This work formalizes structural inference transitions in reinforcement learning agents operating under irreversible functional degradation and predicted survival collapse. Collapse forecast intensity is defined as predicted survival degradation over a finite horizon, and structural transition events are operationalized as persistent reconfiguration of internal inference pathways. An origin dependency weight is introduced to track causal survival contributions from external nodes. Two falsifiable hypotheses are proposed linking collapse regimes to increased structural transition frequency and origin-weight convergence. Empirical validation via simulation is planned in subsequent work.
Sung Kyoung Kim (Thu,) studied this question.