1) In one thread I saw that a Lagrangian that comes from a gauge theory principle is capable to generate interactions, and that would be why we should work with gauge theories. Nevertheless, any lagrangian which have multiplications of diferent fields generates interactions (or am I wrong?) 2) In some books I read that gauge theories are renormalizable, but some non gauge theories can be renormalizable too (so that would no be the reason either) 3) Finally, Symmetrys generates conservation of observables, so every observable that conservates after an interaction should have an undelying symmetry behind. And as we (well, to be honest "you") make experiments where we ("you") scatter particles and see what conservates, this should be the reason. Is the reason 1), 2) , 3) or another? Thanks!