Once one has a solution of the RTE, the most simple procedure of inference can be devised such that a comparison between calculated and observed Stokes spectra suggests modifications in prescribed models of the medium. Iteratively improved models refine the match between observations and theoretical calculations. When the match is good enough, the last model in the iteration is taken as a model of the medium and its characteristic parameters are the inferred parameters of the medium. This trial-and-error method may be useful when the model medium is very simple and contains just a few free parameters. Note that every change of a given free parameter implies an integration of the RTE which is a process requiring a great deal of computer time. If the number of free parameters is large, the manual trial-and-error method can become impracticable, but even automated trialand-error procedures that modify the various parameters randomly (blindly) may not converge to a physically reasonable final model of the medium. The results may even seem reasonable but be greatly in error.