Dataset (as output)

A dataset-as-output is a research dataset deposited in a FAIR-aligned repository with a persistent identifier (typically a DataCite DOI), full metadata, a stated licence, and defined access conditions — treated as a citable scholarly output in its own right rather than only as supporting material tucked behind a paper. The FORCE11 Joint Declaration of Data Citation Principles (2014) and the DataCite Metadata Schema together established the norms that made this possible: a dataset with those properties can be cited directly, the way an article is cited, independent of any paper that describes it.

A research office increasingly sees datasets-as-outputs counted formally: in narrative CVs and structured researcher profiles, in national research-assessment exercises such as the UK’s REF, and, gradually, in promotion and tenure criteria that have historically recognised only articles and books. A dataset may be accompanied by a data paper describing it, but it remains independently citable — the DOI resolves to the dataset itself, not to an article about it.

How it relates to sibling output types

• vs. Model (as output) — a model is a function trained on data; a dataset is the data itself. Both can be deposited and cited the same way, but they are conceptually different artefacts.
• vs. Workflow (as output) — a workflow describes the steps used to produce or analyse a dataset; the dataset is what those steps act on or produce.
• vs. Data validator — validation is a contribution to a dataset’s quality; the dataset-as-output is the resulting citable artefact.

References

• FORCE11 Joint Declaration of Data Citation Principles (2014).
• DataCite Metadata Schema v4.5.