Wikibench: Community-Driven Data Curation for AI Evaluation on Wikipedia

要旨

AI tools are increasingly deployed in community contexts. However, datasets used to evaluate AI are typically created by developers and annotators outside a given community, which can yield misleading conclusions about AI performance. How might we empower communities to drive the intentional design and curation of evaluation datasets for AI that impacts them? We investigate this question on Wikipedia, an online community with multiple AI-based content moderation tools deployed. We introduce Wikibench, a system that enables communities to collaboratively curate AI evaluation datasets, while navigating ambiguities and differences in perspective through discussion. A field study on Wikipedia shows that datasets curated using Wikibench can effectively capture community consensus, disagreement, and uncertainty. Furthermore, study participants used Wikibench to shape the overall data curation process, including refining label definitions, determining data inclusion criteria, and authoring data statements. Based on our findings, we propose future directions for systems that support community-driven data curation.

著者
Tzu-Sheng Kuo
Carnegie Mellon University, Pittsburgh, Pennsylvania, United States
Aaron Lee. Halfaker
Microsoft, Redmond, Washington, United States
Zirui Cheng
Tsinghua University, Beijing, China
Jiwoo Kim
Columbia University, New York, New York, United States
Meng-Hsin Wu
Carnegie Mellon University , Pittsburgh, Pennsylvania, United States
Tongshuang Wu
Carnegie Mellon University, Pittsburgh, Pennsylvania, United States
Kenneth Holstein
Carnegie Mellon University, Pittsburgh, Pennsylvania, United States
Haiyi Zhu
Carnegie Mellon University, Pittsburgh, Pennsylvania, United States
論文URL

https://doi.org/10.1145/3613904.3642278

動画

会議: CHI 2024

The ACM CHI Conference on Human Factors in Computing Systems (https://chi2024.acm.org/)

セッション: Curating Online Content B

313B
5 件の発表
2024-05-16 18:00:00
2024-05-16 19:20:00