Unfitness Scaling by Problems-Checking

A criterion of “intrinsic loss-of-value”, related to fitness decaying (eg due to health deficiency, etc.) on a certain problems-checklist, is outlined and then formally interpreted on a probabilistic setting. The question of interest concerns a structured, operative possibly automatized, way for extracting standardized “increment-of-unfitness” scores, conditionally on individual status x in a finite states-space, to quantify the ordinal levels of unfitness raw scale. In a Dirichlet-multinomial setup, a profile of hyper-parameters (from the which ““increment-of-unfitness” scores could be backward recovered) is defined which would regulate (according to assumptions about underlying dynamics) evolutionary hidden processes, underlying the problems-checklist, conditionally on status x; profile could be viewed as coordinate which, varying in a proper space, identifies a distributions-sub-manifold, modeled by constraints (specified by ex-ante knowledges and logics about such hidden dynamics). To extract value , tuned on “reference training-data-set” (in a machine-learning like sense), a (pseudo-Bayesian-like) “principle of orthogonality” is applied, which would minimize the residual information, beyond that implicitly already accounted through specifying constraints, carried by the training-data set