Generative AI Leader Practice Question: The primary purpose of Google's Datasheets for…

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  To serve as a legal contract for data sharing

  Why it's wrong here

  Datasheets for Datasets are transparency documentation describing a dataset's motivation, composition and collection process; they carry no contractual force. It is tempting because they accompany data sharing, but their purpose is to inform downstream users about a dataset's characteristics and appropriate uses.

• ✗

  To list all models trained on the dataset

  Why it's wrong here

  Datasheets describe the dataset itself, not a register of every model trained on it; such tracking is not their function. It is tempting because they support provenance and accountability, but their purpose is documenting dataset composition, collection and recommended uses for downstream consumers.

• ✓

  To document the dataset's creation, composition, and intended use

  Why this is correct

  Datasheets for Datasets document how a dataset was created, what it contains, and its intended use, giving downstream users the provenance and composition context needed to judge suitability and bias risk. This satisfies the requirement for standardised dataset-level documentation accompanying the data.

• ✗

  To provide a template for labeling data

  Why it's wrong here

  Datasheets are not annotation templates; labelling guidance is a separate concern from documenting a dataset's provenance and composition. It is tempting because both support data quality, but the datasheet's purpose is to record how and why a dataset was created and how it should be used.

Go deeper

Related to this question