Summary of Position: Measure Dataset Diversity, Don’t Just Claim It, by Dora Zhao et al.
Position: Measure Dataset Diversity, Don’t Just Claim It
by Dora Zhao, Jerone T.A. Andrews, Orestis Papakyriakopoulos, Alice Xiang
First submitted to arxiv on: 11 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Computers and Society (cs.CY)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Machine learning datasets are often perceived as neutral, but they actually contain abstract and disputed social constructs. The terms used to describe these datasets, such as diversity, bias, and quality, lack clear definitions and validation. Our research explores the implications of this issue by analyzing the concept of “diversity” across 135 image and text datasets. We apply principles from measurement theory to identify considerations and offer recommendations for conceptualizing, operationalizing, and evaluating diversity in datasets. This has broader implications for machine learning research, advocating for a more nuanced and precise approach to handling value-laden properties in dataset construction. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Machine learning datasets are not as neutral as they seem. They contain hidden social constructs that can be positive or negative. The terms used to describe these datasets, like diversity, bias, and quality, don’t have clear meanings. We looked at 135 image and text datasets to see how people use the term “diversity”. By applying rules from measurement theory, we identified important points and provided recommendations for defining, creating, and checking diversity in datasets. This is important for machine learning research because it shows that we need to be more careful when working with value-laden data. |
Keywords
» Artificial intelligence » Machine learning
