NSTD.3:11 - SoTA-Echoing

Nguyen’s A Review of Mechanistic Models of Event Comprehension (2024) compares accounts of prediction, hierarchy, updating and causal representation in comprehension. Hoffmann’s The Tensions of Scientific Storytelling (2014) illustrates how scientific accounts organize mechanisms and unresolved tensions. Gatt and Krahmer’s NLG survey (2018) distinguishes content selection and realization. This pattern’s synthesis is to compare the connection the passage suggests with the relation its source or premises support.

Operational payload:

Event-comprehension models motivate checking the reader’s event boundaries, dependencies and predictions when those matter to the use. They do not establish that every reader builds the same model from every narrative.
From scientific storytelling, unresolved tension is narratively useful. It must stay unresolved when the source is unresolved. A good story may preserve tension rather than close it.
From NLG, inspect content as well as realization: a fluent connector cannot supply an omitted premise.
For evidence, inference and causal claims, inspect the complete claim before selecting additional work. The occurrence of “shows”, “proves” or “because” alone does not require an assurance procedure.
From architecture work, a mechanism-like story about structure must keep trade-offs, residual exceptions, and telemetry distinct. A before-and-after story is not by itself an architecture evaluation.

For a passage that implies an unsupported connection, the selected line is to inspect the event or dependency relation before polishing its connector. A serious alternative is local wording repair: replace “proved” by “suggested” or “caused” by “preceded”. At comparable effort on the one disputed sentence, that can be sufficient when it states what the source establishes; it fails when the weaker replacement also invents a fact. The incident case shows the decisive difference: retain the logged sequence and leave the cause unresolved instead of supplying a plausible mechanism.

Adapt event-model and content-planning distinctions in :4’s source comparison and :5’s diagnostics. Adopt scientific storytelling’s use of unresolved tension: the next investigation can remain part of the account. The accepted cost is consulting the relevant source relation rather than guessing a safer-sounding sentence. Keep a strong conclusion when that source supports it. Reopen if a new observation or argument establishes the connection, if the intended use needs a different relation, or if the reader still infers more than the repaired passage supports.