It depends what your starting place is for the fundamentals of a model. For example general relativity describes gravity in terms of a metric, a way of measuring distances. This metric defines distances along a path by $$d = \int_{\textrm{path}} \sqrt{g_{ab}\dot{x}^{a}\dot{x}^{b}}dt$$. The metric is the object $$g_{ab}$$ which basically generalises Pythagoras's theorem to allow for curved space. There are 2 indices on this object. Electromagnetism describes the electromagnetic field in terms of a gauge potential, $$A = A_{a}dx^{a}$$. This has 1 index. It is possible to show in quantum field theory that if you quantise such formulations you find that the force due to the 1 index objects can both repel and attract but for those forces due to 2 index objects only attraction is possible. This doesn't require us to explicitly formulate quantum gravity, it is based on more general methods.
We can go a little further in justifying/explaining these things. If a force is described by a mathematical object with 0, 1 or 2 indices then it is known as spin 0, 1 or 2. For spinor fields (ie matter like electrons etc) the spin is half integer, so the electron is spin 1/2, while there's a hypothetical extension which includes spin 3/2. But what about spin 5/2 and above? Well you can again show from relatively high level arguments in quantum field theory that if you obtain spin 5/2 particles you are going to get inconsistent dynamics (causality violation and all that). So we've now narrowed down a reason for why we don't see other, weirder, forces.
Using some pretty general arguments about the structure of quantum field theories we can explain why the photon is massless and doesn't interact with itself, assuming the electron obeys a particular symmetry. We can explain why in the absence of any knowledge of the weak force we can still predict its existence. We can also motivate why, despite being allowed by the spin argument I just mentioned precluding 5/2 and above, we don't see spin 3/2 particles. Using arguments of gauge invariance, Lorentz invariance, renormalisation, causality and a hint of Occam's razor you are lead pretty much to the form of the Standard Model without needing to do many experiments. Experiments then amount to trying to find the values of the coefficients which are not fixed by these arguments.