Data Feminism

Catherine D'Ignazio and Lauren F. Klein argue that data science is shaped by unequal distributions of power and that intersectional feminist thought, a framework attentive to how overlapping systems of oppression interact, offers the best means of identifying and challenging those inequalities. Organized around seven principles, the book draws on examples from journalism, art, community organizing, and technology to show how data can both reinforce and dismantle oppression.

The book opens with Christine Mann Darden, a Black woman who joined NASA's Langley Research Center in 1967 as a data analyst. Darden's work on rocket reentry physics contributed to the Apollo 11 mission, yet she faced systemic racism and sexism at the agency. Women at Langley served as human "computers," performing crucial calculations but treated as unskilled temporary workers. When Darden noticed that men with identical credentials received engineering positions while women were sent to computing pools, groups assigned routine calculation work, she raised the issue with her division chief, who responded dismissively. Her colleague Gloria Champine compiled gender-and-rank statistics, confirmed the disparity, and presented the evidence to Darden's director, securing Darden's promotion. The authors define data feminism as a way of thinking about data informed by direct experience, commitment to action, and the concept of intersectionality, coined by legal theorist Kimberlé Crenshaw to describe how race, class, and gender produce overlapping privilege and oppression.

Chapter 1 establishes the first principle, examine power, through tennis star Serena Williams's near-fatal childbirth complications and the disproportionate maternal mortality rate for Black women in the United States. Drawing on sociologist Patricia Hill Collins's matrix of domination, the authors show how power operates across structural, disciplinary, hegemonic, and interpersonal domains. They introduce the term "privilege hazard" to describe how those in dominant positions fail to recognize oppression. The chapter details MIT researcher Joy Buolamwini's discovery that facial-analysis software fails to detect dark-skinned faces and that a widely used test dataset is 78 percent male and 84 percent white. Artist Mimi Onuoha's The Library of Missing Datasets, a file cabinet of empty folders, highlights how absent data about marginalized groups reflects structural indifference, while independent researcher María Salguero's mapping of femicides, killings of women because of their gender, in Mexico fills a vacuum left by government neglect.

Chapter 2 presents the second principle, challenge power, contrasting two Detroit maps: a 1971 map created by Black young adults under community organizer Gwendolyn Warren to document where white commuters killed Black children, and a 1939 redlining map encoding racial segregation into lending practices. The first exemplifies counterdata, community-collected information used to challenge official records. ProPublica journalist Julia Angwin's investigation reveals that the Equivant risk assessment algorithm mislabels Black defendants as high risk more often than white defendants. Sociologist Ruha Benjamin terms such systems "the New Jim Code," in which software code and a veneer of objectivity constrain Black lives. The authors distinguish concepts that secure power from concepts that challenge it, such as co-liberation, the pursuit of collective freedom through shared struggle.

Chapter 3 argues for the third principle, elevate emotion and embodiment. The authors open with Periscopic's animated visualization of 2013 US gun deaths, which frames each death as "stolen years." They critique the belief that visual plainness equals neutrality, drawing on feminist philosopher Donna Haraway's description of apparent objectivity as "the god trick of seeing everything from nowhere." The authors introduce Haraway's concept of situated knowledge, the idea that all knowledge is shaped by the knower's position, and feminist scholar Sandra Harding's strong objectivity, which holds that centering marginalized standpoints strengthens knowledge. The chapter profiles the theater troupe Elevator Repair Service's A Sort of Joy, in which three minutes of male artists' names from the Museum of Modern Art's collection pass before the first female name is spoken. Cartographer Margaret Pearce's Coming Home to Indigenous Place Names in Canada depicts First Nations, Métis, and Inuit place names without colonial reference points, asserting Indigenous sovereignty through the aerial view.

Chapter 4 presents the fourth principle, rethink binaries and hierarchies. The chapter opens with Maria Munir, a British college student who is nonbinary and does not identify exclusively as male or female, describing the toll of being forced into a binary choice on every form. The authors trace how both sex and gender are social constructs that solidified in eighteenth-century Europe alongside the race binary. They show how Facebook expanded visible gender options while resolving all users' genders into a binary for advertisers, and introduce the "paradox of exposure": Those who would benefit most from being counted face the greatest danger from that act. The Colored Conventions Project at the University of Delaware recovers nineteenth-century Black American convention records while requiring teaching partners to introduce a woman associated with each convention alongside every named male delegate.

Chapter 5 argues for the fifth principle, embrace pluralism, centering the Anti-Eviction Mapping Project (AEMP) in San Francisco, a collective documenting displacement and gentrification whose analysis shows that 69 percent of no-fault evictions between 2011 and 2013 occurred within four blocks of a tech company bus stop. The authors critique the "cleaning paradigm," warning that tidying data can strip away contextual information, and advocate for discussing "data settings," data understood through their local collection context, rather than abstract datasets. They distinguish "data for good" from "data for co-liberation," which requires community leadership and data ownership.

Chapter 6 presents the sixth principle, consider context. The authors open with FiveThirtyEight's retracted story about kidnapping in Nigeria, based on the Global Database of Events, Language, and Tone (GDELT), which logs media mentions rather than discrete events. They coin the term "Big Dick Data" for projects that ignore context and inflate their capabilities. A case study of campus sexual assault data reveals that higher reported rates often reflect better institutional support rather than worse conditions. Researcher Desmond Patton's SAFElab at Columbia University hires formerly gang-involved youth as domain experts to code social media posts, preventing misclassification of cultural references as violent threats.

Chapter 7 presents the final principle, make labor visible, tracing invisible labor from the 1970s Wages for Housework campaign to digital platforms. Scholar and activist Angela Davis added a crucial dimension by observing that Black women had long been poorly compensated for domestic labor in others' homes. Workers on Amazon Mechanical Turk, a crowdwork platform for piecemeal online tasks, earn less than minimum wage. Researchers Kate Crawford and Vladan Joler diagram all the labor, data, and material resources behind a single Amazon Echo in their project "Anatomy of an AI System," revealing chains of exploitation at every level.

The conclusion opens with the 2018 Google Walkout for Real Change, in which over 20,000 workers protested a $90 million exit package given to an executive after credible allegations of sexual misconduct. Google failed to meet the organizers' demands, and lead organizers Meredith Whittaker and Claire Stapleton faced retaliation and ultimately left the company. The authors catalog emerging resistance, including the Tech Workers Coalition, the #TechWontBuildIt movement, the Design Justice Network, and Data for Black Lives, an organization working to make data a tool for social change. They highlight emerging work in Indigenous data sovereignty, the right of Indigenous peoples to govern data about their communities; legal scholar Dean Spade's application of queer theory to data institutions; and model cards, standardized documentation describing a machine-learning model's uses and limitations. The book closes with a call to nurture these ecosystems of resistance before the norms of the data economy are fully determined.