Plus, part of the problem of measuring justice is that what we consider to measure is also a product of our time and place; part of the progression of justice could be discovering new moral dimensions

i think what it comes down to is if 1) people can discern on average if state of affairs x+1 is more just than x, and 2) if this compels people towards x+1. Justice may be too multidimmensional to isolate and measure a set of variables to track

The bubble from diminishing returns of prediction might "burst" but gen-ai still has a lot of potential to integrate into many new products. We still cannot (assuming user consent) ask our pc "find me the pdf I was looking at sometime last week on turtles". There may not be any bubble in this way