Why the Data Revolution Needs Qualitative Methods

Tanweer, Anissa; Gade, Emily Kalah; Krafft, P. M.; Dreier, Sarah K.

doi:10.1162/99608f92.eee0b0da

Cited by 17 publications

(14 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More generally, we echo calls in the literature for more thoughtful and inclusive data collection (Jo and Gebru, 2020;Bender et al, 2021;Tanweer et al, 2021). This could include, but is not limited to a) intentionally curating data from people and viewpoints that are not otherwise well represented; b) including a greater diversity of genres; c) more nuanced or intentional exclusion criteria; d) more thorough interrogation of what text is being excluded; e) developing standard checks for prominent biases in inclusion; f) abandoning the notion of a general-purpose corpus.…”

Section: Discussionsupporting

confidence: 58%

Whose Language Counts as High Quality? Measuring Language Ideologies in Text Data Selection

Gururangan¹,

Card²,

Dreier³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Language models increasingly rely on massive web dumps for diverse text data. However, these sources are rife with undesirable content. As such, resources like Wikipedia, books, and newswire often serve as anchors for automatically selecting web text most suitable for language modeling, a process typically referred to as quality filtering. Using a new dataset of U.S. high school newspaper articles-written by students from across the country-we investigate whose language is preferred by the quality filter used for GPT-3. We find that newspapers from larger schools, located in wealthier, educated, and urban ZIP codes are more likely to be classified as high quality. We then demonstrate that the filter's measurement of quality is unaligned with other sensible metrics, such as factuality or literary acclaim. We argue that privileging any corpus as high quality entails a language ideology, and more care is needed to construct training corpora for language models, with better transparency and justification for the inclusion or exclusion of various texts. 1 We note that the term quality is often ill-defined in the NLP literature. For example, Brown et al. (2020) and refer to "high-quality text" or "high-quality sources"-both citing Wikipedia as an example-but without explaining precisely what is meant.

show abstract

Section: Discussionsupporting

confidence: 58%

Whose Language Counts as High Quality? Measuring Language Ideologies in Text Data Selection

Gururangan¹,

Card²,

Dreier³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Subsequently, digital health research should meaningfully engage with vulnerable population groups to ensure that new digital tools meet their individual needs [ 57 , 60 , 62 , 83 ]. Specifically, a new relationship between qualitative and quantitative research fields needs to be established that considers qualitative and quantitative research complementary to each other [ 84 ]. Related to this is the need for software developers to partner with user experience and user interface experts in designing digital health technologies accessible to everyone [ 85 ] and reduce rather than add to the workload of health care professionals.…”

Section: Principles For Policy Research and Practicementioning

confidence: 99%

Digital Health Paradox: International Policy Perspectives to Address Increased Health Inequalities for People Living With Disabilities

Kessel¹,

Hrzic²,

O’Nuallain³

et al. 2022

J Med Internet Res

View full text Add to dashboard Cite

The COVID-19 pandemic accelerated the uptake of digital health worldwide and highlighted many benefits of these innovations. However, it also stressed the magnitude of inequalities regarding accessing digital health. Using a scoping review, this article explores the potential benefits of digital technologies for the global population, with particular reference to people living with disabilities, using the autism community as a case study. We ultimately explore policies in Sweden, Australia, Canada, Estonia, the United Kingdom, and the United States to learn how policies can lay an inclusive foundation for digital health systems. We conclude that digital health ecosystems should be designed with health equity at the forefront to avoid deepening existing health inequalities. We call for a more sophisticated understanding of digital health literacy to better assess the readiness to adopt digital health innovations. Finally, people living with disabilities should be positioned at the center of digital health policy and innovations to ensure they are not left behind.

show abstract

“…Together these communities are developing theories, methods and systems that better understand and mitigate the adverse effects of automated decision-making tasks performed by machine learning systems. As others have argued [28,37,60,87], the concepts of reflexivity and positionality can provide a framework for data scientists to more systematically address and communicate the influence of discretionary decisions in the development of machine learning models. Put another way, reflexivity and positionality help data scientists to contextualize and situate their own knowledge, the knowledge held by data annotators, and the knowledge represented by computational models in a scalable, intuitive, and visual manner.…”

Section: Discussionmentioning

confidence: 99%

“…Recent work has argued for the incorporation of qualitative thinking, particularly reflexivity. Tanweer and colleagues [87] describe qualitative 'sensibilities' and connect them to data science work while providing three examples of reflexive techniques that data scientists can use. Miceli and colleagues [60] argue that reflexive practices may help address some of the more salient issues with documenting datasets for computer vision.…”

Section: Computational Reflexivitymentioning

confidence: 99%

“…More generally, the practice of machine learning and data science often come from the perspective that there is a singular, objective ground truth regarding the relationship between a target concept and the underlying data. Most methods for data exploration, learning algorithm design, and model evaluation have evolved from this central notion, and data scientists are rarely taught practices which promote acknowledgement, reflection, and discussion of their discretionary and subjective choices, though researchers have called for the need to do so (e.g., [60,77,87]).…”

mentioning

confidence: 99%

See 1 more Smart Citation

Model Positionality and Computational Reflexivity: Promoting Reflexivity in Data Science

Cambo¹,

Gergle²

2022

Preprint

View full text Add to dashboard Cite

Data science and machine learning provide indispensable techniques for understanding phenomena at scale, but the discretionary choices made when doing this work are often not recognized. Drawing from qualitative research practices, we describe how the concepts of positionality and reflexivity can be adapted to provide a framework for understanding, discussing, and disclosing the discretionary choices and subjectivity inherent to data science work. We first introduce the concepts of model positionality and computational reflexivity that can help data scientists to reflect on and communicate the social and cultural context of a model's development and use, the data annotators and their annotations, and the data scientists themselves. We then describe the unique challenges of adapting these concepts for data science work and offer annotator fingerprinting and position mining as promising solutions. Finally, we demonstrate these techniques in a case study of the development of classifiers for toxic commenting in online communities. CCS CONCEPTS• Human-centered computing → HCI theory, concepts and models; HCI design and evaluation methods; • Computing methodologies → Machine learning.

show abstract

Why the Data Revolution Needs Qualitative Methods

Cited by 17 publications

References 0 publications

Whose Language Counts as High Quality? Measuring Language Ideologies in Text Data Selection

Whose Language Counts as High Quality? Measuring Language Ideologies in Text Data Selection

Digital Health Paradox: International Policy Perspectives to Address Increased Health Inequalities for People Living With Disabilities

Model Positionality and Computational Reflexivity: Promoting Reflexivity in Data Science

Contact Info

Product

Resources

About