Model Cards for Model Reporting¶
Gebru, T., Morgenstern, J., Vecchione, B., Vaughan, J. W., Wallach, H., III, H. D., & Crawford, K. (2021). Model Cards for Model Reporting. Communications of the ACM, 64(12), 86-92.
Cited by¶
7 citations across 7 artifacts.
Each citation links to the sentence it supports in the citing article.
Primes¶
- Boundary Disclosure Card
- Across these substrates the structural force is the same: when the producer-consumer interface is wide and reuse is impersonal, a standardized disclosure attached to the boundary substitutes for the trust, inspection, or shared context the consumer cannot otherwise obtain — and the fact that ML data cards explicitly borrowed the food-label pattern is direct evidence of the structure's portability.
This sourceProposes dataset documentation explicitly borrowing the electronics-datasheet and consumer-label idea for datasets and models.
Supported in partVerified against the work's full text
The paper supports the standardized-disclosure-for-consumers framing and explicit borrowing from the electronics datasheet, but not the food-label pattern or a general structural-force claim.
“In the electronics industry, every component, no matter how simple or complex, is accompanied with a datasheet describing its operating characteristics, test results, recom- mended usage, and other information.”
- Across these substrates the structural force is the same: when the producer-consumer interface is wide and reuse is impersonal, a standardized disclosure attached to the boundary substitutes for the trust, inspection, or shared context the consumer cannot otherwise obtain — and the fact that ML data cards explicitly borrowed the food-label pattern is direct evidence of the structure's portability.
- Information Use License
- Remedy varies wildly and the propagation setting varies deliberately; what does not vary is that a party already holding the information is told which of its acts are licit.
This sourceProposes documentation travelling with a dataset that records recommended and out-of-scope uses for the party already holding it.
- Remedy varies wildly and the propagation setting varies deliberately; what does not vary is that a party already holding the information is told which of its acts are licit.
- Reference Standard Decay
- The metrology discipline of explicit reference traceability and versioning is moving into ML benchmark practice as datasheets for datasets, versioned benchmarks, and reference-card requirements — a direct port of the role-asymmetry insight from instruments to label sets.
This sourceImports metrology's traceability-and-versioning discipline into ML, documenting dataset/benchmark provenance and revision.
Supported in partVerified against the work's full text
Datasheets for Datasets proposes documentation of dataset provenance and versions by analogy to electronics datasheets, not metrology, and it does not cover benchmarks or reference cards.
“In the electronics industry, every component, no matter how simple or complex, is accompanied with a datasheet”
- The metrology discipline of explicit reference traceability and versioning is moving into ML benchmark practice as datasheets for datasets, versioned benchmarks, and reference-card requirements — a direct port of the role-asymmetry insight from instruments to label sets.
- Rhetorical Velocity
- Machine-learning training data: content authored with awareness that it will be ingested and re-emitted by models, with provenance markers and explicit licensing as recomposition-aware moves.
This sourceProposes documenting a dataset's provenance, composition, intended uses, and distribution so creators and downstream re-users (including ML training pipelines) can reason about recomposition and re-emission — a provenance-and-licensing layer for content authored to be ingested.
Supported in partVerified against the work's full text
Datasheets for Datasets supports documenting dataset provenance, uses and licensing, but addresses dataset creators documenting datasets, not authors writing content in awareness of model ingestion and re-emission.
“dataset be accompanied with a datasheet that documents its motivation, com- position, collection process, recommended uses, and so on.”
- Machine-learning training data: content authored with awareness that it will be ingested and re-emitted by models, with provenance markers and explicit licensing as recomposition-aware moves.
Domain-specific¶
- Positionality
- in clients or students. AI dataset and model documentation — datasheets-for-datasets and model cards apply the identical constitutive-location-plus-declarability move to an artifact
This sourceThe datasheets-for-datasets proposal, which documents a dataset's motivation, composition, collection process and recommended uses.
Supported in partVerified against the work's full text
“dataset be accompanied with a datasheet that documents its motivation, composition, collection process, recommended uses, and so on”
- in clients or students. AI dataset and model documentation — datasheets-for-datasets and model cards apply the identical constitutive-location-plus-declarability move to an artifact
Mechanisms¶
- Dataset Datasheet or Data Card
- Shipped inside the download archive — not merely posted on a landing page — is a datasheet in the style of "Datasheets for Datasets."
This sourceIntroduces datasheets for datasets as structured documentation accompanying a dataset.
- Shipped inside the download archive — not merely posted on a landing page — is a datasheet in the style of "Datasheets for Datasets."
- Model Card or Datasheet Linkage
- Its strength is that it makes an abstraction's warrant inspectable and portable — the reason model cards and datasheets for datasets became standard practice is precisely that models and data outlive and out-travel the knowledge in their authors' heads.
This sourceProposes portable dataset documentation that makes motivation, composition, collection, and recommended use inspectable by downstream consumers.
- Its strength is that it makes an abstraction's warrant inspectable and portable — the reason model cards and datasheets for datasets became standard practice is precisely that models and data outlive and out-travel the knowledge in their authors' heads.
Verification¶
Does it exist? Confirmed. This work's DOI resolves to a registered record, which fixes its identity. That is all it fixes.
Does it back the claim? Read against the text for 4 of 7 citations: 4 supported in part. Each verdict is shown under its citation below, with what in the work backs the sentence.
Was it audited? Yes. A second, independent pass read the citation against the article text and recorded a verdict.
Support is checked per citation rather than per work — the same source can be cited soundly in one article and wrongly in another. Per-citation recording began recently, so a citation with no recorded check is a gap in the record rather than evidence it went unchecked.
See how references were verified.
Registry ID ref:6716addd0f16 · see in the full table