There's a pretty good discussion of this problem here:
Physics strings us along, By Margaret Wertheim, Commentary, LA Times: There has been much talk of late about the scientific method, which usually takes place in the context of distinguishing science from other "less rational" practices, such as religion and magic. But in recent years science itself has been showing increasingly magical tendencies. In the field of theoretical physics, it is now common practice to talk about other dimensions of reality, entire landscapes of universes for which there is no empirical evidence whatever. ...The extra dimensions of string theory and the other universes they might entail have never been observed and, in principle, they may not be observable, at least not directly. At present they are pure fictions. String theory is so fecund in its descriptive power that one physicist has estimated there may be as many as 10 to the power of 100 different versions of its equations! Each one articulates a different set of possible universes and, at present, there is no way of determining if our universe matches any of them. Once upon a time, the sine qua non of scientific practice was supposed to be empirical verification. Experimental evidence was the core principle of Francis Bacon's much-vaunted "scientific method." In truth, the picture has always been more complex. Science is also an engine of the imagination, leading our minds beyond ... what is to ... what might be. Nowhere is the speculative dimension of science more prominent than theoretical physics, which has given us such magical possibilities as time machines made from spinning black holes, wormholes that become portals to the far ends of the universe and the "parallel worlds" of quantum mechanics, which, in theory, make every possible version of history a realized physical fact. The stories that theoretical physicists tell us are written in the language of mathematics, but for all its formal rigor, the science has become in effect a form of speculative literature. Unchained by the fetters of verification, string theorists are free to dream, articulating through their equations vast imagined domains in which almost anything that is mathematically possible is deemed to be happening "somewhere."
In the link above, Not Even Wrong states:
As I’ve explained in detail on other occasions, the simple fact of the matter is that string theory does not make any predictions, unless one adopts a definition of the word “prediction” different than that conventional among scientists. A scientific prediction is one that tells you specifically what the results of a given experiment will be. If the results of the experiment come out differently, the theory is wrong. String theory can’t do this, since it is not a well-defined theory, but rather a research program that some people hope will one day lead to a well-defined theory capable of making predictions.