Fieldmapping “Data Feminism”

Introduction

Rapid technology advancements and the shift from real life to online spaces at large notably has consequences. At an individual level, datafication happens daily as digital media users generate countless data points, shaping audience profiles and patterns. At the state level, data driven governance is emerging, making society simultaneously more connected and distanced from its citizens (Fourcade and Gordon, 2020). Most people can be traced and tracked by data, as well as potentially be overlooked due to being reduced as a number of a statistic. This data driven approach is inherently shaped by embedded biases in predictive models and risk assessments, and often amplifies existing inequalities, particularly for marginalised groups.

“Data Feminism”, a term coined in 2020 and proposed by Catherine D’Ignazio and Lauren Klein, is a framework that critiques such data practices by introducing an intersectional, analytical lens. This approach addresses how data is collected and used, and challenges the structures that give rise to systematic discrimination, privacy invasions, and the exploitation of vulnerable groups under the guise of technological progress.

Data Feminism

Historically, marginalised groups, especially women, have been misrepresented or made invisible by data. Before the data, there are individuals who share their experiences, those who analyse them, those who visualise them and those who remain unaccounted for (D’Ignazio and Klein, 2020) and data feminism approaches aim to integrate feminism into data science to make these individuals and their work visible. Data feminism- “(…) a way of thinking about data, both their uses and their limits, that is informed by direct experience, by a commitment to action, and by intersectional feminist thought” (D’Ignazio and Klein, 2020: 8).

In order to move away from data science that has been perpetuating inequality and discrimination for decades, the authors suggest 7 core principles to embed in the approach to data science and analytics. The core principles include: examining and challenging power, elevating emotion and embodiment, rethinking binaries and hierarchies, embracing pluralism, considering the context and making labour visible. These principles merge data science with feminist ideas and critical race theory, offering guidelines for data scientists to rethink the nature of their work and its impact.

Key Figures in the Field

Although data feminism as a concept was introduced recently in the book of the same name, influential figures in tech and feminism have paved the way for new approaches to data science.

Donna Haraway’s “A Cyborg Manifesto” challenged rigid categories of identity and power in technologically mediated worlds, a theme relevant to how data feminism addresses the complexities of identity in quantitative forms. Her critical essay “Situated Knowledges” links to the beginnings of data feminism, emphasising that knowledge is situated, always contextual and shaped by socio-political contexts. D’Ignazio and Klein build on Haraway’s ideas of challenging the notion of data objectivity, neutrality and power dynamics, advocating for a more equitable approach.

Furthermore, scholars like Kimberlé Crenshaw have contributed to the field through research on feminist epistemology and intersectionality. Crenshaw’s concept of “intersectionality”, which examines how overlapping identities, such as gender and race, shape individual experiences, has played a key role in informing this field’s commitment to addressing coexisting forms of oppression.

Additionally, in “Counting Feminicide” (2022), D’Ignazio in collaboration with data activists, further explores data feminism through digital activism, critiquing conventional data science’s hegemonic logics. Together, these scholars have shaped the core principles of data feminism, pushing the field to critically examine the power dynamics in data-driven technologies. Their influence is visible in fields like gender studies, critical data studies, and digital humanities, seeking to stress the importance of interdisciplinary collaboration for a more fruitful discussion on justice and equity.

Methodology

Data feminism emphasises qualitative, participatory approaches to address power imbalances in data practices. Drawing on “situated knowledge”, these approaches aim to engage marginalised voices and address systemic biases in data collection. As “data is never neutral” and it reflects social and political contexts (Kitchin, 2014), data feminism seeks to involve the affected communities in data collection processes or to expose inequalities in areas like racial disparities in policing or inequities in healthcare (Fourcade and Gordon, 2020).

Social Relevance

The overarching theme of data feminism is the mitigation of risks and consequences associated in all stages of data production and processing. While the term itself is not (yet) found in many publications, its core principles have resonated across many disciplines. Journals such as Feminist Studies, Women’s Studies, and those focused on technology and digital humanities, explore the societal impact of data and technology. Ethical concerns around AI biases, such as those in dataset creation, are central to this discourse (Crawford, 2020). Such datasets need to be examined in terms of what they represent, because they pose a risk for the emergence of “representation” and “historical” bias (Suresh and Guttag, 2020). Data collection, processing and handling missing values, are crucial to the epistemology that shapes how AI systems operate. Identifying the underlying harms in machine learning, as underlined by Suresh and Guttag, directly draw on the principles of data feminism, because they also aim to consider the socio-political, as well as economic, implications of data-driven science.

Additionally, Savoldi et. al. (2021) highlight how gender biases persist in automated translation systems, and offer critiques on the reinforcement of gender stereotypes within datasets. For example, translations often historically assign male or female professions with corresponding pronouns. Ultimately the way datasets are shaped can amplify and reflect societal biases.

Conclusion

Data Feminism represents a paradigm shift in data science that applies feminist principles to data methodologies and applications. It questions dominant paradigms, asking questions such as:

“Howcan intersectional approaches be operationalised in large-scale, quantitative data projects?”

It strives to make data practices more inclusive, equitable, and representative of diverse experiences. By prioritising qualitative and participatory approaches, data feminism addresses and aims to examine power imbalances in data generation and processing whereby the voices and experiences of marginalised people are silenced and ignored. Although D’Ignazio and Klein present a clear and concise theoretical framework for using intersectional feminism ideas and give examples that exhibit the issues of current methodologies, practical implementation remains a challenge. They discuss the “paradox of exposure” (people who would benefit significantly from being accounted for, actually pose the most risk from this counting), but do not mention the ways in which it could be addressed through data feminism. Furthermore, how do we question and deconstruct embedded binaries and classifications while ensuring practical feasibility? These questions highlight the need for ongoing exploration and adaptation of Data Feminism principles to advance equitable and inclusive data science.