Decision models map stated rules and observed states to discrete answers. Aggregate accuracy leaves unanswered which rule or information changes drive those answers. We connect controlled input variations to rule transfer, joint reliability, and directed information use via executable arithmetic rules. Semantic rule separation and crossed rule‑state tests make transfer effects and joint success separately measurable. We compare equivalent wordings, calibrate joint correctness through marginal‑preserving reassociation, and alter tables to define alternative answer targets. Adaptation gains rise with agreement of the nearest‑source answer; fully conflicting quartets yield only small pooled gains and show opposite gain signs across source splits. Equivalent rewrites produce opposite task‑level effects, linking transfer gains to rule presentation. Joint‑score deviations depend on which rule‑state associations the reference preserves. A decision service changes answers more across equivalent wordings than across identical‑text repeats and usually follows the answer targets implied by altered table values or mappings.
Review