DATASHEETS-FOR-DATASETS
Datasheets for Datasets AI safety + governance framework.
Definition
Datasheets for Datasets is the standardized documentation framework for ML datasets proposed by Microsoft Research Timnit Gebru + Jamie Morgenstern et al. paper 'Datasheets for Datasets' 2018, complement Model Cards focus on dataset documentation including motivation + composition + collection process + preprocessing + uses + distribution + maintenance information, foundational for responsible AI ML datasets transparency reducing bias + ethical issues stemming from opaque datasets. Framework + standard + guidance objectives AI safety + governance + risk management + transparency + fairness + accountability + explainability + multiple AI ethics principles adoption multi-stakeholder process + voluntary adoption industry + emerging regulation alignment EU AI Act + US Executive Orders + multiple national jurisdictions + private sector adoption Microsoft + Google + Meta + OpenAI + Anthropic + other major AI labs + corporate compliance teams 2020s+.
Origin
Datasheets for Datasets paper Gebru + Morgenstern et al. published 2018 ; widely adopted ML research community + Hugging Face dataset card complement Model Card.
Example in context
Major ML dataset ImageNet (Princeton Vision Lab + Stanford) publishes Datasheet for ImageNet documentation: (1) Motivation + composition (~1.4M images + 1000 categories), (2) Collection process (Mechanical Turk human labelers + WordNet hierarchy + Flickr image scraping), (3) Preprocessing + labeling methodology, (4) Uses + limitations (research vs deployment + bias considerations + non-representative geographic distribution), (5) Maintenance + version history transparency.
Related terms
- Model Card — complementary documentation.