Diversity and Inclusion Metrics in Subset Selection

By Margaret Mitchell, Dylan Baker, Nyalleng Moorosi, Emily L. Denton, Ben Hutchinson, A. Hanna, Timnit Gebru, Jamie H. Morgenstern. 2020

95 citations recorded by the source corpus.

The ethical concept of fairness has recently been applied in machine learning (ML) settings to describe a wide range of constraints and objectives. When considering the relevance of ethical concepts to subset selection problems, the concepts of diversity and inclusion are additionally applicable in order to create outputs that account for social power and access differentials. We introduce metrics based on these concepts, which can be applied together, separately, and in tandem with additional fairness constraints. Results from human subject experiments lend support to the proposed criteria. Social choice methods can additionally be leveraged to aggregate and choose preferable sets, and we detail how these may be applied.

Subfields: Artificial Intelligence, Computer Science Applications, Safety Research, Statistics and Probability, AI Ethics, Fairness & Society

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Douwe Kiela
2. Mohit Bansal
3. Ben Hutchinson
4. Vinodkumar Prabhakaran
5. Matteo Negri
6. Marco Turchi
7. Samuel R. Bowman
8. Nazneen Rajani

In Computer Science Applications

1. Tristan Thrush
2. Douwe Kiela
3. Ben Hutchinson
4. Alexandra Sasha Luccioni
5. Vinodkumar Prabhakaran
6. Margaret Mitchell

In Safety Research

1. Ben Hutchinson
2. Vinodkumar Prabhakaran
3. Marco Turchi
4. Tristan Thrush
5. Angelina McMillan-Major
6. Marco Gaido
7. Beatrice Savoldi
8. Douwe Kiela

In Statistics and Probability

1. Jason Baldridge
2. Marco Turchi
3. Douwe Kiela
4. Ben Hutchinson
5. Matteo Negri
6. Yacine Jernite
7. Luisa Bentivogli
8. Margaret Mitchell

In AI Ethics, Fairness & Society

1. Marco Turchi
2. Angelina McMillan-Major
3. Marco Gaido
4. Beatrice Savoldi
5. Douwe Kiela
6. Ben Hutchinson
7. Matteo Negri
8. Yacine Jernite

Citations in the corpus

• Measuring Data (2022, background): “, 2020) and inclusion (Mitchell et al., 2020), which are complementary to the diversity measures and can help represent different aspects of (non-)correspondence between the dataset and the populations that it represents, or those that will be impacted by model predictions.”
• The Vendi Score: A Diversity Evaluation Metric for Machine Learning (2022, unknown): “Mitchell et al. (2020) introduce metrics for diversity and inclusion, defining diversity in terms of the representation of socially relevant attributes like gender and race, and using the term heterogeneity to refer to variety in arbitrary attributes; in this paper, we use the term diversity to have…”
• Gender Bias in Machine Translation (2021, background): “, 2020), and not the only possible one (Mitchell et al., 2020).”
• Underspecification in Scene Description-to-Depiction Tasks (2022, background): “Some challenges that arise include how to measure image diversity in a socially appropriate way (Mitchell et al., 2020), as well as what space of possibilities should be represented at all.”
• Stable Bias: Evaluating Societal Representations in Diffusion Models (2023, unknown): “…vision and language models found that societal biases regarding race, gender and appearance propagate to downstream outputs in tasks such as image captioning (e.g., [28, 9]) and image search (e.g., [70, 42]), and has found that such biases are learned by state-of-the-art multimodal models [67, 68].”
• Detecting Independent Pronoun Bias with Partially-Synthetic Data Generation (2020, background): “Our linguistic understanding of (neo)pronouns and inclusive NLP development are areas of ongoing research (Ackerman, 2019; Bradley et al., 2019; Cao and Daum´e III, 2020; Conrod, 2020; Denton et al., 2020; Mitchell et al., 2020).”