Machine Learning Data Practices through a Data Curation Lens - An Evaluation Framework

Aug 2, 2026

Upload by Miles

Dataset development practices in machine learning directly determine model outcomes and ethical behavior. The paper examines ML data practices using a data curation lens, evaluating them as data curation practices. It proposes a rubric-based evaluation framework grounded in data curation concepts and principles, then conducts a mixed-methods analysis of evaluation results for 25 ML datasets. Findings show ML researchers struggle to apply standard curation principles and reveal challenges in shared terminology, interpretative adaptation, required curation expertise, and documentation responsibilities.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages52

Reading Time2 h 11 min

AccessFree to download

Introduction

Dataset reuse and bias origins

Hidden practices and documentation challenges

Data curation lens and evaluation rubric

arXiv :2405 .02703v1 [ cs .CY] 4 May 2024 Machine Learning Data Practices through a Data Curation Lens: An Evaluation Framework ESHTA BHARDWAJ∗ , University of Toronto, Canada HARSHIT GUJRAL, University of Toronto, Canada SIYI WU, University of Toronto, Canada CIARA ZOGHEIB, University of Toronto, Canada TEGAN MAHARAJ, University of Toronto, Canada CHRISTOPH BECKER, University of Toronto, Canada Studies of dataset development in machine learning call for greater attention to the data practices that make model development possible and shape its outcomes. Many argue that the adoption of theory and practices from archives and data curation fields can support greater fairness, accountability, transparency, and more ethical machine learning. In response, this paper examines data practices in machine learning dataset development through the lens of data curation. We evaluate data practices in machine learning as data curation practices. To do so, we develop a framework for evaluating machine learning datasets using data curation concepts and principles through a rubric. Through a mixed-methods analysis of evaluation results for 25 ML datasets, we study the feasibility of data curation principles to be adopted for machine learning data work in practice and explore how data curation is currently performed. We find that researchers in machine learning, which often emphasizes model development, struggle to apply standard data curation principles. Our findings illustrate difficulties at the intersection of these fields, such as evaluating dimensions that have shared terms in both fields but non-shared meanings, a high degree of interpretative flexibility in adapting concepts without prescriptive restrictions, obstacles in limiting the depth of data curation expertise needed to apply the rubric, and challenges in scoping the extent of documentation dataset creators are responsible for. We propose ways to address these challenges and develop an overall framework for evaluation that outlineshow data curation concepts and methods can inform machine learning data practices. CCS Concepts: • Human-centered computing → Empirical studies in collaborative and social computing; • Computing methodologies → Machine learning; • General and reference → Evaluation. Additional Key Words and Phrases: data practices, datasets, dataset creation, datasheets, documentation, evaluation, machine learning, rubric ACM Reference Format: Eshta Bhardwaj, Harshit Gujral, Siyi Wu, Ciara Zogheib, Tegan Maharaj, and Christoph Becker. 2024. Machine Learning Data Practices through a Data Curation Lens: An Evaluation Framework. In The 2024 ACM Conference on Fairness, Accountability, and Transparency (FAccT ’24), June 3–6, 2024, Rio de Janeiro, Brazil. ACM, New York, NY, USA, 52 pages. [https://doi.org/10.1145/3630106.3658955](https://doi.org/10.1145/3630106.3658955) 1 INTRODUCTION The pervasive usage of predictive machine learning (ML) models has not dwindled in the face of ever-growing research discussing cases of biased results [2, 6, 9, 15, 26, 30, 31, 41, 54, 70, 71, 78, 89, 100, 114, 121, 141] . Bias in ML models often causes discriminatory, unfair, or unethical judgements Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than the author(s) must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires [prior specific permission and/or a fee. Request permissions from permissions@acm.org](prior specific permission and/or a fee. Request permissions from permissions@acm.org). FAccT ’24, June 3–6, 2024, Rio de Janeiro, Brazil © 2024 Copyright held by the owner/author(s) . Publication r

FAQs

What perspective does the paper use to study machine learning dataset development?

It uses a data curation lens to examine ML dataset development, treating data practices as data curation practices.

How is the evaluation framework built and what is it used for?

The framework is built using data curation concepts and principles, operationalized through a rubric to evaluate machine learning datasets.

What major difficulties are identified when applying data curation principles to ML work?

Researchers struggle with shared terms that carry different meanings across fields, flexible concept adaptation without prescriptive restrictions, insufficient scope of curation expertise, and uncertainty about how much documentation dataset creators should provide.

Aug 2, 2026

Upload by Miles