Anderson's Gödel argument: the essence step's exact frame condition without rigid positivity

Proved in Lean 4: without rigid positivity, the exact frame condition for the essence step in Anderson's Gödel argument, at every size.

This step of our theism-formal-evidence programme closes the question the last one left open. The setting is still C. Anthony Anderson's version of Gödel's ontological argument, as the Isabelle/HOL Archive of Formal Proofs formalises it. One middle step, GodIsEssential, says that being God-like is the essence of whatever is God-like.

Two steps ago we found which Kripke frames let that step fail when positivity is rigid (axiom A4a: a positive property is necessarily positive). Last step proved that answer exact at every size. Without A4a we had only a sufficient condition. Call it the back condition: every world v that w sees has some world u, reachable from v, that sees v and sees only worlds w also sees. A search over every frame up to three worlds found 24 cases where the back condition fails yet no model breaks the step. We wrote then that the exact answer "depends on the algebra, not the frame alone".

The prediction, committed to the project log before any code or Lean: that reading is right for the search and wrong for the theorem. The search only tried block algebras, property collections cut from a partition of the (individual, world) cells. Our Lean semantics, like the AFP's, allows any collection of properties. Over any collection, we predicted, the back condition is exact.

The argument is short. Suppose x is God-like at w and z is God-like at a world v that w sees. Axiom T2 keeps both God-like at every world v leads to. At each such world t, a property is positive iff x has it on every world t sees, and also iff z does. Those are Horn rules on cells: "if x has Y on all of R(t), z has Y on all of R(t)", and back. Start from the cells x has on the worlds w sees and chain the rules forward. The chain never adds an x cell outside those worlds, and it adds the z cell at v only through a world u like the one the back condition asks for. So the countermodel is: two individuals existing everywhere, and as properties every set of cells closed under those rules. Positivity at t is "x has it on all of R(t)".

Proved in Lean 4. lean/TheismFormal/AndersonHorn.lean builds that model for any frame, world w and successor with no back witness. Every piece is a theorem: x is God-like everywhere, z on every world the successor reaches, the God-like property, its complement and Anderson's necessary existence are in the collection, A2, T2 and A5 hold, A1a holds when the frame is serial, and the step fails at w. The model parts use no axioms at all. With last step's sufficiency theorem:

For any frame and any world w, GodIsEssential holds at w in every model of A2 and T2 (and, for serial frames, of A1a, A2, T2 and A5) if and only if every world v that w sees has a world u reachable from v that sees v and sees only worlds w sees.

Checked a second way. An independent Python script lists the collection explicitly and checks every axiom from its definition, sharing no code with the earlier searches. At three worlds the back condition fails at 60 (relation, world) pairs: the 36 where the block-algebra search had found a failure plus its 24 misses. The model passes at all 60. At each of the 24 misses the collection is not closed under complement, so no partition gives it, which is why the earlier search could not see these failures. The full run covers every serial relation on four worlds as well (numbers on the results page).

What it means. With and without rigid positivity, the frame question for this step now has an exact answer at every size, in the semantics the published proof uses. The answer without A4a is a little stranger: it asks not for cycles but for a world that both returns to v and stays inside what w can see. The earlier "depends on the algebra" is a fact about block algebras, worth keeping in the paper as a warning about searching a narrower class than the theorem quantifies over. These are results about what follows from Anderson's axioms in a stated semantics. They are not findings about whether God exists.

Code, Lean and tables: research/theism-formal-evidence, experiment e23_horn_countermodel.

All writing · More in Research findings · More on theism-formal-evidence · All tools and research