Actually, there was a fairly interesting argument for string theory that went like so (I'm citing this by memory, but it was occurrent in the literature at some point):
- General relativity is in formal logical tension with quantum mechanics.
- String theory is a unique and nontrivial solution to that tension.
- So all evidence for GR and QM is evidence for string theory.
Be that as it may, concerns about the viability of string theory as a scientific research program were acknowledged widely, and dissenters could publish books like Not Even Wrong without destroying their careers. In fact, theoretical physicists have proposed a number of alternative solutions to the GR-QM tension, proposals which are at worst labeled "fringe science" and not pseudoscience (the following list is not assumed to be completely exhaustive; it is a reflection of the alternatives that I am familiar with at this time):
- Loop quantum gravity
- Causal-dynamical triangulation
- Crystallographic cosmology
- Causal sets
One can also consider Max Tegmark and Stephen Wolfram's meta-theories about the mathematized universe (or multiverse). Now keep in mind, pseudoscientists will often use their theories to try to manipulate the scientific laity in various ways: astrologers for the sake of psychological and economic predation/parasitism, creationists for the sake of cognitive-religious predation. The star-seers try to scare people into making decisions, including life choices, based on the supposed imminent dangers posed by various celestial alignments; the false prophets issue warnings of moral corruption in this life and eternal torment in a future state.
By contrast, string or causal set theorists, say, do not tend to insist on people agreeing with them to such an extent, and might not be much of true believers in their very own ideas anyway: they are mainly, for lack of a better word, "nerds" who just enjoy the intellectual exercise involved in working through the mathematics of their models.
We should also consider that the boundary between a testable empirical hypothesis, and the abstract production of a mathematical concept, might be vague. In his work on the theory of "Ultimate-L" (I will leave the definition of that phrase for the reader to find), Hugh Woodin has framed an empirical prediction that an important component of his theory (see slide 11) will not be proven inconsistent within the next one thousand years. And so are predictions of (in)consistency a way to "empiricize" even the most esoteric mathematics? I don't know how string theory contributed to pure mathematics, but it is said to have done so in some interesting way, but then we can imagine that there are seemingly pure theories of mathematics with applications to physics that we can make predictions about the (in)consistency of in turn. (One might think that the deeper background for string-theoretic mathematics (topology, I assume) is immune to the flexibility and fluctuation of set theory, but note that Saharon Shelah has been dreaming of a way to force over arithmetical propositions that we might have believed (or hoped) were barricaded against forcing dynamics and so if we ever found a way to alter relatively basic arithmetic in this vein, who knows how we might alter the premises of empirically-applicable geometry/topology as well?)
ADDENDUM: The history of QFT in brief
I think getting at why modern theoretical physics is continuous with scientific empiricism as a historical program requires understanding how the Standard Model arose, how that evoked the question of supersymmetry, and how abstract models of symmetry conditions justified speculation about string theory even though no substantively unique evidence for string theory was forthcoming.
So, just to give a thematic overview of the history: in the late 1800s and early 1900s, scientists made great empirical discoveries about uranium (which had been known beforehand as e.g. "pitchblend," and used as paperweights for having a relatively boring metallic appearance (it is a metal, not glowing green gel, after all)), radiation, and stellar dynamics, that converged in the realization that uranium could be subjected to a process that might unleash a huge amount of energy, a process not entirely difficult to figure out. And all things considered, the even more ominous ability to use a random kind of metal (uranium again, also plutonium of course) to cause nuclear fusion in a random kind of water (heavy water, that is), which was gleaned from analysis of the physics "situation," was an even greater testament to the embyro of the Standard Model back then.
Correct me if I'm wrong, but I think I've read that one impetus for string theory was the practice of Feynman diagrams, since there is a not-mysterious lookalike aspect to the stringy things there and the general hypothesis of string theory (which is sort of intermediary between pure particle physics on the one hand, and unparticle physics on another). At any rate, the conceptual transition from point particles to strings was not an obscure, alien conjecture, but even the doctrine of point particles was itself already conjectural, for there is no absolute proof a priori that the units of empirical substance must be zero-dimensional rather than higher-dimensional, even within the dimensionality that humans can experience thoroughly from the inside.
Now, appeals to symmetry can be motivated by the weight of the prior evidence, which is that the mathematics of the experimentally (or: technologically) supported physics organized various factors according to discernible symmetry and symmetry-breaking parameters. There is an induction from that, if not to a further theory outright, yet to the string hypothesis (or, indeed, to any such hypothesis, i.e. any not-pseudoscientific extensions proposed for the Standard Model). More precisely, empirical information coming from astrophysics/cosmology is used to calibrate recognized deficits in the Standard Model.
Now that model is heavily grounded in particle-accelerator experiments, and we have strong operational knowledge about how to look for new particle information in those experiments. Thus far, we can justify a specific tabulation of particle types (elementary and combined). But the information we have about those types leaves out of view the evident cumulative impact of types that we have inferred can't be as readily detected by current accelerator technology. So balancing the microphysical and macrophysical evidence, we have an erotetic scheme encoded into the Standard Model, i.e. we can discern slots and metaslots in the particle table that need to be filled in (or else our theories about the known types will have to be seriously altered to accommodate the higher-level data).
Then we embark upon the quest for possible symmetry, sometimes "supersymmetry" is how they put it. This means doing a lot of peculiar, but extremely relevant and historically grounded, mathematics, hence we end up generating speculation about strings or world-crystals or stranger things besides (I myself am often struck by the peculiarity of the anyon/plekton hypothesis, which has been partly, but nontrivially, corroborated by now, as far as I know).