This information will show up in the header of the published Label

This includes subject matter experts, community advocates, academics, etc.

Summarize this dataset in a few sentences, including the purpose and contents.

Does the dataset have a metadata repository or data dictionary? If yes, provide the link or access point and if not, explain the features in this dataset.

If yes, provide the link and if not, explain the features in this dataset.

For example, a dataset my be published in 2022 but collected between 2005-2010. If the data continues to be collected, choose "ongoing" for the end date.

What tools, services, or technologies were used to collect the data? If there were multiple steps, please include all.

Tools and services would include: software API, hardware apparatus or sensor, app, door-to-door, etc. Also indicate how they were used.

Upstream sources can include: other datasets, research papers, tweets, websites, etc.

Name these sources and provide access point to the upstream dataset(s) where possible. (Link, Description)

Upload visuals, such as data flow diagrams, that would illustrate the relationship to upstream data.

For example, to correct labeling errors, add new instances, delete instances, etc.

Include the names of companies, institutions, grant award numbers, etc.

Include the names of companies, institutions, grant award numbers, etc.

For example: checking for data type consistency, gaps in data completion, etc.

For example, quality reports, a list of quality metrics and expected distributions, etc.

For example: institutional review board approval, internal or third party ethical review process, a report on the potential impact on data subjects, etc.

Provide a description of these review processes, including the outcomes, as well as a link or other access point to any supporting documentation.

You may upload any visuals that indicate distributions if that is available as well.

This could include inferences or conclusions from individual instances, as well as models created from trends and correlations in the dataset.

Are there other uses, beyond those intended by the dataset producers, for which you could imagine this dataset being used repsonsibly?

This could include inferences or conclusions from individual instances, as well as models created from trends and correlations in the dataset.

Provide a description and links to papers or systems that use this dataset.

This could include inferences or conclusions from individual instances, as well as models created from trends and correlations in the dataset.

What concerns might you have about extrapolating trends or making generalized inferences from this dataset at a population level?

Broader populations could include geographic populations, demographic populations, or other groupings

Describe any known mitigation strategies, technical or otherwise, to address the concerns stated above.

Mitigation strategies could include data imputation or removal techniques, grouping approaches, or domain-specific knowledge.

Are there concerns around using this data to make decisions or predictions at the individual level?

Aggregate data may not capture variation across individuals, and therefore decisions made with that data may lack important information. In some cases, this lack of information can make those decisions incomplete, inaccurate, or discriminatory.

Describe any known mitigation strategies, technical or otherwise, to address the concerns stated above.

Mitigation strategies could include data imputation or removal techniques, grouping approaches, or domain-specific knowledge.

Describe any cultural or domain assumptions in data the field definitions that are not made explicit in the data dictionary.

This could include: representing information on a continuum in a set of categories, definitions that are domain-specific, grouping multiple categories into one broader category, assuming certain concepts are common knowledge across cultures/domains, etc.

A proxy characteristic is a variable or field that is implicitly connected to another variable or field, often one that is sensitive or connected to sensitive features (e.g. protected categories such as race, gender, sex).

Did people from the communities, groups, or identities represented in the dataset participate in the process of defining and planning this dataset?

This could include the participation of the racial, gender, socioeconomic, and ability, as well as geographic, identities that are represented in the dataset.

Describe any domain-specific knowledge that is necessary to ensure the dataset is used as intended.

This could include knowledge about terms, processes, or strategies for cateogrization that are domain-specific.

Were there any protocols for collecting and labeling data that required interpretation by individuals or trained systems?

This could include categorizing or labeling data manually using a protocol or leveraging a trained system.

Did people from the communities, groups or identities represented in the dataset participate in the data collection process?

This could include the participation of the racial, gender, socioeconomic, and ability, as well as geographic, identities that are represented in the dataset.

In many cases a dataset is a sample that represents a larger population; it is thus important to indicate where undersampling, oversampling, etc. may have taken place.

Describe any protocols followed to impute or fill in missing data, and indicate which fields were affected and in what quantity (%).

Describe any protocols followed to manipulate or adjust existing data, and indicate which fields were affected and in what quantity (%).

This includes actions such as cleaning, transforming, grouping data.

This includes both excluded items and lost data, such as unreadable data, assumed false survey answers, information retrieval malfunction, etc.

Include the URL or API endpoint where the information can be accessed.

Upstream Dataset Information (repeated per dataset)

Upstream Dataset Information (repeated per dataset)

Upstream Dataset Information (repeated per dataset)

This includes questions of: respresentation, sampling, labeling, communities engaged or excluded, any other domain interpretation or knowledge

Upstream Dataset Information (repeated per dataset)

Upstream Dataset Information (repeated per dataset)

Is there anything else that someone who is using this data should know about this upstream dataset?