That was a very intelligent response. What I mean is that your response was exactly what one would expect of any careful and thoughtful reader. The problem is that you are looking at what you think I am saying and not at what I am actually saying. You have in your head a model of what I am doing, but that is not an accurate appraisal. There is a very subtle difference and, in this case, the whole significance lies in that subtle fact. Seeing what I am talking about is not easy so I hope you will think out my comments carefully.
As I said, the goal is to create a model which will model any explanation of A obtained from C. It has already been specified that the model must include a method to obtain expectations. The absolute first requirement which must be placed on any valid "explanation" is that those expectations must be 100 % consistent with C. Now think about that for a minute. That means that the algorithm for obtaining those expectations can not be a function of that mapping procedure described in the fabrication of the model. If it is a function of that procedure, it is most certainly the wrong answer as the expectations become a function of that procedure and not a function of the elements of C: i.e., the "observations" defined as B. Note, I call them "observations" though I avoided that word when talking to Harv because he always brings in his personal ideas about what words are supposed to mean and apparently totally lacks the ability to think about things in any way except as he has been indoctrinated to think.
With regard to your comment about how one establish C, I think you miss the central issue of my argument (don't feel bad, I think everyone else dismisses it without thought just as quickly). The central issue is that all relevant information is contained in C. Your explanation must be based on C alone. What you should understand is, if you include any information not contained in C, you are essentially presuming that the explanation of that information is valid. But, from my perspective, you cannot presume any ideas you currently hold are valid. It follows that any relationships you want to include in the modeled explanation must arise from C.
And, yes, any explanation obtained from C can indeed be wrong, However, it happens to be the very best one can do (by definition) and my deductions are certainly not based on the idea that the creation of the algorithm which provides those expectations is done in a way which can be described as "blindly".
Thank you again for your thoughtful responses. I remain here to provide you with a resolution of any difficulties you may confront. In the last forty years, I think I have confronted almost all of them.
Thanks -- Dick