Running models not to seek truth but to find permission — varying parameters until the simulation returns the result that justifies what was already decided. Described in the essay as "p-hacking at civilizational scale"
