Route-Dependent Correction and Credibility Curvature in Online Electric-Vehicle Misinformation: A Computational Test of Quantum Sociodynamics

Corrections to electric-vehicle (EV) misinformation are usually analysed as functions of content, source credibility or timing, yet the same correction can be read differently depending on the relation through which it travels. This article develops Quantum Sociodynamics (QSD), a gauge-covariant formulation in which actor orientations are complex states, relations are directed symbolic transporters, response is signed, and inconsistency between communication routes is represented by holonomy around closed cycles. It then asks two deliberately separate questions. Stage A tests whether a reduced QSD estimator recovers three signed couplings, relational memory, institutional-shock magnitude and a baseline response. Recovery is assessed across 80 baseline replicates and an additional stress grid spanning higher observation noise, shorter panels, fewer actors, missing phase observations, uncertain shock timing, heterogeneous memory, weak couplings and structural misspecification. Stage B is an explicitly exploratory analysis of 101 EV-related historical Birdwatch/Community Notes linked to 75 posts, 3,999 aggregate ratings, 145 status milestones and 139 cited-source records. Because closed directed cycles are not observable in these data, formal holonomy is not measured. Stage B instead evaluates a clearly labelled provenance-memory inconsistency index constructed from cited-source entropy, author-history mismatch and source-type count, together with a full weight-sensitivity grid and a component decomposition. Sparse-event estimation uses Firth penalised logistic regression with cluster bootstrap stability intervals, and inference is reported with effect sizes, confidence intervals and false-discovery-rate control rather than binary support labels. Baseline synthetic recovery was strong (true-estimated correlations 0.986 to 0.998; parameter RMSE 0.005 to 0.032; coupling sign recovered in every replicate) and degraded in an interpretable way under stress. Empirically, longer notes were associated with higher odds of platform resolution and greater cited-domain count with lower odds; the inconsistency index was associated with lower aggregate rating weight but not with greater reason-tag disagreement. Repeated grouped and rolling-origin cross-validation, benchmarked against an explicit prevalence model, showed no predictive gain from the index over a simpler provenance-memory specification. QSD is therefore computationally recoverable but empirically undecided in its distinctive claim: deciding it requires individual ratings, authentic exposure sequences and complete directed cycles. The contribution is a methodological boundary analysis for relational models of low-carbon technology communication, not a predictive instrument for platform governance.