An arxiv find, via David Hogg (via Facebook, via the internet).
The gravitational force law in the Solar System
Authors: Jo Bovy (NYU), Iain Murray (Toronto), David W. Hogg (NYU, MPIA)
Abstract: If the Solar System is long-lived and non-resonant (that is, if the planets are bound and have evolved independently through many orbital times), and if the system is observed at any non-special time, it is possible to infer the dynamical properties of the Solar System (such as the gravitational force or acceleration law) from a snapshot of the planet positions and velocities at a single moment in time. We consider purely radial acceleration laws of the form ar= –A [r/r0]-α, where r is the distance from the Sun. Using only an instantaneous kinematic snapshot (valid at 2009 April 1.0) for the eight major planets and a Bayesian probabilistic inference technique, we infer 1.989<α<2.052 (95-percent confidence). Our results confirm those of Newton (1687) and contemporaries, who inferred α=2 (with no stated uncertainty) via the comparison of computed and observationally inferred orbit shapes (closed ellipses with the Sun at one focus; Kepler 1609). Generalizations of the methods used here will permit, among other things, inference of Milky-Way dynamics from Gaia-like observations.
So: instead of noting that an inverse-square behavior for the force of gravity fits the data, assume that gravity obeys an inverse power law and fit for the power. (It’s two, to within the errors.) Of course there have been many higher-precision tests of gravity in the Solar System than this one; the new thing here is that the data are simply the positions and velocities of all the planets at one particular moment in time, no direct dynamical measurements. A little bit of Bayesian voodoo magic, and there you go.
What I want to know is, what makes the authors so convinced that their instantaneous kinematic snapshot is valid tomorrow?