[ZeroGravitas]

Here is the model:

Individual agents have a mild preference for publishing only models I agree with. They will never publish anything that I 2/3 disagree with (I rank all published papers in order of preference and the bottom 1/3 fail this test) and will immediately retract anything they have published that I disagree with to that extent and replace it with something I do agree with a random amount that places it in the top 66%. We start with them having published things that I randomly agree or disagree with.

No other factor than my agreement will have any impact on their decision to publish or retract.

Surprisingly, with only this mild preference people will automatically self censor and not publish any models I disagree with.

In reality, it's a series of laws that stops them, but we've just proved it's natural so really do those laws even matter?