Datasheet for a dataset
A document, proposed in AI research (Gebru et al., 2018 onward), that ships with a dataset stating how it was collected, what it represents, its known gaps and biases, and its intended uses. Where a dataset you use came with a datasheet, it populates your provenance entry; its absence is itself a provenance gap.
Defined in 2 GAGE programs, which carry 4 distinct definitions of it. The wording above is taught in Certified AI Governance Professional (CAIGP).
How each discipline defines it
The same term does different work depending on who is using it. These are the definitions as each program teaches them, unedited.
A document, proposed in AI research (Gebru et al., 2018 onward), that ships with a dataset stating how it was collected, what it represents, its known gaps and biases, and its intended uses. Where a dataset you use came with a datasheet, it populates your provenance entry; its absence is itself a provenance gap.
A standardized document that travels with a dataset describing its motivation, composition, collection process, and recommended uses, providing the semantic and provenance context a structural schema alone cannot capture; covered in full in Module 4. (see Topic 4.8)
A structured documentation practice, proposed by Gebru et al. in 2018, describing a dataset's motivation, composition, collection process, labeling methodology, and recommended and discouraged uses, modeled on electronics industry component datasheets.
A structured document describing a dataset's origin, collection process, composition, and intended and inappropriate uses. Proposed by Gebru and colleagues (2018; Communications of the ACM, 2021). The data-side counterpart to a model card; your data provenance file is your datasheet. (see Topic 2.6)
Where it is taught
The exact lessons this term appears in. The first module of every program is free with a free account.
- Training data governance: what may teach a model, decided before the model exists · Feeding the Machines, Certified AI Data Governance Professional (CADGP)
- Data contracts: producer and consumer agree in writing, and builds break when they lie · The Catalog That Lives, Certified AI Data Governance Professional (CADGP)
- The provenance file: your data map an auditor could follow · Data Reality, Certified AI Governance Professional (CAIGP)
- Model cards and system cards for your own systems, written so an outsider could act on them · Evidence Engineering, Certified AI Governance Professional (CAIGP)
Where it shows up at work
A definition is where Datasheet for a dataset starts. The lessons above teach it inside the competencies below, which are what job descriptions ask for and what the GAGE credential records. Every link here is drawn from the same careers map that builds a role's readiness check.
Traces where data came from, what transformed it, who owns each hop and where it flows downstream, so a number can be defended.
One lesson on this page teach it as a primary outcome.
Profiles data, sets rules and thresholds, monitors, assigns issues to owners, and knows when data is not fit to train or run a model.
One lesson on this page teach it as a primary outcome.
Collects, labels and preserves the evidence that a control operated, a decision was made, and a claim can be defended to an auditor or regulator.
One lesson on this page teach it as a primary outcome.
Roles that hire for it
Each role page shows the full competency set, what the seat decides, and a readiness check against your own evidence.
- Chief Data Officer, data leadership editionneeds it at depth
- Data Governance Leadneeds it at depth · 2 open postings this week
- Privacy Analystneeds working knowledge
- AI Auditorneeds it at depth · 9 open postings this week
- AI Privacy Engineerneeds it at depth · 6 open postings this week
- Third-Party AI Risk Analystneeds working knowledge
How job descriptions say it
Postings rarely use the textbook word. These are the phrases the careers map reads as asking for the competencies above.
Where the frameworks place it: EU AI Act Article 10; NIST AI RMF MAP 4; ISO 8000; EU AI Act Articles 11 and 12.
One interview question about it
Trace the lineage of a training dataset back to its source. What do you record, and what breaks when you cannot?
Asked of candidates for data lineage and provenance. Answering it in your own words is exactly how a GAGE topic is passed, and the credential records the pass where an employer can check it.
If you come from another field
Some backgrounds already carry this competency, and the readiness check gives credit for it.
- Privacy and data protection carries it as work you can show: Data maps and records of processing are lineage by another name.
- Data analysis, stewardship and business intelligence carries it as work you can show: Source-to-report lineage is what you trace.
- A data science, statistics or analytics degree carries it as a claim to back up: Lineage and provenance are known ideas, rarely yet documented for an auditor.
Terms it appears with
Not an alphabetical neighbourhood: these are the terms taught in the same lessons, ranked by how often they appear together.