Science is empirical. Create a hypothesis and test it. If it matches the actual behaviours observed, then it becomes a theory (I suppose - I've always thought of them just as the best current hypothesis). Keep trying to break it, and if you can, start again with a new or refined hypothesis.
There's a hypothesis at the moment that the reason gravity is so weak (compared to magnetism, for example) is that gravity is dispersed through multiple dimensions. The next trick is to come up with a means of testing this empirically.
Going to be slightly off-topic here, but I can't resist. When it comes to physics at least, a theory (which, by the way, I think is a poor choice of word) is more of a scientific framework, starting with some axioms, then analysing their consequences, and then making testable and falsifiable predictions, which can then be experimentally checked. That's where the hypothesis part comes in.
For example, the general relativity in a nutshell starts with "what if space and time were actually together as spacetime, and matter would curve this spacetime?" Then as a consequence, not only would gravity arise from said curvature, but light would also bend while traveling through this curved spacetime. Working on that some more, we can come up with a prediction of how much would the Sun bend light passing near it, like light reflected from Mercury. As it turns out, the predicted amount differs from what Newtonian mechanics* would predict for the same experiment: so, by measuring it when it actually was possible to do (comparing the position of Mercury in the night sky with the position of it during a solar eclipse), we knew which theory was closer to reality.
By the way, this is sort of the stumbling block for all the string theories and many exotic stuff: to be better than what we have currently, it has to make better predictions that can be verified by experiments within our capabilities(!)
and correctly make all the predictions that the others already make. Put another way, to be just as consistent with all the experimental data that we can get. Otherwise, at best they would be theories fit only for special cases.
*: which would be "what if masses pull other masses towards themselves, which we'll call a gravitic force"