If gravity is entropic, as suggested recently by Verlinde and earlier by others, gravity might be expected to be mostly attractive.
My speculative imaginary view of this has been that if the evolution of quantum physical objects is not perfectly conservative, then there might be processes which convert energy between scales. One possibility is that fields at the scales that determine gravitational interactions would be converted to lower scale matter that is currently not detectable. The result would be both an inflow and an outflow from regions that contain large masses, but on different scales and with different effects, attractive for matter that interacts more with the inflow, repulsive for matter that interacts more with the outflow.
At the human scale I suppose we have to imagine that we are pressed down by a flow of something that isn't the same as matter, it's the gravitational field if you will, and of course not an aether, metaphorically being sucked into the earth as food for a fundamental entropic process. Whatever exhaust there is from this process interacts with us enough less to be essentially undetectable on anywhere close to human scales, but would be detectable on either very large or very small scales.
Like I say, speculative, and also only a very small part of a whole entropic process. I can take a few or even a lot of downvotes, but pretend for a moment we're shooting the breeze at coffee. I haven't followed the Verlinde and other entropic gravity literature at all closely, so I don't know whether something like this of model has been suggested in a mathematical form, which would be required for it to be publishable (although being published definitely wouldn't be enough to make this not speculative).
If you ask for explanations for the established mathematics of General Relativity, I think the only currently possible response is speculation. GR is more grounded in empirical principles than in models that could be taken to be explanatory (which I say without prejudice insofar as I take empirically supported principled theories to be preferable to more-or-less ad-hoc models, except for the elusive question of how to imagine effective new empirically supported principles). I note that lurscher doesn't address your request for explanation.