A data model is also a theory of what counts as real. Tables privilege entities, attributes, and predefined schemas; graphs privilege connection, direction, provenance, and path. For domains such as influence, dependency, identity, or causation, the edge may carry more explanatory weight than either endpoint. Graphs do not eliminate interpretation: choosing labels, ontology boundaries, and temporal semantics still embeds assumptions. Their advantage is that those assumptions become inspectable objects rather than disappearing into application code and join logic.