basis_of_record for machine annotations data

[rdinnager]

Does basis_of_record refer to the observation that is being annotated (which is 'HumanObservation' for iNaturalist data) or to the annotation itself (which for the ML annotations is 'MachineObservation')? The latter would make sense so that human annotations and machine annotations for the same image can coexist. But in this case, do we also need to keep track of the basis of record for the original observation? For example, once we start adding herbarium specimens in, which are 'PreservedSpecimen', they could also have human or machine annotation attached to them. Should we be keeping track of both? A basis of record, and a basis of annotation?

[jdeck88]

I believe MachineObservation would be the basisOfRecord particularly since the record's unique identifier is called machine_learning_annotation_id and the purpose or target of this record really is about the inferred trait on the plant from a machine. we're not inferring traits from the original iNat data...