Excerpt from Datasheets for Datasets
*

Gebru, Timnit; Morgenstern, Jamie; Vecchione, Briana; Vaughan, J.; Wallach, Hanna; Daumé, Hal; Crawford, Kate

doi:10.1201/9781003278290-23

Cited by 45 publications

(67 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We identify Adult, COMPAS, and German Credit as the most utilized dataset in the surveyed algorithmic fairness literature. Following Gebru et al (2018) and Holland et al (2018), we produce in-depth documentation for these three datasets by carefully consulting relevant references previously known by the authors or found querying search engines for academic publications with the dataset name. Appendices B-D contain datasheets and nutrition labels for these three datasets, preceded by a list of relevant publications that informed their compilation.…”

Section: Methodsmentioning

confidence: 99%

“…Different purposes can motivate data collection and annotation of sensitive attributes. Purposes and aims should be documented clearly, while also reflecting on other uses and potential for misuse of a dataset (Gebru et al, 2018). Dataset curators may use documentation to discuss these aspects and specify limitations for the intended use of a resource (Peng et al, 2021).…”

Section: Sensitive Attribute Labellingmentioning

confidence: 99%

“…Good documentation should discuss and explain features, providing context about who collected and annotated the data, how and for which purpose (Gebru et al, 2018;Denton et al, 2020). This provides dataset users with information they can leverage to select appropriate datasets for their tasks and avoid unintentional misuse (Gebru et al, 2018).…”

Section: Transparencymentioning

confidence: 99%

“…Following the widespread study and application of data-driven algorithms in contexts that are central to people's well-being, a community of researchers has coalesced around the growing field of algorithmic fairness, investigating algorithms through the lens of justice, equity, bias, power and harms. A line of work gaining traction in the field, intersecting with critical data studies, human-computer interaction, and computer-supported cooperative work, studies data documentation and proposes standardized processes to describe key characteristics of datasets (Gebru et al, 2018;Holland et al, 2018;Bender and Friedman, 2018;Geiger et al, 2020;Jo and Gebru, 2020;Miceli et al, 2021). Most prominently, Gebru et al (2018) and Holland et al (2018) proposed two complementary documentation frameworks, called Datasheets for Datasets and Dataset Nutrition Labels, to improve data curation practices and favour more informed data selection and utilization for dataset users.…”

Section: Introductionmentioning

confidence: 99%

“…A line of work gaining traction in the field, intersecting with critical data studies, human-computer interaction, and computer-supported cooperative work, studies data documentation and proposes standardized processes to describe key characteristics of datasets (Gebru et al, 2018;Holland et al, 2018;Bender and Friedman, 2018;Geiger et al, 2020;Jo and Gebru, 2020;Miceli et al, 2021). Most prominently, Gebru et al (2018) and Holland et al (2018) proposed two complementary documentation frameworks, called Datasheets for Datasets and Dataset Nutrition Labels, to improve data curation practices and favour more informed data selection and utilization for dataset users. Overall, this line of work has contributed to an unprecedented attention to dataset documentation in Machine Learning (ML), including a novel track focused on datasets at the Conference on Neural Information Processing Systems (NeurIPS), an initiative to support dataset tracking in repositories for scholarly articles, 1 and dedicated works producing retrospective documentation for existing datasets (Bandy and Vincent, 2021;Garbin et al, 2021), auditing their properties (Prabhu and Birhane, 2020) and tracing their usage (Peng et al, 2021).…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Algorithmic Fairness Datasets: the Story so Far

Fabris,

Messina,

Silvello

et al. 2022

Preprint

View full text Add to dashboard Cite

Data-driven algorithms are being studied and deployed in diverse domains to support critical decisions, directly impacting on people's wellbeing. As a result, a growing community of algorithmic fairness researchers has been investigating the equity of existing algorithms and proposing novel ones, advancing the understanding of the risks and opportunities of automated decision-making for different populations. Algorithmic fairness progress hinges on data, which can be used appropriately only if adequately documented. Unfortunately, the algorithmic fairness community, as a whole, suffers from a collective data documentation debt caused by a lack of information on specific resources (opacity) and scatteredness of available information (sparsity). In this work, we survey over two hundred datasets employed in algorithmic fairness research, producing standardized and searchable documentation for each of them, along with in-depth documentation for the three most popular fairness datasets, namely Adult, COMPAS and German Credit. These documentation efforts support multiple contributions. Firstly, we summarize the merits and limitations of popular algorithmic fairness datasets, questioning their suitability as general-purpose fairness benchmarks. Secondly, we document hundreds of available alternatives, annotating their domain and supported fairness tasks, to assist dataset users in task-oriented and domain-oriented search. Finally, we analyze these resources from the perspective of five important data curation topics: anonymization, consent, inclusivity, labeling of sensitive attributes, and transparency. We discuss different approaches and levels of attention to these topics, making them tangible, and distill them into a set of best practices for the curation of novel datasets.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Sensitive Attribute Labellingmentioning

confidence: 99%

Section: Transparencymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Algorithmic Fairness Datasets: the Story so Far

Fabris,

Messina,

Silvello

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Lack of Transparency and Potential Bias in Artificial Intelligence Data Sets and Algorithms

et al. 2021

View full text Add to dashboard Cite

IMPORTANCE Clinical artificial intelligence (AI) algorithms have the potential to improve clinical care, but fair, generalizable algorithms depend on the clinical data on which they are trained and tested.OBJECTIVE To assess whether data sets used for training diagnostic AI algorithms addressing skin disease are adequately described and to identify potential sources of bias in these data sets.DATA SOURCES In this scoping review, PubMed was used to search for peer-reviewed research articles published between January 1, 2015, and November 1, 2020, with the following paired search terms: deep learning and dermatology, artificial intelligence and dermatology, deep learning and dermatologist, and artificial intelligence and dermatologist.STUDY SELECTION Studies that developed or tested an existing deep learning algorithm for triage, diagnosis, or monitoring using clinical or dermoscopic images of skin disease were selected, and the articles were independently reviewed by 2 investigators to verify that they met selection criteria.CONSENSUS PROCESS Data set audit criteria were determined by consensus of all authors after reviewing existing literature to highlight data set transparency and sources of bias.RESULTS A total of 70 unique studies were included. Among these studies, 1 065 291 images were used to develop or test AI algorithms, of which only 257 372 (24.2%) were publicly available. Only 14 studies (20.0%) included descriptions of patient ethnicity or race in at least 1 data set used. Only 7 studies (10.0%) included any information about skin tone in at least 1 data set used. Thirty-six of the 56 studies developing new AI algorithms for cutaneous malignant neoplasms (64.3%) met the gold standard criteria for disease labeling. Public data sets were cited more often than private data sets, suggesting that public data sets contribute more to new development and benchmarks.CONCLUSIONS AND RELEVANCE This scoping review identified 3 issues in data sets that are used to develop and test clinical AI algorithms for skin disease that should be addressed before clinical translation: (1) sparsity of data set characterization and lack of transparency, (2) nonstandard and unverified disease labels, and (3) inability to fully assess patient diversity used for algorithm development and testing.

show abstract