Documentation

Linglib.Theories.Semantics.Causation.Interpretation

Causative Interpretation (force-dynamic dispatch) #

@cite{nadathur-lauer-2020} @cite{talmy-1988} @cite{wolff-2003}

Maps Causative verb classifications to their compositional semantics under the force-dynamic view (@cite{talmy-1988}, @cite{wolff-2003}), which collapses enable/force/make into a single sufficiency predicate (causallySufficient) distinguished post-hoc by isCoercive/isPermissive.

CausativeMechanismEnglish verbsN&L property (derived)
causeCounterfactual dependencecausenecessity
makeDirect sufficient guaranteemake, have, getsufficiency
forceCoercive (overcome resistance)forcesufficiency + coercion
enableBarrier removal (permissive)let, enablesufficiency
preventBarrier addition (blocking)preventpreventSem

Theoretical commitment #

This file commits to the force-dynamic mapping. The competing causal-model-theoretic view (@cite{sloman-barbey-hotaling-2009}) distinguishes enable from make/cause structurally: enable asserts B := A ∧ X (accessory variable required), while cause asserts B := A. Under that view, the present Causative.toSemantics mis-classifies enable. The Sloman alternative dispatch — and a divergence theorem witnessing the disagreement on enable — lives in Phenomena/Causation/Studies/SlomanBarbeyHotaling2009.lean.

This is intentional. linglib does not pretend a single canonical mapping exists; both dispatches coexist as named functions and the disagreement is theorem-provable.

Methods on Features.Causative that depend on heavy semantic machinery (Core.Causal.SEM, CausalGraph, the Necessity/ Sufficiency/Prevention modules) live here rather than in Features/Causation.lean, which is kept import-free per the "Features/ stays lightweight" convention. The namespace Features.Causative block below is the standard mathlib pattern for distributing methods on a type across files based on import weight.

make and force are distinguished by coercion despite sharing truth conditions.

make and enable are distinguished by permissivity despite sharing truth conditions.

Causative encodes force-dynamic mechanisms; CCSelectionMode (@cite{baglini-bar-asher-siegal-2025}) encodes which element of a causal model the construction can select as "the cause." These are orthogonal but connected: each variant has a canonical selection mode.