For any obvious decision, consider secondary modalities of action which might aid or hinder it in the doing.
Robustnes of metrics shared between them: For r=the count of defined and acesible measurement categories expresing both of any pair of states, R=the relevance with a third state by the floor of the set of each pair's r. For any model, find the ceiling of r and R for each pair and triples, etc as notated rR_number of defined and acesible rR framestates, e.g. a comon sense course of action from start to finish, which is compared to the anticipated range of responses, and to the facts of the matter in hand ... is rR₃ ... which denotes a quality of predictive strength in the abstraction of discrete quantization, fitting any pedagogical framework with like-terms.
How good a measure is it? By comparing accuracy among systems of calculating predicted values in terms of 0...1, i.e. percentage of correct predictions within any range for any measure. And how does rR_n scale proportionally with acuracy of prediction?
Decisions might be ethically bound towards the advantages of co-operation and,or friendly competition.
There're practical maths to be had, no doubt, unless i've missed something along the way ... of course, we all have; as a working hypothesis, axiomatically we can say that there's always something else to consider and thereby another way to look at it which is yet to be considered.
Uncounted words carry meanings to be lived. For any sense of a word, it's total meaning is just the collection of all uses as delimited by any range of perspectives.
