Any scientist will tell you that testing forms the basis for legitimate scientific inquiry. So, how can we classify a theory based on alternate universes that scientists can never detect as science? The fact that multiverse models stand on the scientific frontier, where we have sparse hard data, makes answering this question even more difficult. Though many scientists disagree about the usefulness of models containing some form of a multiverse there are some reasons for including these models in the realm of scientific investigation, even though they are speculative (in the theoretical, not pejorative, sense) in nature.
First, multiverse models are not new to the cosmology scene. For example, shortly after Einstein developed the equations for the theory of general relativity, he realized that the solutions to those equations indicated we lived in an expanding universe. This meant the universe began to exist in the relatively recent past (few billion years ago) and was not eternal. Motivated by philosophical opposition to a genuine beginning, Einstein proposed a multiverse model known as the oscillating universe, where the universe alternated between expanding and contracting phases. It begins with a big bang, expands until gravity halts and reverses the expansion, then contracts until it ends in a big crunch. And then the cycle starts again leading to an infinite number of universes, one of which is where we reside. Eventually, calculations and measurements of our universe ruled out the oscillating universe as a viable cosmological model (although scientists have since proposed an updated version known as the cyclic universe).
Second, the current batch of multiverse models gained popularity primarily because they arose from investigations of other phenomena. Scientists did not simply invent a multiverse in order to explain away the beginning of the universe or to account for its life-friendly fine-tuning. The most popular multiverse model (a level II bubble multiverse filled with level I universes) arises from efforts to find an explanation for how inflation works. Granted the multiverse scenario arises after huge extrapolations of well-tested physical models, but most versions of inflation that produce a universe that looks like ours also produce a multiverse.
Third, some current multiverse models do make testable predictions. Stated another way, they have consequences in our universe that future measurements could validate or falsify. For example, some models predict that another universe might have collided with ours during its earliest phases. Such a collision would produce measurable signatures in the cosmic microwave background (CMB). Similarly, a multiverse would naturally cause asymmetries in the CMB that some scientists claim to have found.
Fourth, scientists recognize the need to find a way to test multiverse models although they disagree about whether they we will ultimately have the ability to conduct such tests. Distinguished cosmologist George Ellis rightfully argues that multiverse models require huge extrapolations from known physics and may undermine core scientific principles. Equally distinguished cosmologists Alexander Vilenkin and Max Tegmark agree about the large extrapolations but argue that multiverse models provide important explanations about the ultimate origin and character of our universe. theoretical physicist Sean Carroll argues that evidence has driven scientists to accept the idea of a multiverse and, as a scientific model, it is here to stay. In response, theoretical physicist and mathematician Peter Woit contends that the multiverse evidence rests on circular reasoning.
It seems that, as a whole, the scientific community remains agnostic about the existence of a multiverse. We may find evidence for or against it as scientists continue to investigate—or we may not. This uncertainty means that any attempt to declare the multiverse out-of-bounds scientifically is premature. Yet we must also approach the topic with appropriate caution lest we undermine the foundations of the scientific enterprise.