Path Sufficiency and Generative Consistency in Decision-Path Inverse Reconstruction: Formal Criteria for Constraining Decision Paths from Observed Outcomes

This working paper formalizes the path-sufficiency problem introduced by Decision-Path Inverse Reconstruction (DPIR). It asks when a reported or partially observed decision path can legitimately be judged compatible, sufficient, or insufficient with respect to an observed decision outcome. Rather than assuming a deterministic psychological function, the paper represents a bounded decision architecture as a constrained generative system. Available evidence defines a set of admissible complete paths; conditioning that set on the observed outcome produces an outcome-compatible path set. The paper formally distinguishes outcome compatibility, strong path sufficiency, outcome exclusion, outcome-required latent conditions, Structural Residual as an excluded-path set, and bounded minimal path repair. The paper also states four formal propositions concerning contraction of feasible path sets, sequential evidence, required-condition claims, and model misspecification. It provides a worked example, a probabilistic extension, an empirical validation protocol, and explicit failure conditions. DPIR is distinguished from partial identification, inverse reinforcement learning, structural causal explanation, and process tracing. The formalism is mapped onto the AIM sequence Awareness → Insight → Structure → Translation → Integration → Momentum.