When Coordination Becomes Consensus: Mechanism-Attribution Loss in Social Listening Pipelines

Social listening platforms report volumes, sentiment shares, growth rates, and dominant narratives, and such reporting can be descriptively accurate. This paper argues that descriptive accuracy is compatible with a distinct failure: the loss, inside the analytical pipeline, of the information required to distinguish between competing generative explanations of the samecollective trace. We call this failure mechanism-attribution loss. On the Coordination Consensus Continuum (CCC), the observable signature of collective activity is a many-to-one trace of a partially observed causal state — intentionality, control centralization, participant agency — so a single signature is compatible with configurations from independent expressionto coordinated mobilization. We show how an idealized pipeline — capture, normalization, classification, aggregation, reporting — can render unavailable precisely the relational evidence (timing structure, content lineage, dependency, propagation order) on whichcoordination-detection methods rely to partially separate these configurations. We distinguish three inference types that industry language can conflate — mechanism attribution, population inference, and principal attribution — and state four propositions: (1) descriptive accuracy does not entail attributional validity; (2) aggregation can increase the stability of specified marginal summaries while decreasing the information available for mechanism attribution; (3) mechanism-attribution loss is generally irreversible downstream; (4) provenance requirements are claim-relative. From the fourth we derive the Claim–Provenance Map: a branched mapping from claim families onto the information each transformation must preserve for claims of that family to remain assessable. We define the condition under which a transformation is descriptively acceptable yet attribution-destructive, propose attribution-preserving social listening as a design principle, and audit real public claim language against the map.