Robertson, Roger Penrose, Brian Greene, Eric Weinstein, Tasneem Zehra Husain

It should be noted, however, that both points look far less unusual in other scientific fields. The hypothesis that a meteorite impact caused the extinction of the dinosaurs, to give one example, underwent a decades-long transition from a mere speculation to an increasingly trusted but not conclusively confirmed hypothesis, all the way to the near universal acceptance it enjoys today. During that transition period, an adequate understanding of the hypothesis’s status required acknowledging that a theory can be much more than a speculation without having found conclusive confirmation. To be sure, a lot of confirming empirical data was involved in this process. But equally important was a growing awareness that there probably was no unconceived alternative theory that could provide the correct explanation instead of the meteorite hypothesis.    

Critics of MEA have suggested that, if followed, it would disfavor work on alternative research programs in a scientifically detrimental way. If MEA is handled with care, this is not the case for two reasons. First, as already stated, responsible use of MEA would not lead to full trust in an empirically unconfirmed theory. It would rather make explicit and thereby more transparent considerations on the prospects of research programs that happen anyway. Second, MEA itself depends on the serious investigation of scientific alternatives. If no such investigations were pursued, the argument that no alternatives have been found would have no bite. MEA thus actually adds a new motivation for pursuing a wide spectrum of scientific approaches: pursuing them is essential for having any degree of trust in a theory that may, in the end, emerge as the leading competitor. 

Making explicit the reasons why string theory has a fairly high chance of being correct is so important because, in the current situation, arguments of pursuit-worthiness would be fairly weak without an epistemic component. If thousands of physicists working on the question of quantum gravity for over half a century hadn’t even succeeded in coming up with a theory that could with some confidence be expected to be correct, this would in itself be a substantial drag on the field’s pursuit-worthiness. The inspirational appeal of thinking about the most fundamental physical principles behind the world we live in could not be upheld if the message was that, while it would be great to find them, no one has any idea whether what has been developed in the last half century will stand. Fortunately, the actual situation, though complicated, seems far brighter than that. It is important to clearly spell out that it is.