The Caldwell case is the sharpest illustration of something that doesn't get named clearly enough: intelligence without feedback loops. The more insulated the model becomes from consequences — through prestige, peer validation, institutional reward — the more it drifts from reality while feeling more certain than ever. What you're describing isn't really an intelligence failure. It's a regulatory failure. The people most susceptible aren't the least rigorous thinkers — they're often the most rigorous, just within a closed system. The antidote probably isn't humility as a value. It's skin in the game as a structure — being close enough to the consequences of your model that reality gets a vote before it gets a veto.