The Rambam, in his Morei Nevuchim, says that when it comes to knowledge about G-d, we can only know what G-d is not. I.e. He is not limited in space, or time, He does not have property A, nor does he have property "not A". We can have no positive knowledge of G-d's essence.
Of course this poses an obvious question that statements like "G-d is not A" or "it is not true that G-d is A" are themselves some forms of positive knowledge about G-d. But that's not my question here.
My question is - given the above restriction, how can we then ever know about or relate to G-d, and in particular how are we to understand statements like "G-d is Just" or even "G-d Created heaven and earth". It seems that the these statements are to be given meaning only in terms of observable effects by humans or by an analogy. For example, "G-d is Just" means that G-d's actions are seen to be 'just', as humans define that term. Or that "G-d created heaven and earth" means that the appearance of the physical reality bears some resemblance, via an analogy (or some sort of isomorphism?), to an act of creation as defined in common usage by man.
But even an analogy or even some conception connecting what is physically observed to G-d, also represents some knowledge of G-d. Namely, it is some sort of model inside one's mind, that links the concept of G-d to normal everyday concepts or structures such as "creating", "justice", etc. If we are able to have these models in our heads, then how is this knowledge qualitatively any different from knowing G-d's essence? For in truth, even for regular things (cars, rocks, humans, numbers, etc.) we do not know truly know their 'essence' - we speak only in terms of abstractions, models of these concepts that we have in our minds, and theories that allow us to link our models to observable reality. If so, how is that any different then from knowing G-d in the sense described above? If the knowledge about a computer, say, consists of various abstractions in our head about the mechanisms of how its inputs get transformed to outputs, then how is that knowledge different (qualitatively) from the various abstractions and models that we have in our minds about G-d (both with observable effects or not)?