Skin looks the way it does because light enters it, bounces around inside several layers, and comes back out changed by the pigments it met. Rendering that convincingly means simulating the scattering, and the parameters controlling it have to come from somewhere.

Current practice homogenises the medium, recovers optical parameters by inverting albedo from a reflectance texture, then hand-tunes scattering distance and anisotropy until it looks right. That last step is the tell. The authors describe the resulting art-skill dependency plainly: the model does not determine the appearance, an experienced artist adjusting numbers does.

Inverting for the biophysical chromophores instead means solving for what is actually in the skin rather than for parameters that happen to produce an acceptable image. The payoff is that quantities with physical meaning generalise, where a tuned value good for one face under one light is a fit to that situation.