Gravitational Lens Recovery with GLASS: Measuring the mass profile and shape of a lens

We use a new non-parametric gravitational modelling tool -- \Glass{} -- to determine what quality of data (strong lensing, stellar kinematics, and/or stellar masses) are required to measure the circularly averaged mass profile of a lens and its shape. \Glass{} uses an under-constrained adaptive grid of mass pixels to model the lens, searching through thousands of models to marginalise over model uncertainties. Our key findings are as follows: (i) for pure lens data, multiple sources with wide redshift separation give the strongest constraints as this breaks the well-known mass-sheet or steepness degeneracy; (ii) a single quad with time delays also performs well, giving a good recovery of both the mass profile and its shape; (iii) stellar masses -- for lenses where the stars dominate the central potential -- can also break the steepness degeneracy, giving a recovery for doubles almost as good as having a quad with time delay data, or multiple source redshifts; (iv) stellar kinematics provide a robust measure of the mass at the half light radius of the stars $r_{1/2}$ that can also break the steepness degeneracy if the Einstein radius $r_E \neq r_{1/2}$; and (v) if $r_E \sim r_{1/2}$, then stellar kinematic data can be used to probe the stellar velocity anisotropy $\beta$ -- an interesting quantity in its own right. Where information on the mass distribution from lensing and/or other probes becomes redundant, this opens up the possibility of using strong lensing to constrain cosmological models.