Self Help

Data Feminism - D’Ignazio, Catherine & Klein, Lauren F_

Author Photo

Matheus Puppe

· 64 min read

“If you liked the book, you can purchase it using the links in the description below. By buying through these links, you contribute to the blog without paying any extra, as we receive a small commission. This helps us bring more quality content to you!”



Here is a summary of the key ideas in Data Feminism:

  • The book advocates applying feminist theory and values to the field of data science and technology. It argues data science needs feminism to address issues of power, ethics, and social impacts.

  • It proposes seven principles for doing data work in a feminist way: 1) Examine power 2) Challenge power 3) Elevate emotion and embodiment 4) Rethink binaries and hierarchies 5) Embrace pluralism 6) Consider context 7) Make labor visible.

  • Applying these principles means examining how power dynamics shape data and technology, challenging power structures through data collection and analysis, acknowledging subjective human experiences, avoiding oversimplified categories, including diverse perspectives, understanding data in its full context, and making data labor processes transparent.

  • The goal is to use data and technology to empower marginalized groups and work towards social justice, rather than reinforce existing inequalities and power imbalances. It calls for rethinking common practices and assumptions in data science through a feminist lens.

  • Overall the book advocates integrating feminist theory and values into data science as a way to address issues around ethics, power, inclusion, and the social impacts of new technologies.

  • Christine Darden began working at NASA’s Langley Research Center in 1967 as a data analyst and “computer”, doing calculations by hand. She faced discrimination as a Black woman in this predominantly white, male environment.

  • “Computers” at Langley were mostly women who performed important calculations but were treated as temporary, low-skilled workers without recognition of their contributions.

  • Darden noticed that men with math degrees were promoted to engineering roles while women were kept in computing pools with little opportunity for advancement.

  • She advocated for herself and other women by directly asking her manager why this distinction was made. His response showed the pervasive sexist attitudes of the time that women were happy and suited only for certain roles.

  • Darden’s story illustrates the workplace inequality and restrictive gender roles that were central issues of second-wave feminism in the 1960s, as sparked by books like The Feminine Mystique. However, her experience as a Black woman was not fully represented.

  • Her work and advocacy demonstrate why data science needed and continues to need feminism - to challenge discrimination, promote equality of opportunity, and give proper recognition to the important but overlooked contributions of women.

  • The passage criticizes Betty Friedan’s 1963 book The Feminine Mystique for ignoring how other aspects of identity like race, class, ability, sexuality, etc. intersect with gender to determine women’s experiences.

  • It introduces the concept of intersectionality, which argues these dimensions cannot be examined in isolation. Key scholars and activists were attuned to how racism compounded other forms of oppression for women of color.

  • It discusses bell hooks’ criticism that Friedan only considered the experiences of middle-class white women. She failed to acknowledge how being a maid, factory worker, etc. may have been more fulfilling than being a housewife for poor and non-white women.

  • The passage emphasizes how feminism must consider intersectionality to address current injustices, which often stem from historical power differentials among groups. It will draw from intersectional concepts developed by Black feminist scholars like the Combahee River Collective.

  • The overall message is that examining gender in isolation is incomplete and ignores how other identities shape women’s varied experiences of oppression. An intersectional lens is needed for a more just and equitable feminism.

  • Kimberlé Crenshaw coined the term “intersectionality” to describe discrimination based on both race and gender together. She explained that oppressive systems work in intersecting ways that multiply their effects on marginalized groups.

  • Intersectionality describes not just individual identities, but also the intersecting forces of privilege and oppression in society. Oppression involves the systematic mistreatment of groups without equal power.

  • Christine Darden experienced intersecting systems of oppression and privilege as a Black woman working in a predominantly white, male environment at NASA. Collecting data on representation helped expose these issues.

  • Darden’s experience exemplifies “data feminism” - using data in an intersectional, experience-driven way to understand and challenge power structures. While data reduces lived experiences, it was necessary to persuade those in power.

  • Converting experience to data always entails reduction and conceptual burdens. But listening to those behind the data is important. Data projects involve many people - before, during and after. There are always limitations to what data alone can address.

  • Data feminism, like justice, must remain both a goal and ongoing process of guiding thought and action toward social change. The key focus is on power structures in society and how data science can both reinforce and challenge inequality.

  • The passage discusses the exponential growth of computing power and data collection capabilities over time, enabled by Moore’s Law where transistor counts double every year. This has enabled vast amounts of data to be collected about people and their behaviors.

  • Historically, data collection has been used to consolidate power and knowledge over the people being collected. Examples given include slave ship records, eugenics movements, and modern surveillance of Black communities.

  • Edward Snowden’s leaks revealed extensive data collection by the US government on its own citizens with little oversight. Cities also collect vast amounts of data from public spaces.

  • Corporations systematically collect behavioral data from online activities like searches and social media to target ads and maximize profits through “datafication.” Even offline behaviors are being tracked and quantified.

  • Decisions are increasingly being made by algorithms analyzing big datasets, which can amplify existing biases like PredPol’s predictive policing software that disproportionately targets minority neighborhoods based on historical crime data.

  • Data feminism aims to trace biases in data and systems back to their historical and social roots, understand their impacts, and challenge oppressive data collection and uses through community-driven methods and diverse examples of data science.

The introduction examines why data science needs feminism through an intersectional feminist lens. It outlines seven principles of data feminism:

  1. Examine power - Analyze how power operates in the world.

  2. Challenge power - Commit to challenging unequal power structures and working toward justice.

  3. Elevate emotion and embodiment - Value multiple forms of knowledge, including lived experiences.

  4. Rethink binaries and hierarchies - Challenge systems like the gender binary that perpetuate oppression.

  5. Embrace pluralism - The most complete knowledge comes from multiple perspectives, prioritizing local and Indigenous voices.

  6. Consider context - Data are not neutral or objective, but products of social relations and systems of power.

  7. Make labor visible - Make the often invisible labor that goes into data science and knowledge production visible.

Each subsequent chapter will explore one of these principles in more detail using real-world examples. The goal is to show how data and data science can be used as a tool for social justice from an intersectional feminist perspective.

  • Serena Williams shared on social media about complications she experienced giving birth, which prompted many Black women to share their own stories of difficulties during childbirth.

  • Statistics show Black women are over 3 times more likely than white women to die from pregnancy or childbirth-related causes. Reproductive justice groups have been working on this issue for decades.

  • The mainstream media began covering the racial disparities more after Williams’ experience and reporting by ProPublica and NPR. However, data collection is still weak in the US.

  • Williams recognized her celebrity status contributed to her receiving proper medical care, when many Black women do not. She said “If I wasn’t who I am, it could have been me.”

  • The chapter introduces Patricia Hill Collins’ concept of the “matrix of domination” to analyze how systems of power are configured through four domains: structural, disciplinary, hegemonic, and interpersonal.

  • Examining these intersecting systems of power and oppression is the first principle of data feminism, in order to understand how bias and inequities are baked into our datasets and technologies.

  • The passage discusses how data science and related fields lack diversity and are dominated by privileged groups like white men. This leads to what is called the “privilege hazard” - those in privileged positions have difficulty recognizing oppression or harm because they lack the “empiricism of lived experience.”

  • Examples are given of how biases and discrimination can be encoded into data systems, like Amazon’s recruitment algorithm preferring male candidates or Google images depicting black teens differently than white teens.

  • The lack of diverse perspectives means there is a collective privilege hazard where it is unlikely biases will be identified before systems are deployed at scale. Data science risks “hard-coding” discrimination into digital infrastructure if the field remains dominated by a narrow group.

  • For meaningful progress, data science needs to acknowledge it does not represent the overall population and ask uncomfortable questions about whose goals and interests are prioritized versus marginalized in its work.

This passage discusses Joy Buolamwini, a scholar who studied the problem of bias in facial recognition technology. Some key points:

  • Buolamwini found that facial recognition software had trouble detecting her dark-skinned face, but could easily detect the faces of her lighter-skinned collaborators. She had to put on a white mask for it to recognize her.

  • By examining the training datasets used for facial recognition algorithms, she discovered they contained mostly male (78%) and white (84%) faces. When looking at gender and skin tone together, only 4% of faces were women with dark skin.

  • In tests of commercial facial recognition systems, Buolamwini and Timnit Gebru showed darker-skinned women were up to 44 times more likely to be misclassified than lighter-skinned men.

  • Buolamwini’s work highlighted the issue of algorithmic bias stemming from lack of diversity in datasets. However, just collecting more diverse datasets is not a full solution, as it could enable greater surveillance and oppression.

  • Buolamwini founded the Algorithmic Justice League to work on technical audits of bias, as well as cultural, policy and advocacy efforts to meaningfully address the underlying power issues around AI and its disparate impacts.

So in summary, Joy Buolamwini is highlighted as a scholar who studied and drew attention to the problem of racial and gender bias in facial recognition technology through her technical findings and multifaceted approach toward solutions.

The passage discusses missing or lacking datasets that reveal social biases and structural disregard for certain groups. It gives examples of potential but missing datasets like “People excluded from public housing because of criminal records” and “Mobility for older adults with physical disabilities or cognitive impairments.”

It describes an art installation by Mimi Onuoha that compiles a list of missing datasets to call attention to cultural indifferences. It also mentions the “gender data gap” where most research data is based on men’s bodies, overlooking women, as highlighted by initiatives like Data2X and the book “Invisible Women.”

The passage discusses how some address these gaps through “counterdata” or “data activism,” like Ida B. Wells collecting lynching statistics in the 1800s or current groups crowdsourcing names of women who died in childbirth in the US.

It provides an example of this work through María Salguero, who maintains a map of over 5,000 cases of femicides (gender-related killings of women) in Mexico since 2016, filling a data void where the government has failed to adequately record such crimes. Her map collects names and details of victims to make them visible and help seek justice.

  • The example of missing data about femicides in Mexico highlights a broader issue of missing data about marginalized groups in societies with unequal power structures. This results from willful disregard, deferral of responsibility, and organized neglect.

  • Individuals and communities often work to collect their own “counterdata” to fill gaps and hold powerful institutions accountable. This shows how data science can empower marginalized people.

  • However, data systems of powerful institutions are often built on excessive surveillance of marginalized groups, overrepresenting them in the data. This can disadvantage them, as seen in an algorithm that predicts child abuse risk based more on data about poor families.

  • Data science priorities the goals of its creators - efficiency for governments, profit for corporations, not the goals of the people impacted. This is due to resources required, so only powerful institutions can work with data at large scales.

  • This results in data science serving the goals of science, surveillance and selling, rather than other possible goals that may be underserved. Examples given are of physical infrastructure like data centers required to store large amounts of data.

This section discusses the climate change impacts of data centers and data storage in the cloud. Some key points:

  • Building and operating data centers requires massive financial resources, with Facebook’s new data center in New Mexico expected to cost $1 billion to construct and $31 million per year in electricity costs alone.

  • Only large corporations like Facebook and smaller institutions like wealthy governments and elite universities have the resources to build and maintain huge data centers.

  • The people who run these well-resourced institutions and benefit the most from data collection and use are disproportionately rich, powerful white men.

  • Data is often described as “the new oil,” highlighting both its potential for profit extraction from people’s data as well as its significant ecological costs from energy-intensive data storage and processing.

  • The massive infrastructure and energy needs of data centers have a substantial climate impact through the associated greenhouse gas emissions from electricity generation to power the centers.

So in summary, it discusses how data storage in large centralized data centers controlled by powerful corporations and institutions comes at both financial and environmental costs, including exacerbating global climate change through increased energy usage and emissions. The section critiques who benefits most from this system.

  • The passage discusses two maps of Detroit - a 1939 “Residential Security Map” and a map from 1968-1971 created by the Detroit Geographic Expedition and Institute (DGEI).

  • The 1939 map was an early example of “redlining” where Black neighborhoods were literally outlined in red, signaling them as high-risk areas for bank loans. This reinforced existing racial inequalities and segregation.

  • The DGEI map challenged this unequal distribution of data and power in three ways: 1) They compiled their own counterdata on social issues like child deaths since official data was missing. 2) They intentionally mapped structural oppression issues. 3) It was made by Black community organizers and youth with support from white academics.

  • The passage analyzes how today’s risk assessment algorithms used in criminal justice resemble redlining maps in being neither neutral nor objective. Studies have shown they often mislabel Black defendants as higher risk than white defendants.

  • The algorithms rely on proxies for race like questions about family structure, continuing structural biases embedded in the criminal justice system. Counteracting this requires compiling alternative data sources and auditing algorithms to expose oppressive outcomes.

  • Risk assessment algorithms used by companies like Equivant in the criminal justice system are designed to predict recidivism (likelihood of reoffending). However, they often rely on proxy variables that are linked to race like being raised by a single mother or having friends/family who were arrested.

  • Even though the creators claim not to consider race directly, these proxy variables incorporate racial bias and disadvantages because things like single parenthood rates differ significantly along racial lines. As a result, the algorithms tend to disadvantage black people.

  • Julia Angwin’s investigation at ProPublica found the COMPAS algorithm used by Equivant was 77% more likely to incorrectly flag black defendants as higher risk compared to white defendants. So it reinforced existing racial disparities.

  • The investigation also found gender disparities, where high risk scores meant different likelihoods of reoffending for men versus women. However, Angwin decided to focus reporting on racial bias due to prior work highlighting gender issues and her own experience with sexism in journalism.

  • Auditing algorithms through data analysis is an important tool for journalists to make opaque harms from algorithms more visible and hold systems accountable. It helps challenge unequal power dynamics reproduced through new technologies.

  • The passage discusses efforts to use data and analysis to expose problems like discrimination and oppression, with the theory being that revealing the extent of issues will prompt those in power to take action.

  • Examples are given where data-driven analysis in mainstream media did prompt some policy changes. However, two caveats are discussed.

  • First, data alone may not be enough and needs to be accompanied by other tools like community organizing and protest. Data can also be minimized or dismissed by those with power.

  • Second, the analysis could unintentionally contribute to harmful deficit narratives that reduce groups to problems rather than strengths. It could place an undue burden of proof on marginalized groups.

  • Overall, while exposing issues through data is important, it is equally key to involve communities directly and amplify their existing efforts, in order to counter deficit narratives and ensure the subjects of data have ownership and agency in the process.

The passage discusses issues with focusing solely on data ethics and bias when addressing problems in data and algorithms. It argues that this approach does not get at the root causes of structural oppression and inequality.

Some key points made:

  • Addressing bias is important but is also a “captivating diversion” from deeper issues of unequal power and privilege resulting from history.

  • A data justice framework that considers past inequities is needed to avoid algorithms that replicate historical harms like redlining.

  • Concepts like fairness and transparency are not enough on their own and limit the possible solutions. A feminist approach recognizes oppression is real and worth dismantling.

  • Restorative justice is proposed as an approach, like considering historical admissions discrimination in college admissions today.

  • Equity, not just equality, is the goal as it considers power differentials over time rather than just the present.

  • The source of bias is structural oppression, so systems need to address this, not just fix bias retroactively.

  • The framework of co-liberation toward mutual benefit is proposed rather than just “helping” marginalized groups. Both dominant and minoritized groups must work together towards dismantling oppression.

  • The project initially focused on data profiling and resistance but shifted to specifically studying surveillance in response to community concerns.

  • Even within big tech, employees are increasingly pushing back against projects like using AI for military drone strikes or border enforcement due to ethics concerns. This pushback led Google and Microsoft to cancel some military contracts.

  • Building understanding and addressing the systems and contexts that lead to discrimination is important. Simply improving algorithms is not enough without considering history, culture and power structures.

  • Data science education needs to move beyond the “Horace Mann factory model” which mostly benefits elite white men. Alternative models like Local Lotto engage communities and teach concepts through place-based issues of equity and justice.

  • The Local Lotto project taught high school students in New York City about data analysis and statistics concepts by having them study the New York State lottery.

  • Rather than taking an abstract, technical approach, it grounded the lessons in a real-world issue relevant to students’ lives - examining whether the lottery is good or bad for their neighborhoods.

  • Students collected both qualitative and quantitative data through interviews, observations, and statistical analysis. They mapped locations that sell lottery tickets and analyzed maps of income levels.

  • Students worked collaboratively to analyze the data and present their findings. They created opinion pieces arguing whether the lottery benefits their communities.

  • The project challenged power by taking a justice-oriented approach rather than just teaching technical skills. It valued students’ lived experiences and qualitative data from community members.

  • While successful in improving students’ data analysis skills, organizers also acknowledged shortcomings like not explicitly addressing issues of race that students raised. They are refining the curriculum based on lessons learned.

  • The project shows how data science education can be grounded in social justice by examining real issues, valuing community voices, and continuously reflecting on power dynamics.

The passage discusses the tension between objective, data-driven visualization approaches and more emotionally evocative styles. It uses the example of the Periscopic project that visualized gun deaths in the U.S. in 2013 as arcs representing lost years of life. While praised for its message, the project raised concerns from some in the data visualization field about using emotion and persuasion.

Traditionally in fields like statistics and data communication, visualizations are meant to be plain, neutral and devoid of subjective elements in order to be objective. However, the passage argues this concept of neutrality is problematic and masks underlying perspectives. It draws on feminist philosophy to challenge the false dichotomy between reason and emotion. Exploring how emotion can be constructively leveraged in visualization, rather than resisted, is key to the data feminism principle of embracing emotion and embodiment. Overall, the passage examines debates around subjectivity, objectivity and rhetorical aspects in data visualization design.

  • Early data visualizers like W.E.B. Du Bois, Shanawdithit, and Elizabeth Palmer Peabody created visualizations to communicate specific messages and perspectives, understanding the rhetorical power of images.

  • However, modern data visualization has been approached more from a technical/engineering perspective focused on objectivity and neutrality.

  • All visualizations involve editorial choices that frame the data in certain ways, even those presented as purely factual or neutral. For example, the New York Times created graphs of unemployment data from Democratic and Republican perspectives that emphasized different aspects of the numbers.

  • Conventions used in data visualization like two-dimensional views, clean layouts, and citation of sources contribute to perceptions of neutrality but still involve framing effects and rhetorical influences on interpretation.

  • Feminist philosophers propose alternatives to the ideal of complete neutrality, such as standpoint theory, which calls for centering excluded perspectives, and positioning, which acknowledges knowledge arises from multiple cultural and contextual positions.

  • The goal should not be an unattainable universal objectivity but rather more inclusive and complete knowledge that accounts for the situated nature of all perspectives.

  • The passage discusses critiques of scientific research that is conducted primarily on male subjects and perspectives, without considering gender differences or the experiences of women. It argues this approach treats the male experience as the “norm” that women vary from.

  • Feminist approaches advocate disclosing one’s subject positions and perspectives to acknowledge the limits of any individual’s knowledge claims. Researchers should embrace multiple perspectives rather than viewing them as biases.

  • The concept of “data visceralization” is introduced, going beyond just visual data representations to involve other senses. This can make data more accessible and capture emotional and cultural aspects beyond just facts and reason.

  • An example performance art piece called “A Sort of Joy” is discussed, which uses metadata from an art museum collection to depict gender biases in whose work is represented, through non-visual and experiential means.

  • Throughout, the passage advocates moving beyond just rational, objective viewpoints to incorporate emotional, cultural and multiple subjective perspectives into how data is represented and understood.

  • The piece describes an art performance called “A Sort of Joy” that visualized gender disparities in the MoMA art collection by having performers read out male and female names from the collection.

  • It took over 3 minutes for the first female name (“Mary”) to be said, highlighting the male dominance in the collection. Additional female names were also spaced further apart compared to male names.

  • From a data perspective, the performance simply counted and grouped names, but presenting it as an experience over time made the audience wait, listen, and feel the insights rather than just seeing them.

  • Other examples are given of projects that viscerally present data through embodied experiences like walking tours, interactive chairs, fashion shows, sound installations, maps you can listen to, and data-driven pies.

  • Viscerally presenting data can engage emotions, leverage different senses, and enable different types of learning compared to standard charts and graphs.

  • A challenge is representing uncertainty, as people struggle to recognize it even when explicitly shown in visualizations. Experiencing uncertainty perceptually through affects and emotions may be more effective.

  • The New York Times election needle that jittered based on forecast uncertainty is discussed as an example of showing rather than just telling about uncertainty. Its affective experience aimed to convey the real-time ranges and shifts.

  • The New York Times’ election forecast graphic in 2020 that jittered represented current best practices for communicating uncertainty visually. It gave readers an intuitive sense of the uncertainty through movement.

  • However, context is important, and there is no single rule that a design choice made in one context will work in another. It’s never a good idea to say “never” in design.

  • The “god’s-eye view” or overview perspective (sometimes called the “god trick”) can be used positively to advocate, empower, or contest dominant narratives, not just negatively.

  • The map “Coming Home to Indigenous Place Names in Canada” leverages the authoritative overview perspective to assert Indigenous sovereignty and presence by mapping Indigenous place names across Canada.

  • Elevating emotion and embodiment is an important principle for data feminism. Features like emotion and the body are often excluded from data visualization in favor of rationality, but can be valuable tools when included effectively. The margins, liminal perspectives, and marginalized groups should be centered in design.

In summary, the passage discusses the importance of context in design, provides examples of how the “god’s-eye view” can be used positively, and argues for elevating emotion and embodiment as valuable aspects of data communication that are often excluded.

  • The chapter discusses the importance of counting and classifying people in a way that is inclusive and does not perpetuate oppression. It uses examples of non-binary individuals who struggle with gender binary options on websites.

  • Counting what is included and excluded shapes policies and visibility. Feminist scholars advocate using both quantitative and qualitative data to identify collective oppression beyond individual experiences.

  • While Facebook expanded gender options in 2014, a study found it still classifies users into a male/female binary for advertisers. This shows corporations control data terms even as individuals experience limits of classification systems.

  • Feminists have analyzed gender and sex as social constructs rather than essential or natural categories. The binary developed historically from a hierarchical view of sexes, not an inherent biological distinction. Data collection needs to challenge binaries and hierarchies to be truly representative and empowering.

  • The passage discusses how classification systems have historically been used to categorize and discriminate against marginalized groups, like women, people of color, disabled people, and LGBTQ+ individuals.

  • It traces how the concept of race developed in the 18th century through “scientific racism” that purported to scientifically classify people by race. This led to differential treatment and pseudosciences justifying oppression.

  • All data and information systems require some form of classification to function, but these systems often encode hidden values and hierarchies that are not questioned.

  • The US Census is used as an example, where debates emerged around categorizing multiracial people and how census data impacts political representation and funding.

  • The story of Michael “Mikey” Hicks, an 8-year-old who is routinely flagged by TSA due to having the same name as someone on a terrorist watchlist, is presented to show how rough classification criteria can unfairly impact individuals.

  • In summary, the passage critiques how classification systems have historically masked oppression and calls for critically examining the values and impacts inherent in how data and information is categorized.

Here are the key points made in the passage:

  • The airport security scanners classify travelers as male or female based on a TSA agent visually deciding their gender and selecting the corresponding option on the scanner. This loads a gender-specific algorithm profile against which the person’s measurements are compared.

  • If a person’s measurements diverge from the statistical norm for their assigned gender, they will trigger a “risk alert” and be subjected to a full-body pat down, even if the discrepancy is just because the TSA agent guessed wrong or because the person is non-binary/transgender.

  • This flawed classification system is an example of why design justice is needed in relation to data systems. Systems should not rely on binary gender assumptions that exclude and pathologize people who do not fit into the male/female boxes.

  • The passage argues that collecting accurate gender data beyond male/female categories and representing the diversity of gender identities could help address inequality. However, accurate representation alone does not always lead to positive change, so alternative data collection models still need to be carefully considered.

The key point is that the airport security scanner system relies on oversimplified male/female gender classifications, which can result in discriminatory screening and invasive pat downs for non-binary and transgender travelers whose bodies do not neatly fit the predefined algorithms. This illustrates the need for design justice in technical systems to avoid inadvertently excluding or harming certain groups.

  • The passage discusses the complexity of collecting and representing gender data. While having non-binary gender options makes sense, it can also pose risks by making those identities more visible.

  • It notes the risks depend on the context, such as the laws and norms of a given place. In some contexts, collecting detailed gender data could expose people to discrimination, violence or other harms.

  • Ethical decisions around gender data collection should consider these potential harms. Depending on the situation, it may be best to avoid collecting gender data, make it optional, collect only binary options, or take other precautions.

  • Beyond gender, binaries also structure how we think about race and other attributes. Challenging binary thinking can work towards more just and equitable data practices and outcomes.

  • Even when inheriting binary data, designers can “hack” binaries through techniques like color choice to better represent complexity and avoid stereotypes. Representing an absence of data can also be meaningful.

  • Visualizing gender itself is highly challenging given its complexity beyond binaries. Effective representations acknowledge its multidimensional nature.

The passage discusses how counting and representing data about human characteristics can be complex due to nuances and changes throughout a person’s life. It focuses on a diagram called “Beyond XX and XY” that visualizes the sex spectrum and how sex is not fixed but changes over time. The diagram rejects flawed categories and research in order to provide a more nuanced visualization of sex as a spectrum rather than a simple binary.

The passage also discusses how counting and classifying people can paradoxically harm those who are exposed without consent due to social and historical factors. It argues that counting must consider consent, safety, cultural context and dignity. Examples of how counting policies on platforms like Facebook can endanger marginalized groups are provided.

Finally, the passage discusses how counting can also empower and heal when done by and for communities in a respectful way that humanizes those being counted. The Colored Conventions Project and breast pump hackathons are used as examples of counting efforts that contribute to representation, recognition and improving lived experiences. In summary, the complexities of ethically and respectfully counting aspects of human identity and experience over a lifetime are examined.

  • The Anti-Eviction Mapping Project (AEMP) is a collective in San Francisco that maps evictions through a collaborative, multimodal approach grounded in antiracist, feminist, and decolonial methodologies.

  • They have over 70 maps on their website related to evictions, displacement, tech buses, property owners, declining black population, and more.

  • They are also working on an atlas called Counterpoints that will cover topics like migration/relocation, gentrification, Indigenous and colonial histories of the Bay Area, and speculation about the future.

  • One of their long-standing collaborations is with the Eviction Defense Collaborative (EDC), a nonprofit that provides legal representation for evicted tenants.

  • The EDC collects demographic data on evicted tenants that the city does not, including race and income. They share this data with AEMP.

  • AEMP and EDC have worked together on EDC’s annual reports and additional analyses of evictions with a focus on race.

  • AEMP takes a pluralistic, multimodal approach to mapping evictions through collaboration with community groups and using various data sources and storytelling methods.

  • The AEMP collects data on tenant evictions and gentrification through maps, timelines, oral histories and other methods. Some maps like the Tech Bus Stop Eviction Map clearly show correlations between evictions and factors like proximity to tech company bus stops.

  • Other maps like Narratives of Displacement and Resistance intentionally don’t show clear patterns, obscuring the map with bubbles representing thousands of evictions and oral history locations. Clicking on bubbles plays audio stories of displaced residents.

  • This map enacts dissent against both san francisco policies and clean/clear data visualization conventions. It aims to document displacement and resist through critical creative means.

  • The AEMP embraces pluralism by including diverse voices and perspectives at all stages, rather than telling one clear story. It values plurality over clarity and cleanliness.

  • Data cleaning is often seen as necessary to impose order and structure on “messy” data. But this process could lose certain perspectives or impose others. Not all contexts value cleariness and control above diversity of voices.

This section discusses some key ideas around data cleaning and the histories and assumptions underlying it:

  • The ideas of cleanliness and control in data have troubling historical roots in eugenics from the 19th century. Many early statisticians were also leaders in the eugenics movement.

  • While the most extreme eugenics ideas have been stripped away, an underlying belief in the benefits of clean and controlled data still remains.

  • Cleaning data can function as a “diversity-hiding trick” by assuming there is a single correct order or structure for data. The messiness contains important context about how the data was collected.

  • Data is shaped by its “data settings” - the technical and human processes of collection. Cleaning can separate data from this important context.

  • Once data is analyzed by “strangers” removed from the original collection, such as data scientists from different places, cleaning may be needed but can also be destructive.

  • Common metaphors used for data scientists like “unicorns,” “wizards,” “ninjas,” and “rock stars” portray them as solitary genius males, ignoring the support and education behind their work. Terms like “janitors” also have problematic gender and class connotations.

  • There are risks of “epistemic violence” by privileging dominant ways of knowing over local ways in the data collection context. More attention needs to be paid to data settings and histories.

  • The Anti-Eviction Mapping Project (AEMP) collects eviction data in California to better understand the housing crisis. They aim to build community solidarity among participants.

  • The Eviction Lab approached AEMP to request data, but AEMP had concerns about privacy protections and how the data might be used. Instead of addressing these concerns, Eviction Lab bought lower quality data from a real estate broker.

  • AEMP and other housing groups found that their data showed 3 times as many evictions in California compared to Eviction Lab’s numbers, highlighting inaccuracies in Eviction Lab’s data for that state.

  • Eviction Lab prioritized clean, standardized national datasets that could be purchased quickly, rather than building trusted relationships locally for more accurate data. This maintains the image of a “solitary genius” rather than recognizing collaborative, community-led work.

  • AEMP involved many non-experts in mapping to build technical skills and relationships. This reflects a feminist view that plural voices yield a richer understanding than a single expert view. Transparency about methods and positions can also be a feminist approach.

  • Designing from the margins argues that designers and engineers should actively work to dismantle the distinction between centers and margins in society, rather than just engaging people at the margins.

  • The Design Justice Network promotes centering the voices of those directly impacted by design processes.

  • Data for co-liberation aims for designs where dominant and minoritized groups work together to free themselves from oppressive systems. It requires participation from a diverse set of experts and shifting the goal from “doing good with data” to co-liberation.

  • Key differences between data for good and data for co-liberation are that the latter emphasizes leadership from minoritized groups, community ownership of data and resources, grounding analysis in community perspectives, facilitating rather than experts, knowledge transfer, and building social infrastructure.

  • Examples are given of data murals created through partnerships between communities and the Bhargavas that employ both data analysis and civic processes to tell community stories and build social solidarity.

  • The chapter discusses several examples of data projects designed for “co-liberation” rather than just “good”. These projects originate from community-identified needs and involve community members in all stages of the work.

  • One example is a data mural project in Somerville, MA where youth helped collect data, paint a mural, and present it to officials.

  • Other examples include projects using data to defend Indigenous land rights and push for affordable housing.

  • A key aspect of these projects is knowledge transfer from external collaborators to the community and building community solidarity around issues.

  • The chapter argues this model of “data for co-liberation” can scale up in a way that remains community-focused, as demonstrated by the Global Atlas of Environmental Justice, a large global database of environmental conflicts built through pluralistic, relationship-based processes.

  • In summary, the chapter advocates embracing pluralism at all stages of data work as a feminist strategy to mitigate risks of epistemic violence and extractive approaches, and shows how this is compatible with large-scale projects.

  • The article discusses how context is important when analyzing data but often missing from open datasets found online.

  • It uses the example of FiveThirtyEight erroneously reporting kidnapping statistics in Nigeria based on raw data from the GDELT project, which counts media reports rather than actual events. This shows how lack of context can lead to misinterpretation.

  • The concept of “Big Dick Data” is introduced to describe big data projects that make grand claims but lack context and transparency about limitations.

  • Many open datasets provide little to no metadata or documentation about how the data was collected and the social/cultural conditions that shaped it.

  • Without understanding the context, it’s difficult to properly analyze the data and avoid misinterpretations. Feminist data analysis emphasizes understanding this situated knowledge and wider context around how data was produced.

  • More context is needed for datasets, like one shown from Brazil about government procurement, to really understand what the variables mean and how the system works on the ground. Lack of context poses challenges for data exploration.

So in summary, the article advocates the importance of understanding the context and limitations of data sources to conduct accurate and ethical analysis, using examples to show how missing context can lead to errors.

  • Open data aims to make government and other public data freely available, but often lacks proper context and documentation to make the data meaningful and usable. This is known as “zombie data”.

  • Chris Anderson argued in 2008 that with big data we no longer need hypotheses or theories, as “the numbers speak for themselves.” However, data is always situated within social and historical contexts.

  • Algorithms like Google Search perpetuate harmful stereotypes due to correlation without understanding context, as demonstrated by Safiya Noble. Correlation is not enough and can reinforce oppression without acknowledging context.

  • Sexual assault statistics reported under the Clery Act also demonstrate how numbers alone can be misleading without context. Further investigation by students found schools with higher reporting may actually be providing more support to survivors, while less reporting does not necessarily mean fewer assaults. Understanding context is key to interpreting any data.

In summary, the passage argues that open data initiatives often lack context, which can lead to misinterpretation of numbers and perpetuation of harmful assumptions without understanding the social contexts and systems that produce the data. Context is crucial for ethical and responsible use of data and knowledge.

  • The Clery Act requires colleges to report annual crime statistics like sexual assaults, but there are incentives for underreporting to avoid negative perceptions. Reporting is also self-reported rather than independent audits.

  • Studies have found large discrepancies between Clery-reported sexual assault numbers and anonymous campus climate surveys, indicating underreporting. Survivors may not report due to stigma, trauma of reliving the experience, or lack of support.

  • Power dynamics like race and sexuality also impact reporting. A complaint alleged Columbia mishandled LGBTQ sexual assault cases from lack of proper training.

  • Data is never “raw” but already “cooked” through social, political and historical factors in its collection and reporting. Taking numbers at face value ignores this context.

  • Context is needed to understand what datasets actually measure and potential biases. Simply analyzing the data alone cannot reveal these deeper insights about structural biases in its collection and the realities it aims to represent.

So in summary, the key points are around the limitations of self-reported crime data due to incentives for underreporting, lack of survivor support encouraging reporting, and need for contextual analysis to understand biases and gaps in what data represents. Simply analyzing numbers alone ignores important context around their collection.

Here are the key points about communicating context:

  • Context matters not just in data acquisition and analysis, but also in how results are communicated and framed.

  • Simply reporting numbers and findings without context can undermine and misrepresent the actual meaning and implications of the research.

  • In the example of mental health diagnosis disparities in jails, communicating the numbers without mentioning race/ethnicity, racism, or disparities fails to properly represent the study’s findings of discrimination.

  • Choosing language like “inmates” rather than referring to people can be dehumanizing in some contexts like mass incarceration.

  • Naming forces like racism is important for accurate representation, though some fields like journalism conventionally avoid such “accusations” and “characterizations” in the name of neutrality.

  • However, for many people racism exists as a factual reality supported by evidence, so refusing to name it can serve to protect the status quo rather than promote understanding.

  • Communicating context fully involves framing results in line with the actual forces, such as racism, identified by the underlying research.

So in summary, context is crucial not just in data work but also in how results are presented and framed to outside audiences. Failing to communicate context can undermine and distort the meaning and implications of findings.

  • When communicating data results, it is important to place numbers in context and name structural forces like racism, sexism, etc. if they are present in the data. Simply letting numbers “speak for themselves” often leads to misinterpretation.

  • Members of dominant groups especially have a responsibility to recognize and discuss systemic oppressions like racism and sexism that influence data.

  • Graph titles and subtitles should avoid perpetuating deficit narratives about marginalized groups and instead focus on advantages given to dominant groups.

  • Tools are being developed to better attach context to data, like data biographies, datasheets for datasets, and data user guides. But responsibility for providing context remains unclear.

  • Collecting qualitative data in addition to quantitative data is important for researching groups experiencing oppression, as their lived experiences are not fully captured by statistics alone. Context around limitations and social power differentials is also needed.

Here is a summary of the relevant passage:

The passage discusses whose responsibility it is to properly document data sources and limitations. It argues that placing the full burden of extensive background research on individual people and small teams working on tight deadlines is unreasonable.

It examines potential responsibilities of different actors:

  • Data publishers: In some cases like GDELT, data publishers overstate capabilities and fail to adequately document limitations.

  • Universities/governments: When they self-report data like campus sexual assault statistics, they are governed by their own interests rather than transparency. Governments are under-resourced to fully verify and document limitations.

  • Data intermediaries: Entities like librarians, journalists, nonprofits could help clean and contextualize data for public use, but need more funding and capacity-building to do this at a large scale.

Overall, it argues that until we invest as much in providing context for data as in just publishing raw data, public information resources will remain subpar and potentially dangerous due to lack of context. Contextualization should be a major focus for open data advocates and organizations going forward.

Here is a summary of key points from Chapter 7:

  • Much of the invisible labor that goes into creating data products and visualizations is not acknowledged or credited. This includes the people who collected and processed the raw data.

  • Corporations have an incentive to keep this invisible labor out of public view. If work is unpaid or underpaid, it is less valued culturally and economically.

  • The term “invisible labor” encompasses unwaged, underwaged, and waged work that happens behind the scenes or takes non-physical forms. Examples include domestic labor, social media “likes” and posts, and crowdsourced data tasks.

  • Crowdsourcing projects by companies like Netflix, the Guardian, and for text recognition contribute unpaid labor but are framed as opportunities to help the public good. However, not all groups have equal time, ability, and motivation to participate.

  • Platforms like Amazon Mechanical Turk exploit “on-demand workers” with low wages for tasks that support data science. Many are internationally outsourced and paid less than minimum wage.

  • Even at large companies, data entry work is profoundly undervalued compared to the knowledge created, as seen with Google’s book scanning workers. This labor force also tends to be disproportionately women and people of color.

  • The passage discusses the work of information studies scholar Lilly Irani, who argues that today’s hierarchy of data labor mirrors older hierarchies in technologies like computing that disadvantaged women, minorities and lower classes.

  • Irani studied Amazon Mechanical Turk, showing how it exploits “workers” with low pay and poor working conditions. In 2008 she created a tool called Turkopticon to help workers assess tasks and report issues.

  • However, the tool had limited reach as workers lacked time due to the same poor conditions it aimed to address. This highlighted the precarious nature of much digital labor.

  • This “cultural data work” involves content moderation, transcription, captioning, and is often performed by vulnerable groups like women of color with little power or protections.

  • The passage draws parallels between this modern digital exploitation and the historical exploitation of slavery. It cites the infamous Zong slave ship incident as an example of how capitalism reduces humans to economic assets.

  • Scholars are examining “data production” to understand how power dynamics shape the creation of data, algorithms and technologies, revealing hidden human labor and linking products to real human and environmental impacts.

  • The goal is to “show the work” and make invisible labor visible in order to properly value it and understand technology’s true costs to people and the planet.

  • The diagram “Anatomy of an AI System” by Kate Crawford and Vladan Joler attempts to chart all the human labor, data, and planetary resources that go into creating a product like the Amazon Echo. It emphasizes giving credit to the broad range of invisible work that supports technological systems.

  • Feminist practices of citation aim to make visible contributions that are often erased or overlooked, like the labor of women and minorities. Formally crediting this invisible work is a way to resist its erasure.

  • Any technological project relies on many types of labor that may be difficult to document fully, like project management, design, writing, and customer support. Visualizing this “underwater” labor is important to give a more complete picture of what supports a project.

  • Even data itself can reveal invisible labor, like Benjamin Schmidt’s visualization of Library of Congress cataloging records showed the efforts of cataloguers over decades.

  • Emotional and affective labor involved in customer service, technical support, and minority groups navigating workplace biases are also important types of invisible work that support technological systems and should be recognized.

This passage discusses efforts to visualize and acknowledge types of labor that are often invisible, such as emotional labor, caregiving labor, and data maintenance work.

It summarizes the Atlas of Caregiving project, which uses data collection tools, interviews and logs to visualize the range of physical and emotional labor involved in caring for a chronically ill family member. It also describes the “Bruises—the Data We Don’t See” project, which uses alternative visualization techniques like color and fluid timelines to depict the experience of one family providing care.

The passage discusses how care work has historically been undervalued, despite efforts by feminist scholars and artists dating back to the 1960s/70s to recognize care and maintenance as important labor. It outlines ongoing work by groups like the Maintainers to apply theories of feminist labor studies and make data maintenance labor more visible in tech fields. Overall, the key point is that visualizing different types of unseen labor can help bring acknowledgement and value to work that often goes unrecognized.

  • Maintainers refers to people who work in libraries, archives, and preservation fields to ensure current knowledge remains accessible for future generations. Their work facilitating access to future knowledge can be viewed as a form of care work.

  • There is increasing attention paid to types of invisible labor like care work, as more work becomes virtual. Professional care workers have long dealt with issues like undercompensation and precarious work.

  • Unions and advocacy groups are using new technologies like apps and data to organize workers and resist inequities. However, apps connecting caregivers to employers often do not solve systemic problems and can reinforce unequal power relations.

  • Alternatives centered around worker needs are emerging, like an app developed by a domestic workers’ alliance that allows clients to contribute to worker benefits accounts. While better than nothing, systemic changes are still needed.

  • Showing the labor behind data work through “show your work” practices can help value invisible work and understand the true costs and impacts of data science. Data can also be used to highlight undervalued care work.

  • The summary emphasizes the challenges faced by care and domestic workers, efforts to organize and advocate for their rights, and the importance of acknowledging all labor that goes into data science and knowledge production.

The passage discusses various forms of collective organizing and resistance happening within the technology industry. It outlines how workers at companies like Slack or Google could hypothetically organize strikes or slowdowns using the technology platforms themselves.

It then provides examples of actual organizing efforts, like the Tech Workers Coalition which brings together tech employees and service workers. Groups like the Design Justice Network and Data for Black Lives are building multi-stakeholder movements around issues like racial justice in technology and data.

The Google Walkout is discussed as one high-profile action that achieved some wins but not all demands. Organizers have reportedly faced retaliation, and the company continues to resist full transparency and inclusion.

Overall, the passage argues technology workers and supporters are experimenting with creative ways to channel digital connections into real-world solidarity and social change. However, entrenched power structures within corporations represent ongoing challenges to be addressed through continued grassroots organizing and intersectional feminism.

  • The authors introduce the concept of data feminism, which aims to challenge power imbalances and promote equality through a feminist lens in data science.

  • They outline seven principles of data feminism: examine power, challenge power, elevate emotion/embodiment, rethink binaries/hierarchies, embrace pluralism, consider context, make labor visible.

  • The principles were derived from intersectional feminist thought over decades. However, they acknowledge there are many valid starting points for challenging oppression in data.

  • They provide examples of work from different fields and communities that align with data feminist values, like spatial analysis of displacement, queer theory, Indigenous data sovereignty, decolonizing design, model documentation, civic tech projects, and artworks.

  • The goal is not uniformity but nurturing diverse approaches and building links between them to mobilize resistance to power differentials in data and imagine alternatives.

  • They conclude by emphasizing the importance of multiplying these efforts now, before data structures and norms are fully cemented, to shape the emerging data landscape in an equitable way.

  • The passage advocates for equity and prioritizing marginalized voices and perspectives, specifically in how information is presented.

  • It emphasizes looking at both individual and structural issues of injustice. Those with relative power should listen differently to understand injustice from those experiencing marginalization.

  • It argues for prioritizing the knowledge and perspectives of communities most proximate to issues, as they know problems and solutions intimately.

  • It acknowledges that data often lacks context and complexity when human experiences are reduced to data, and has historically been used to oppress. It aims to present data with more context and acknowledge limitations.

  • The authors situate themselves and their institutions, recognizing knowledge is shaped by one’s positionality. They strive for reflexivity, transparency and accountability in their work on issues of injustice.

  • Draft and final metrics are presented to track how well the book lives up to values of centering marginalized voices. Some goals were better met than others in the final version.

The authors worked to audit data references in their book “Data Feminism” in order to remain accountable to their values statement. They categorized each reference by demographic details like gender, race, region of origin, and whether they represented examples of good or bad data practices.

Some categories like importance and whether a reference provided a non-visual example were straightforward, but identifying details like race and gender required research and assumptions without self-identification. The audit process highlighted difficulties in clearly establishing identity categories.

Future audits should take these challenges seriously and acknowledge limitations without identifiers. The goal was not just representation but accountability to intersectional values. Some royalties from the book will be donated to community organizations modeling data feminist principles, like Indigenous Women Rising and Charis Circle.

Here is a summary of the figure credits from pages 228-231:

The figures come from a wide range of sources, including academic papers, news articles, government datasets, nonprofit organizations, artistic works, and social media platforms. Many figures credit the original artists, researchers, or organizations who created the work. There is also attribution for figures that have been adapted or reproduced from their original sources. The sources provide context and background for the data visualizations, images, and maps included in the text. Proper attribution is given to respect intellectual property and follow academic standards.

  • The Eviction Mapping Project is a collaborative effort that maps evictions and displacement in San Francisco. It collects data and stories from residents facing eviction to document the human impact of gentrification and rising housing costs.

  • The project collaborates with the Ruth Asawa School of the Arts to produce interviews and videos telling the stories and narratives of those displaced. Students are involved in conducting, filming, and editing the interviews under the guidance of facilitators.

  • The project’s website hosts these narratives and stories to put a human face on eviction data and statistics. It aims to document resistance to displacement and raise awareness of the issues surrounding affordable housing and gentrification.

  • By mapping evictions and collecting personal stories, the project provides both a quantitative and qualitative understanding of eviction and displacement in San Francisco neighborhoods experiencing rapid economic and demographic change.

So in summary, the Eviction Mapping Project documents evictions and displacement through maps, data collection, and narratives in order to raise awareness of housing issues and the human impact of gentrification in San Francisco. It’s a collaborative effort that involves community members sharing their stories.

  • Feminism is described as having occurred in waves, with the third wave beginning in the 1990s and characterized by increased attention to intersectionality. Some propose we are now in a fourth wave since the 2010s with social media.

  • However, other scholars reject the wave model for how it overlooks the long-term work of activists, especially women of color, that took place both during and between waves.

  • Intersectionality originated even earlier with thinkers like Anna Julia Cooper in the 19th century and was a key concept developed by the Combahee River Collective in the 1970s.

  • Kimberlé Crenshaw is credited with coining the term intersectionality in the 1980s-90s through her legal writings examining how anti-discrimination frameworks failed Black women.

  • Positionality, oppression, patriarchy, sexism, and cissexism are discussed as relevant concepts to understand how different identities intersect and relate to systems of power and discrimination.

  • The summary emphasizes examining feminism and feminist history through an intersectional lens that centers the sustained efforts of marginalized groups, rather than just prominent waves or figures.

  • The passage references three digital humanities projects conducted by Lauren F. Klein and others: Data by Design, Vectors of Freedom, and the Floor Chart Project. More information on these projects can be found at or in certain published articles.

  • It discusses several sources that relate to concepts of fact, data, and objectivity in history and science, including works by Mary Poovey, Miriam Posner/Lauren Klein, and Daniel Rosenberg.

  • It compares computing power between an iPhone XR and an IBM System/360 Model 30 from the 1960s to demonstrate increases in processing capability over time.

  • It lists several sources that examine historical practices of counting and collecting statistics on populations, such as death tables, colonial censuses, and statistics on ethnic groups.

  • It references scholarship on the histories of slavery, eugenics, and surveillance of black populations in relation to data practices.

  • It discusses concepts such as “dataveillance”, predictive policing technologies, and criticism of how data can perpetuate discrimination and societal inequalities.

  • It provides examples of community organizations and artists addressing these issues through projects focused on data and social justice.

Based on the information provided, there is no information about the book cited (York: The Feminist Press, [1973] 2010). The passage provided citation details for other sources but did not include any discussion of or quotes from the cited book. A summary cannot be provided based on the information given.

Here is a summary of the key points from the provided references:

  • Women remain significantly underrepresented in computing and technology fields. While they make up around half of all college graduates, women receive only around 26% of computer science degrees. Their representation is even lower in some subfields like machine learning.

  • Gender and racial biases are present in AI systems due to the biases of their predominantly male and white creators and the uneven representation in the training data used. Facial analysis and recruitment tools have been shown to perform poorly on women and people of color.

  • Collecting more diverse training data, including faces from more countries and ethnicities, can help improve the accuracy and fairness of facial recognition systems. However, care must be taken to obtain informed consent and avoid exploiting vulnerable populations.

  • Groups are calling for regulation and oversight of facial recognition to prevent harmful uses like secret government surveillance and by law enforcement without public accountability. An ethical pledge has been proposed to commit users of this technology to principles like preventing unlawful discrimination.

  • Joy Buolamwini, a researcher at MIT, found that facial recognition systems from major tech companies like Amazon, Microsoft, and IBM performed worse at identifying women and darker-skinned individuals.

  • She brought this issue of bias in AI to public attention through research papers, conferences talks, and advocacy efforts. But Amazon has actively opposed her research and not signed her calls to limit sales of its Rekognition software to police.

  • Other top AI researchers have defended Buolamwini and called on Amazon to stop selling Rekognition to police given its biases. She continues working to make algorithms fairer and more just.

  • More broadly, there are issues of missing, biased or incomplete data when it comes to marginalized groups. Researchers and activists are working to collect counter-data to address these gaps and push for more inclusive and just data and systems. Areas discussed include gender data, maternal health outcomes, environmental health, police killings, violence against women, and Indigenous women. Community-driven data collection efforts aim to fill information voids and give voice to impacted populations.

Here is a summary of the key points from The Curious Journalist’s Guide to Data:

  • The book provides an introduction to collecting, analyzing, visualizing and reporting with data for journalists.

  • It covers the basics of working with data including finding publicly available datasets, scraping data from the web, obtaining private data through records requests.

  • Key concepts in data journalism like open data, data verification, data ethics and data visualization are explained.

  • Examples of successfully executed data journalism projects are used to illustrate best practices in working with data to uncover stories.

  • Guidelines are given for analyzing data through tools like spreadsheets, SQL and coding languages. Basic data analysis and visualization techniques are demonstrated.

  • Legal and ethical issues around using data in reporting like privacy, attribution and intellectual property are discussed.

  • The book aims to equip journalists who may not have an extensive technical background with the knowledge to leverage the power of data in their storytelling.

  • The passage discusses important early works that examined how whiteness became constructed as a racial identity in the US, and how that shaped the American working class.

  • It references a 1917 Supreme Court case (Buchanan v. Warley) that declared race-based zoning unconstitutional, but many local laws continued such exclusionary practices.

  • It brings up the concepts of racial capitalism and how credit scores, tax codes, and other systems have been shaped by and reinforce racism.

  • Cheryl Harris’ theory of whiteness as property is cited, and the passage discusses how technologies currently distribute and secure life chances unequally along racial lines.

  • Redlining, both historical and in newer forms like digital/discursive redlining, is discussed as still impacting communities of color.

  • The risks of bias in algorithms and predictive systems are examined through the example of ProPublica’s investigation into the COMPAS risk assessment tool.

  • The need for algorithmic accountability efforts and pushing for transparency is emphasized.

This summary covers key aspects of the provided source material:

  • The book Indigenous Statistics: A Quantitative Research Methodology by Chris Andersen and Jude Walker outlines an approach to statistics and quantitative research from an Indigenous perspective.

  • It discusses doing numbers “our way” and centering Indigenous communities and knowledges in statistical work.

  • The book advocates for community-based research approaches, sovereignty over data, and addressing biases in mainstream statistical methods.

  • It presents alternative concepts like “relationality” and places-based methods drawn from Indigenous epistemologies.

  • Overall, the book presents a framework for Indigenous self-determination in quantitative research against the historical context of data-based oppression of Indigenous peoples. It discusses decolonizing statistics to respect Indigenous rights and knowledges.

Here is a summary of the key points from es_Single.pdf and related sources:

  • The Our Data Bodies Digital Defense Playbook provides guidance on digital self-defense strategies for communities. It focuses on building connections across communities to advocate for co-liberation.

  • The Appolition app takes users’ spare change and converts it to bail money to address racial inequities in the cash bail system.

  • The #TechWontBuildIt hashtag saw tech workers publicly reject opportunities to work for companies involved in controversial projects, like Microsoft’s military contract or Amazon’s work with ICE.

  • Teachers developed a math curriculum called City Digits around lottery statistics to teach concepts in a real-world context and address spatial justice. Students analyzed lottery data and probabilities.

  • Teaching approaches like City Digits aim to disrupt the traditional “factory model” of STEM education by integrating social justice and humanities perspectives.

  • Participatory projects that engage youth in analyzing and presenting data can build skills while amplifying marginalized voices. City Digits involved student presentations to faculty and at conferences.

  • Leadership that reflects diverse backgrounds and priorities is important to avoid “privilege hazard” and elevate minoritized perspectives within organizations.

So in summary, it discusses various digital defense, advocacy and education projects aiming to address social inequities through technology and by reforming STEM pedagogy to be more inclusive and justice-oriented.

This summary covers discussions from multiple sources on the topic of rational, scientific, and objective viewpoints versus mythical, imaginary, and impossible viewpoints in data visualization and analysis. Some of the key points discussed include:

  • The importance of objectivity and distance in statistical and data analysis, though this is challenging to fully achieve given personal biases (paragraphs 1-2).

  • Critiques of the “god trick” of visualization and the pretense of complete objectivity and distance, arguing this can obscure power dynamics and oppression (paragraphs 3-4).

  • The influence of emotions, values, and partial perspectives that are inherently situated in any knowledge or analysis (paragraphs 5-6).

  • Calls for recognizing subjectivities, emotions, and situated knowledges rather than pretending neutrality or universality (paragraphs 7-8).

  • Issues of binarism, inclusivity, and representing diversity in data and analyses (paragraphs 9-10).

  • Examples of feminist scholars and techniques emphasizing partial perspectives, perceptions beyond vision, and critiquing exclusion in traditional frameworks (paragraphs 11-13).

So in summary, it discusses moving beyond pretenses of pure objectivity to more situated, inclusive, and critical approaches that acknowledge subjective and emotional factors.

Here is a summary of the paper:

The paper by Burks et al. describes a technique for supporting the exploration of viscous fingering phenomena in large-scale fluid simulation ensembles. Viscous fingering refers to the instability that occurs when a low-viscosity fluid displaces a higher-viscosity one in a porous medium. The authors present an approach called “Details-First, Show Context, Overview Last” that leverages brushing and linking to connect views of simulation details, contextual information, and global overviews. This enables researchers to efficiently navigate large ensembles of viscous fingering simulations at different levels of granularity. An interactive web-based system called VFView is developed to demonstrate this multi-view exploration approach. User studies suggest the technique facilitates insight generation and understanding of simulation behavior compared to conventional visualization methods. The work thus provides useful tools to analyze complex phenomena in large-scale simulation data.

  • A new generation of data visualizers and designers is challenging the anti-emotion and anti-embellishment dogma in data visualization. This includes people like Jessica Bellamy, Giorgia Lupi, Stefanie Posavec, Federica Fragapane, and Kelli Anderson.

  • They are bringing artistic and design techniques to data visualization to make it more emotive and rhetorical, moving beyond just abstract representation of data.

  • Productive collisions between data scientists (who focus on analytic methods and abstraction) and artists/designers/humanists (who focus on rhetoric, form, and embodiment) can help overcome the false dichotomy between reason and emotion.

  • Engineering such collaboration can be the surest way to advance data visualization by combining different skills and perspectives from data scientists, artists, designers, and humanists. This brings together their differing strengths in analytics, abstraction, rhetoric, form, and embodiment.

Here is a summary of the visual timeline of racial categories on the US census according to the Pew Research source:

  • 1790 - The first US census only had categories for “Free White males”, “Free White females”, and other people (enslaved Africans and Native Americans).

  • 1850 - “Mulatto” was added as a category to designate people with mixed black-white parentage.

  • 1860-1930 - Racial categories became more complex, adding categories like “Chinese”, “Japanese” etc. to account for immigrants.

  • 1930 - The “Mulatto” category was replaced by “Negro” to cover all non-white black populations.

  • 1960 - “Negro” was replaced by “Black” as the main category for black Americans.

  • 1970 - People could check more than one race for the first time.

  • 2000 - “Some other race” was added to accommodate newer immigrant groups. The current categories of “White”, “Black or African American”, “Asian”, “Native Hawaiian/Pacific Islander” and “American Indian/Alaska Native” were introduced. People could also select “Two or more races”.

So in summary, the timeline shows how US census categories around race and ethnicity have evolved over time to be more inclusive and complex, from just “White” and “Non-White” to accommodate diverse populations.

Here is a summary of the key points from the article “The Design of Nothing: Null, Zero, Blank” by Andy Kirk:

  • The article discusses the concepts of absences, nulls, and zeros, which Kirk refers to as the “design of nothing.” These concepts represent the absence of data or information.

  • Null values represent unknown or undefined data. They indicate a lack of information rather than a value of zero. Nulls are important for representing missing or unknown data in databases.

  • Zeros can represent either a numeric value of zero or the absence of a quantity. The meaning of a zero value depends on the context.

  • Blanks or empty fields are used to represent missing or absent information in visualizations and user interfaces. They allow for spaces where data is not yet present.

  • Nothing or absence of data needs to be consciously designed and represented, just as the presence of data does. How nulls, zeros, and blanks are designed can have important implications for how data is interpreted.

  • The article argues that designers need to carefully consider the meaning and communication of absent data, not just present data. Representing nothing is a critical part of data design.

  • Historically, eugenics promoted the reproduction of white people while discouraging reproduction of others through forced sterilization and even murder. Karl Pearson, who held the first Eugenics chair at University of London, developed many statistical concepts still in use today.

  • Eugenic assumptions of genetic superiority have never fully gone away and there is a long history of sterilization programs targeting incarcerated populations in the US.

  • Certain statistical techniques could be inherently racist depending on how they are designed and applied. Critiques note the need to recognize implicit biases.

  • Cleaning and organizing data often erase important context about how and why the data was collected. It is important to retain and make visible information about data collection processes and limitations.

  • Roles of data workers like annotators and labelers are often invisible in final datasets and results. More needs to be done to document and credit their contributions.

  • Intersectionality as a framework emphasizes that people have multiple, intersecting identities and modes of experiencing oppression. This perspective needs to be better incorporated into data science work.

  • Community-based approaches to data collection and environmental monitoring, like those used by Public Lab, can help empower local groups and support advocacy and regulatory actions.

  • Design justice principles emphasize starting projects from the perspectives of marginalized communities in order to build technologies and systems that respect human diversity and experience.

Here is a summary of the key points about the Science for Social Good program at the University of Chicago:

  • The Science for Social Good program uses scientific research and tools to help address social problems and improve lives. It brings together students and faculty from various fields like computer science, social sciences, biology and more.

  • The program focuses on translating academic knowledge into solutions for issues like education, health, justice and economic opportunity. It advocates for an interdisciplinary, collaborative approach between researchers, community partners, policymakers and others.

  • Major projects under the program include work on improving K-12 education, reducing gun violence, advancing criminal justice reform and addressing homelessness. The projects involve data analysis, technology development and policy research.

  • The goal is to apply scientific rigor while also partnering with communities to ensure research addresses real-world problems andempowers participants. They aim to produce socially useful knowledge through participatory processes.

  • The program highlights how universities can promote social good through strategic collaboration across disciplines and sectors. It presents a model of academic research tackling important societal challenges in an interdisciplinary, collaborative way.

Here is a summary of the key points from the sources provided:

  • Source 1 describes a Storify thread and tweet by Erin Simpson pointing out that data from GDELT on kidnappings in Nigeria really refers to news stories about kidnappings, not the actual number of kidnappings.

  • Source 2 provides information about the GDELT project and its mission to code and catalog world’s news media.

  • Source 3 and 4 discuss emerging fields of study around videogames/masculinity and queer feminist interventions into big data scales.

  • Source 5 explains what APIs are and how they allow programs to access and retrieve data from other entities over the internet.

  • Source 6 provides examples of unconventional open data sources like dog names in Zurich, UFO sightings, lost items on NYC subway, and abandoned shopping carts in Bristol.

  • The remaining sources provide additional context and examples regarding data dictionaries, responsible data practices, open data initiatives, zombie data, predictive modeling, algorithms and bias, sexual assault reporting data from the Clery Act, and underreporting of sexual assault to the Department of Education.

  • Institutions often face reduced fines for violating Clery Act requirements around reporting crimes like sexual assault on campus. However, high-profile cases like the Penn State sexual abuse scandal result in larger fines.

  • Coming forward to testify about sexual assault, like Christine Blasey Ford did regarding Brett Kavanaugh, involves relinquishing privacy and reliving trauma publicly.

  • Schools like Columbia University have been accused of mishandling LGBT rape cases.

  • Not speaking about violence reproduces a culture of violence and abuse, according to scholar Sara Ahmed.

  • The dichotomy between “raw” and “cooked” data is misleading, as data always requires context and interpretation.

  • Google Flu Trends showed the dangers of prioritizing prediction over context - it overestimated flu cases when its algorithms unexpectedly changed.

  • Data platforms like Twitter make data vulnerable by opaque sampling practices.

  • Reddit data used in multiple studies was found to be missing millions of comments and submissions, affecting study validity.

  • Analyzing language data can quantify decades of gender/ethnic stereotypes reproduced in society.

  • Context is crucial for analyzing social media conversations but hard for automation to grasp.

Here is a summary of the relevant points from the text:

  • The authors did not appear to explicitly consider nonbinary genders in their study of gender differences on GitHub. Mozilla undertook an experiment to conduct gender-blind code reviews in response to the study’s findings.

  • GitHub’s founder resigned after an investigation into allegations of sexist behavior and harassment. A former executive also sued GitHub claiming she was paid less and fired for complaining about gender and racial pay disparities.

  • The interactive Ship Map project requires complex cognitive work from volunteers to analyze historical records. Crowdsourcing relies on “free labor” from volunteers to perform digital piecework.

  • The “wages for housework” movement in the 1970s advocated for remunerating domestic labor, especially unpaid women’s work. This addressed both gender and racial divides, as domestic work was racially stratified. Intersectionality was a key consideration.

  • Various terms like “invisible labor” and “free labor” describe uncompensated, undervalued or unseen forms of work, often gendered, that crowdsourcing and other digital platforms may exploit. Standardization and transparency around data collection practices are important issues.

Here is a summary of the key points about programming diversity and inclusion from the sources referenced:

  • Studies have found gender gaps in participation on Wikipedia, with fewer women contributing as editors. Outreach programs like #VisibleWikiWomen aim to address this.

  • Unpaid labor like cooking, cleaning, childcare falls disproportionately on women around the world. This limits women’s ability to take on additional volunteer work like open source contributions.

  • Surveys of Mechanical Turk (MTurk) workers found they were more likely to be female, younger, lower-income, and reside in India or the United States compared to typical American Internet users. MTurk jobs pay very little.

  • Content moderation jobs reviewing graphic or disturbing online content can cause psychological stress. This work is typically outsourced and poorly paid.

  • Technology and AI systems are often developed and trained without consideration of bias, as the developers themselves may lack diversity. This can negatively impact marginalized groups.

  • The materials and manufacturing of technology often rely on exploitative labor practices, like mining of conflict minerals in dangerous conditions. E-waste is also often processed by vulnerable workers in unsafe recycling centers.

  • Gender, racial, and economic inequality persist both within technology companies and in their supply chains and outcomes. More diverse and inclusive participation in technology is important to address these issues.

Here is a summary of the key points from the suggested readings:

  • “Anatomizing Echo” analyzed Amazon’s voice assistant Echo and found it reflects and amplifies problematic power dynamics around labor, data collection, and control over resources. It raises concerns about how technologies like Echo are developed through invisible forms of human labor that are often unpaid or underpaid.

  • Scholars argue more recognition and empowerment is needed for different types of invisible labor like community management, digital curation, collaboration, care work, emotional labor, maintenance, and more. This includes efforts like bills of rights for collaborators and students.

  • Self-tracking data from tools like Fitbit that is often treated as objective can reflect invisible forms of care labor. Data is shaped by social and political forces, not just technological design.

  • The concept of “crip time” highlights how notions of time and productivity are often coded to ignore barriers faced by people with disabilities or care responsibilities.

  • Movements like the Google walkout show the potential for tech workers to organize and advocate for issues like gender equity and ethics in technology development. This builds on a longer history of labor organizing in the tech sector.

  • Alternatives like tech cooperatives aim to give workers more democratic control and address power imbalances, while declarations like the Toronto Principles advocate for protecting human rights in machine learning and AI.

  • The text discusses principles of design justice put forth by the Design Justice Network, which include centering marginalized communities, collective access, sustainablilty, spirituality, and cooperation over competition.

  • It discusses workshops held by the Design Justice Network to generate shared principles for design justice.

  • Sasha Costanza-Chock is mentioned as writing a book on design justice that will be published by MIT Press in 2020.

  • The Design Justice Network Principles document is cited, which provides more details on centering marginalized groups, collective benefit, and sustainability.

  • Abolishing big data is discussed, arguing it should be put in the hands of those who need it most and naming the “data-industrial complex”.

  • Design justice seeks to re-Imagine design processes to be more equitable, work to dismantle unjust systems, and place marginalized communities at the center of the design process.

Here are the key points from the references in the provided text:

  • Sharpe, Christina (183) - Discuses the intersection of identity categories like gender, geography, and disciplinarity. Argues identity is shaped by positionalities.

  • Özkaynak, Begüm (147) - Looks at issues of citizenship, geography, and identities through analyzing place names in Canada. Shows how places are sites that construct identities.

  • Shetterly, Margot Lee (3–7, 31) - Provides a biography of mathematician and physicist Margaret Murray who faced gender barriers. Details the invisible labor of women in science.

  • Pak, Agnes (173) - Examines issues of gender, labor, and inclusivity in open source software communities like GitHub. Finds biases in participation and recognition.

  • Snowden, Edward (12) - Known for leaking National Security Agency documents and raising awareness about mass surveillance. Revealed privacy and data issues.

  • Spivak, Gayatri (133, 147) - Influential postcolonial feminist theorist. Work discusses concepts like epistemic violence and subjugated knowledges from marginalized groups.

  • Philip, M. NourbeSe (183) - Poet who discusses fluidity of gender and sexuality identities and intersections with race. Part of Africadian and Black Canadian literary movements.

Here is a summary of the key sections from , 91– 96:

  • Image-processing algorithms are discussed on pages 11, 65, and 160 in relation to their use in fields like computer vision. The challenges of encoded bias and lack of diversity in training data are mentioned.

  • Google is referenced for its work on image-processing algorithms, as well as issues around data collection, transparency, and the diversity crisis in AI/tech. Google Brain, an AI research group, is mentioned on page 141.

  • Data sets and their proprietary control by companies like Google and Instagram is discussed on page 152 in relation to limiting access and analysis by outsiders.

  • The use of data and algorithms in domains like medicine is covered, highlighting issues of bias, lack of diversity, and historical injustices and oppressions like racism and sexism that remain issues today in fields like healthcare.

  • Journalism’s role in investigative computing using data as well as increasing transparency around data use is summarized.

That covers the main topics referenced for the page range 91-96 based on the summary you requested. Let me know if you need any clarification or have additional questions.

Author Photo

About Matheus Puppe