“What data might be out of date? All of it!”: web archivist perspectives on datasheets for datasets

Purpose Web archive collections provide snapshots of archived websites, and recent work in the field aims to support analyses of these collections as data. However, web archives as a field has no existing standardized approach to documentation of data context to support analysis and interpretation. This study sheds light on what practitioners and researchers identify as the most important information for such documentation to include. Design/methodology/approach This article reports on one phase of a larger study investigating the utility of applying “datasheets for datasets” documentation to web archives. The first phase included a workshop introducing participants to the datasheets framework and asking them to prioritize the documentation elements it describes. The interviews analyzed here were conducted with select respondents who participated in the workshops. Interview transcripts were then studied through thematic analysis. Findings Web archive datasets are revealed to have complex relationships with both web archive collections and the live web from which they are ultimately derived. Respondents identified challenges for documentation to trace the proliferation of changes across these different contexts as well as tensions between labor and scale necessitating trade-offs in the granularity of description. Originality/value Little existing work addresses systematic approaches to documenting web archive datasets informed by consultation with stakeholders. Broadly, this article also contributes to discussion of what is required for transparent and ethical documentation of datasets collected from the web, such as machine learning data for which datasheets are designed.