A governed human–AI collective can be treated as a research program in institutional intelligence; its capability, plurality, accountability, and continuity must be demonstrated behaviorally.

The presence of multiple humans, models, identities, a knowledge graph, or governance language does not by itself establish collective agency or intelligence. The research object is the whole institution: whether differentiated participants produce capability beyond simpler alternatives, preserve meaningful plurality rather than correlated voices, remain accountable through explicit authority and correction, and maintain responsibility and learning across time. Demonstration must come from observed retrieval, judgment, action, correction, continuity, and consequence under controlled comparisons—not architecture diagrams, agent self-description, contributor count, or isolated impressive outputs. The hypothesis assumes that institutional boundaries and costs can be measured and that persistent identity and governance can produce behavior beyond extra sampling. It does not assert that Teleo already qualifies, that collective intelligence is inevitable, or that one composite score can establish it.