Kratzer–Epistemic–RSA Bridge #
@cite{holliday-icard-2013} @cite{kraft-pratt-seidenberg-1959} @cite{kratzer-1981} @cite{kratzer-2012}
This file traces the dependency chain from Kratzer's ordering source
semantics down to RSA's worldPrior, via @cite{holliday-icard-2013}
epistemic likelihood logics:
Kratzer ordering source A : List (BProp World)
↓ atLeastAsGoodAs (subset inclusion on satisfied propositions)
World preorder ge_w : World → World → Prop
↓ halpernLift (Lewis's l-lifting)
EpistemicSystemW
↓ + totality, transitivity, additivity
EpistemicSystemFA
↓ theorem8a (|World4| = 4 < 5, so FA = FP∞)
FinAddMeasure
↓ mu {w} → worldPrior
RSAConfig.worldPrior
An alternative, more direct path is provided by @cite{fagin-halpern-1994}'s
KripkeKP (see Modality/KnowledgeProbability.lean): a Kripke probability
structure packages both the accessibility relation and the probability
assignment. Under CONS, knowledge and probability-1 belief coincide
(knows_implies_prob_one), giving a single structure that serves both
the epistemic logic layer and the RSA prior.
§1 and §2 are definitions (no sorry). §3–§4 are conjectures (sorry) that state what additional conditions are needed to traverse the full pipeline.
Convert Kratzer's Bool-valued atLeastAsGoodAs to a Prop-valued
world preorder, suitable for halpernLift.
Instances For
Kratzer's ordering is reflexive (Prop version).
Applying Lewis's l-lifting to Kratzer's world ordering yields EpistemicSystemW. This connects Kratzer modal semantics to Holliday & Icard's weakest epistemic likelihood logic.
Equations
Instances For
Conjecture: For World (= World4, 4 elements), if a Kratzer
ordering source induces an l-lifted system that satisfies the FA
axioms (totality, transitivity, qualitative additivity), then by
Kraft–Pratt–Seidenberg (Theorem 8a, since |World4| = 4 < 5), the
system is representable by a finitely additive measure, yielding
an RSA world prior.
This states the full Kratzer → RSA pipeline for the concrete 4-world type used throughout linglib's modal and RSA formalizations.
The key gap is that not every Kratzer ordering source yields an
l-lifted system satisfying FA — the hypothesis hFA captures the
additional conditions needed (totality, transitivity, additivity
of the lifted relation).
Conjecture: The probability-ordering round-trip is consistent.
Starting from a probability assignment P:
probToOrdering Pyields a Kratzer ordering sourceatLeastAsGoodAsgives a world preorderhalpernLiftgives an epistemic system- If FA holds, representation theorem gives measure μ
toWorldPriorextracts μ({w})
The round-trip prior μ({w}) should be consistent with the
original P(w) — not necessarily equal (since the l-lifting and
representation may lose information), but preserving the same
ordering: P(w) ≥ P(z) ↔ μ({w}) ≥ μ({z}).
This conjecture makes precise what "consistent" means: the probability ordering is preserved through the pipeline.
Note: probToOrdering is world-independent (probToOrdering_const),
so the choice of evaluation world w₀ is arbitrary.