Decoherence doesn't actually cause collapse of the wave function, but instead makes superpositions unobservable.
A lot of people say that the idea of the wave function of the universe doesn't make sense, but to me, it helps to understand conceptually what's going on with decoherence. Suppose you prepare a particle to be in a superposition of states corresponding to two different spatial locations. That means that the particle is described by the pure state \alpha |\psi_1\rangle + \beta |\psi_2\rangle where let's suppose |\psi_1\rangle is the state in which the particle is most likely to be at one location, x_1, and |\psi_2\rangle is the state in which the particle is most likely to be at a macroscopically distant location, x_2. Let |\Psi_0\rangle be the state of the entire rest of the universe. So the total wavefunction for the universe would be the state |\Psi_0\rangle \otimes (\alpha |\psi_1\rangle + \beta |\psi_2\rangle). Now, suppose that there is photographic paper spread across the region from x_1 to x_2 so that the particle interacting with the paper produces a dark spot on the paper (small, but visible to the naked eye). Then the particle interacting with the paper will result in a transition of the form:
|\Psi_0\rangle \otimes (\alpha |\psi_1\rangle + \beta |\psi_2\rangle) \Longrightarrow \alpha |\Psi_1\rangle + \beta |\Psi_2 \rangle
where |\Psi_1\rangle is a state in which the photographic paper now has a dark spot at location x_1 and |\Psi_2\rangle is a state in which the photographic paper has a dark spot at location x_2.
It isn't that the position of the particle has become more definite; it's that the multi-valuedness of the particle's position has "infected" the rest of the universe, to make it multi-valued, as well. In the language of the Many-Worlds Interpretation, the world has "split" into a world in which the particle is at x_1 and a world in which the particle is at x_2. But you don't have to accept MWI to agree that the interaction between particle and the rest of the world will cause the multi-valuedness to spread beyond just the particle. The two possibilities |\Psi_1\rangle and |\Psi_2\rangle will each evolve independent of each other's existence, so for each possibility, it's as if the other possibility didn't exist.