Gödel's God proof only works in a world where nothing could have been otherwise

A model search shows Scott's version of Gödel's ontological argument has models only with one possible world, plus what a Bayesian ledger on God leans on.

Two more checks from the theism formal-evidence programme, which puts arguments about God into machine-checkable form. All numbers are model outputs from stated inputs, not findings about whether God exists.

Gödel's argument. Take Scott's axioms (positive properties are closed under necessary implication, a property or its complement is positive, being God-like is positive, necessary existence is positive) and search every S5 model with up to five (world, individual) cells, counting every function from worlds to subsets of the domain as a property. There is exactly one model per individual when there is one world, and none with two to five worlds. That is Sobel's 1987 modal collapse seen from the outside: if the axioms hold, everything true is necessarily true, so the argument is satisfiable only where possibility is trivial. It is a bounded search, not a proof; the proof is Sobel's.

What the evidence ledger leans on. Dropping one item at a time, the midpoint probability of naturalism (0.76) falls to 0.09 without natural evil, 0.42 without moral evil and 0.50 without hiddenness, and rises to 0.99 without fine-tuning. I had predicted no published item could move it by 0.1; that was wrong, and the variance shares (prior and fine-tuning first) do not show it. A reader should look at both views.

Code, tests and the log, including the predictions I made before running: research/theism-formal-evidence/.

All writing · More in Research findings · All tools and research