Aumann’s theorem assumes that assumptions can be enumerated and argued in some real time.
The reality seems to be far more complex than that, in that a human brain can generate hypotheses and relationships to existing datasets far faster than it can communicate the relationships, let alone reach any sort of agreement about datasets.
We are vastly more complex than that theorem allows for.
It is an oversimplification of reality that leads to invalid conclusions, even as it is true within its own set of assumptions, those assumptions clearly do not apply to reality.
Yet one more thing Robin Hanson and I can agree to disagree about 😉