Understanding Structured Knowledge Production: A Case Study of Wikidata's Representation Injustice

TL;DR

This paper examines how structured knowledge contributions in Wikidata can lead to epistemic injustice, highlighting disparities in coverage and proposing solutions to improve fairness and inclusivity.

Contribution

It provides a case study analyzing coverage disparities in Wikidata and offers actionable solutions to address knowledge representation injustice.

Findings

Identified coverage disparities between countries in Wikidata.

Showed that automated edits significantly influence content representation.

Proposed strategies to enhance fairness and inclusivity in knowledge bases.

Abstract

Wikidata is a multi-language knowledge base that is being edited and maintained by editors from different language communities. Due to the structured nature of its content, the contributions are in various forms, including manual edit, tool-assisted edits, automated edits, etc, with the majority of edits being the import from wiki-internal or external datasets. Due to the outstanding power of bots and tools reflecting from their large volume of edits, knowledge contributions within Wikidata can easily cause epistemic injustice due to internal and external reasons. In this case study, we compared the coverage and edit history of human pages in two countries. By shedding light on these disparities and offering actionable solutions, our study aims to enhance the fairness and inclusivity of knowledge representation within Wikidata, ultimately contributing to a more equitable and comprehensive global knowledge base.