Skip to main content
SearchLoginLogin or Signup

Chapter 4 — Researching

Published onNov 02, 2022
Chapter 4 — Researching

It was 2018 and the group Mumalá in Argentina had been documenting feminicide for three years. Mumalá is the short name of the organization Mujeres de la Matria Latinoamericana (Women of the Latin American Motherland). The group has existed since 2001 when it was formed in a period of economic crisis and great social need in Argentina. Aligned with Argentina's piquetero movement, which advocates for the right to secure employment and livelihood, Mumalá characterizes themselves as federal, feminist, popular and dissident1. They work across urban, rural and remote parts of the country on issues of gender rights, including and especially those issues which intersect with neoliberal economic violence. They had started a femicide observatory in 2015, following the #NiUnaMenos uprising. At the time, the Argentine government had no central database of femicides (and still does not have one that activists consider comprehensive)2. Mumalá saw a need for independent monitoring which integrated a feminist analysis of the issue with specific political demands, and so they initiated their registry. Over three years they had amassed an extensive, centralized database of hundreds of cases, carefully researched and recorded by a handful of their members. But in early 2018, those same members walked away over political disagreements with the rest of the group – leaving the organization and taking the database with them.

Losing their data was a wake-up call for Mumalá. They used that painful moment to regroup and reformulate the way they researched cases and stored their data, as well as to expand their focus to monitor other forms of violence in addition to femicide. In late 2018, they re-launched the project and it is now called the "Mumalá Observatory: Women, Disidencias, Rights"3. Members research and record cases of femicide, femicide attempts, missing women, and LGBTQ+ hate crimes.

Their research structure is now "federated", meaning they have Mumalá affiliates based across Argentina's 24 provinces who monitor those specific locales for cases. This is better aligned with Mumala's organizational structure which is based on chapters that self-organize and coordinate local protests for gender rights around the country (see figure 4.1a). Thirty-two people in total produce data for the observatory. Like the majority of data activists we interviewed, their primary source of information is media reports, combined with information from families or friends of victims, or other members of the public, who reach out directly. Mumalá members – whom they refer to as compañeras and compañeres – scan news media reports, conduct web searches, monitor Google alerts, and scan social media in order to discover new cases in the province they are responsible for. Once they locate a case, they register it via a web form into their database and continue following the case as further information reaches the public. Following the data loss in 2018, there is now not a single database but rather many multiple copies of the database. Activists work with their own local copies and then synchronize cases periodically. After registering a case, they may also seek out official sources of information on it, and work with their network of feminist journalists who have more success than activists in speaking directly with prosecutors and judicial officials. Mumalá publishes monthly and yearly figures to their social media accounts, often accompanied by analysis and infographics like the one in figure 4.1b.

Four female-presenting activists wearing T-shirts with ‘Mumala’ printed on them, are raising their hands and spraying a purple colored gas as a demonstration of protest. More activists and buildings are visible behind them.
Figure 4.1 (a) Mumala has a national group as well as chapters based across Argentina. They stage many mobilizations around gender rights across the country such as this one which took place on the seventh anniversary of #NiUnaMenos in June 2022. Courtesy of Mumalá.
Figure 4.1(b) Mumalá's observatory registered 221 femicides in 2021, including 6 transfemicides/travesticides. The organization publishes monthly and annual figures to their social media accounts, often accompanied by analysis and infographics like this. Courtesy of Mumalá.

The subject of this chapter is the work of researching – how counterdata groups seek, find and verify cases of gender-related killing and related information, especially in the absence of official data. Researching is the second stage of a feminicide counterdata science project (see figure 2.4) and is at the heart of counterdata production work. As we will see, activists develop deep expertise in the information available in their contexts and engage in a vast array of creative informatic strategies to research missing data. These often involve collaborative and innovative forms of human relationships, as exemplified by Mumala's federated monitoring structure which engages dozens of members, based regionally, who contribute to the national observatory.

Researching follows after the first stage – resolving – in which activists determine to address structural violence through counting, and it precedes recording, in which groups register and classify cases with specific variables into structured data sets. But the divisions between these stages are not always so clean in practice. Groups' analysis of power is sharpened by encountering missing data from media and institutions during the research process. And researching and recording cases is often a back and forth process – as activists follow cases, sometimes over the course of months, years or even decades, they continually seek and find new information and they proceed to log that into more complete cases in their databases.

Compared to the groups we interviewed, Mumalá's research staff is quite large. Not all groups have dozens of volunteers dedicated to finding cases and covering particular territories. Indeed, many feminicide data projects are conceived and run primarily by individuals, yet they still develop deep expertise in the information ecosystem surrounding feminicide and use creative, collaborative strategies for researching cases in the face of missing data. Such is the case with Dawn Wilcox who runs Women Count USA from her home in Texas in the United States4. She has a full-time job as a school nurse, but she spends many hours on nights and weekends scanning news media articles and following up on tips sent by strangers. Her goal is to document every case of femicide in the United States going back to 1950. A survivor of domestic violence herself, Dawn became curious about femicide statistics in the US in 2016. She started what she thought was a simple online information search and was shocked when she realized how difficult it was to find systematized information on women's murders. She found that the FBI's Uniform Crime report was woefully incomplete, not only for women but for all homicides, since they rely on data voluntarily reported from law enforcement agencies. Other citizen-run archives, like the National Gun Violence Memorial, only register gun deaths. And domestic violence organizations tended to limit their scope to intimate partner relationships, which left out women killed by other family members, by neighbors, by strangers, in the context of sex work, and more. Dawn felt that the fragmentation of the information contributed to the invisibility of the problem: "I felt like if I could bring all of this data into one place that, first of all, it could tell a story about what was really happening to women and where was happening, how it is happening, who was doing it, who was killing women, what sort of relationships they had. And I felt like it would memorialize these women, which was very important. It would show that they were more than just statistics on a page."5

Dawn sources cases of femicide daily from digital news articles that she finds by using a search engine and typing a set list of queries such as "woman's body found", "woman stabbed to death", and "husband kills wife". Since Dawn started the project, her work has become more widely known and people will often email her news articles about femicide. Like many activists that we spoke with, she cross-references or "triangulates" multiple news media articles against each other as well as to other sources of data in order to verify information and arrive at the details she needs to log a case in her database. Reading a news article will reveal a victim's name that she will then use to search for more details. She is adamant that every woman in her database needs a photo, and this is the piece of information that takes the longest amount of time to find. When the photos that she finds are low quality, she will retouch them. The work is immense: "I think even if I did this work full time, I would still need help. It's just... the sheer number of cases is just staggering."6 Dawn is always working with a backlog of articles to review, cold cases to search up, and tips to follow up on.

While individual activists face challenges of time, resources and missing data in the researching stage, groups that monitor MMIWG2 (Missing and Murdered Indigenous Women, Girls and Two spirit people) and racialized feminicide face even more hurdles. News outlets cannot be their sole source of discovering new cases because they know that the media systematically neglect to report on this violence. Such is the case with the Sovereign Bodies Institute (SBI), which is based in Northern California and monitors MMIWG2 across the Americas and beyond. Annita Lucchesi, now the executive director, started the work as part of her masters thesis in geography, which used mapping to tell stories of intergenerational violence and genocide in Indigenous communities: "it was a way to make sense of what happened to me and kind of the broader maze of trauma in our communities."7 Official data about MMIWG2 are woefully inadequate, and Annita and SBI have quantified exactly how inadequate they are. In a scan of official records in California, SBI found that 91% of missing Native girls in California are also missing from at least one official database. Of those actually recorded in state databases, 56% of missing Native women and girls in California were classified incorrectly as a different race8. Part of the problem has to do with legal and "jurisdictional mazes": it is often unclear – even to officials themselves – which county, state, federal or tribal agency is responsible for investigating a crime (see figure 4.3)9 [CITE]. Yet trying to detect cases from media articles is also challenging because the media, first, don't report on MMIWG2 cases and, second, even when they do consider those cases newsworthy, according to Annita, "press don't really do a great job of acknowledging victim indigeneity."10

Instead of relying on official data or news media as a first line of information for learning about cases, then, Annita describes how Sovereign Bodies Institute researchers "get creative with the data."11 They use Indigenous networks, both digital and physical to discover and gather information about new cases. They review social media posts and do direct outreach. They have done FOIA requests, partnered with tribal enrollment offices and used historical archives. Sovereign Bodies Institute is a family-centered, survivor-led and survivor-centered organization. As Annita frames it, " thing that became clear starting to build a database and do this work is that a lot of families were going to come to us directly and that still happens constantly – families wanting to make sure their loved one is in the database and those families had a lot of acute needs." SBI pairs their informational work with direct services and support and tries to respond to family requests for help on specific cases. At the time of our interview in 2020, Annita was working directly with a medical examiner for the state of Montana to carefully review state autopsies for a handful of cases that families wanted re-opened, challenging the state ruling of "suicide" as a misclassification. Sovereign Bodies Institute sources of information end up needing to be "multi-pronged" and "diverse" to counter the additional research hurdles faced by groups monitoring violence against Indigenous women.

Researching cases

The three stories narrated above describe the diverse research strategies used by three groups – Mumalá, Women Count USA, and Sovereign Bodies Institute – to find information about feminicide in their contexts. Researching is arguably the most time consuming part of any counterdata science project. This is where an individual or group spends the bulk of their time and effort: seeking and finding examples of the phenomenon of interest. For the activists we interviewed, the unit of analysis is a case of feminicide or gender-related killing. The researching stage includes discovery of new cases as well as on-going research to follow, add to and verify information for other existing cases. During this stage of a counterdata project, activists continually assess existing sources of information, including official databases, other counterdata projects, news media, social media and more.

In their research process, all groups encounter many examples of missing data – data that are either incomplete, incorrect, biased, inaccessible, biased – or some combination of the above – of just wholly absent. Missing data can function as both a motivating factor for beginning a project as well as an on-going challenge to navigate as activists research cases. For Mumalá, it was the failure of the government to create a federal registry of feminicide in Argentina, even after the #NiUnaMenos movement had made that a central demand and the government had committed to doing it. For Dawn, it was the incomplete and underreported federal homicide database combined with civil society data projects focused on subsets of femicide, but not femicide itself. And in Sovereign Bodies Institute's case, they encounter incomplete and misclassified official databases, "jurisdictional mazes", and heightened media bias, making cases of MMIWG2 especially challenging to source.

In the researching process, all groups seek to use multiple sources of information about a given case of feminicide to verify details. Once they obtain the name of the victim or some identifying details, activists can search for more media articles about that case as well as seek out institutional data such as police reports, court records or death registries. Two groups described this process as "triangulation" – an apt name for the idea of using multiple sources of information to cross reference and verify details about the killed person, the perpetrator, the event and the judicial response12. As I described in Chapter 2, the activists working on the Persistence of indolence report also did this. They used cases of feminicide sourced from the Puerto Rican media and triangulated those with official numbers to demonstrate systematic gaps in police data13. Annita from Sovereign Bodies Institute describes their process: "we also kind of go on a deep dive to see if we can find other information. So for example, the media aren't going to include the victim’s birthday or how many kids they had, or what tribe they're from, but we might be able to find that in their obituary, so we try to do as much of that as we can."

All activists are deeply concerned with getting the facts and details of a case straight. They learn to be suspicious of early media reports about a feminicide, because these can often be filled with errors; they need to follow a case for some time until details are clarified and they can verify one source of information against others. They learn which outlets cover the issue, how it is reported, and which terms and language and framing are used. Nerea Novo, who researches cases for, says it like this: "...the methodology [for searching for cases] can be taught quickly but then practice gives you certain tricks to get to know which sources are more reliable, which sources less, which sources dedicate more time to research and which do not."14

The work of researching cases – reading and scanning stories of violence, sometimes for hours every day – takes an emotional and mental toll on counterdata producers. Betiana, from Mumalá, concedes, "Yes it's true…the work we do is quite pessimistic. The compañeras never cease to be surprised at the amount of violence in new cases and old."15 Researching feminicide involves the continual emotional labor of reading about brutal violence, and the secondary witnessing of the trauma and loss of others, and we will return to this at the end of this chapter. For survivors of gender-related violence, it is work that hits close to home.


Reasons for missing data mentioned by activists

Research tactics activists use overcome missing data

State - laws

  • Absence of laws on feminicide

  • Narrowness of laws on feminicide

  • No public information laws or public reporting requirements

  • Law doesn't recognize certain groups (trans people, Indigenous status of people)

  • "Jurisdictional mazes" and fragmentation of laws across county, state, federal, tribal territories

  • Include cases based on structural analysis of power

  • Count people that are excluded by present laws, including LGBTQ+, children, Indigenous people not recognized by state

  • Use non-state information sources to discover and track cases

State - implementation of laws

  • Law on feminicide is comprehensive but interpretation is narrow

  • Law requires state disclosure of info about feminicide but state fails to do it or finds legal loopholes to avoid giving public records

  • Lack of state resources to investigate cases and/or publish information

  • State employee turnover (police, prosecutor, judicial)

  • Lack of state expertise in gender, race, feminicide; minimization of gender-related violence as crime

  • Public information not published in timely or disaggregated way (gender, race, sexuality)

  • Public information shows signs of possible political manipulation

  • State is absent in rural/remote areas

  • State consistently misclassifies cases (e.g. suicide cases, cases related to Indigenous women, trans people)

  • Information collection and reporting is fragmented across state agencies

  • Manually review state websites, state social media feeds, state-run WhatsApp & Telegram groups

  • Search court records

  • Follow up on individual misclassification/suicide cases, especially where family and community are contesting state ruling.

  • Triangulate state records with media reports to see what's missing/conflicting

  • Visit morgues, coroners & medical examiners

  • Work in networks & coalitions to discover cases and share information

  • Solicit/receive crowdsourced reports of cases

  • Use federated media monitoring structure to cover large territories

  • Use media sources – TV, radio, print and digital – to discover cases

  • Source cases from other counterdata databases (e.g. Fatal Encounters in the U.S. with gender and/or race filter)

  • * Partner with tribes and cross-reference their records

  • * Mine state-run historical archives

  • * File formal public information requests, e.g. FOIA in the US

  • * Call friends who work for the state and discuss discrepancies


  • Media report sensational details but do not provide info on race, tribe, sexuality

  • Media report on some deaths and not others (e.g. often don't report on trans, Black, Indigenous, migrant, rural/remote, poor people)

  • Media reporting about feminicide is deeply biased, toxic and sensational

  • Media draw from police reports and framings, replicating state biases (like misgendering trans people) and often using victim-blaming frame

  • Media are absent in rural/remote areas

  • Media do not report on killings related to organized crime/paramilitary activity for fear of retribution & violence

  • Media do not follow cases through the justice system

  • Early media reports on killings are often full of incorrect information

  • Use stigmatizing language to search for cases in media e.g. "crime of passion", "man dressed as woman", but reject their framing

  • Seek cases from hyperlocal and regional news outlets and blogs

  • Use social media (esp. Facebook) & private groups (esp. WhatsApp) to find and validate cases outside mainstream media

  • Use networks & partnerships to confirm details/verify information

  • Collect humanizing information (e.g. photos, details about life)

  • Triangulate official state data with media to see what's missing

  • Follow individual cases through judicial system

  • * Use death certificates to confirm race

  • * Follow prominent journalists reporting on feminicide


  • Family/community doesn't report because too much work, already traumatized

  • Stigmatizing for families to report feminicide (e.g. elite families try to keep cases out of the press)

  • Minoritized communities have good reasons not to trust the authorities so they don't report

  • Partner & share info with groups that offer families acompañamiento (accompanying) through the justice system

  • Families/communities contact activists directly to request inclusion (or, less frequently, removal) of their loved one or to share more details

  • * Get individual case files from families, friends or state leaks

  • * Contact families directly to verify information or offer support/services

* = Strategy mentioned by only 1-2 groups. Not a common pattern.

Table 4.1 maps out activist analyses of missing data in their context as well as the creative strategies they used to research cases and acquire information. The entries in the table represent something mentioned by three or more counterdata groups. Entries with a * were mentioned only by 1-2 groups.

Power and resistance in the information ecosystem

In undertaking research, activists operate in an information ecosystem: a dynamic constellation of actors that includes infrastructure, tools, technology, producers, consumers, curators, and sharers of information about feminicide. The metaphor of the ecosystem is designed to capture the dynamic nature of information – it moves and flows across scales and sites and actors as it is produced, curated, transformed and used16. In the case of feminicide information, the sites and actors that surfaced most frequently in our interviews included the state, the media and families as key producers of information about cases (and, paradoxically, also key drivers of missing and biased information).

As we interviewed activists, they provided analysis and insider perspective on their information ecosystems. They have expert answers to questions like: Who produces information about feminicide cases? How reliable is it? How can it be verified? How public is it? If it is not public what are other ways to acquire it? When is information biased or unreliable? Why and how is information missing, biased or dubious? Indeed, we found that activists connect their analysis of power (discussed in the prior chapter on resolving) into concrete observations about how that power is made manifest in the information ecosystem in their country or locale, and then use creative strategies to address those informatic gaps and biases.

Table 4.1 places activist analyses of missing data in counterpoint to the strategies they use to overcome that missing data through research17. Activist analyses of power are grouped into the domains that surfaced most frequently: the state (including both laws and the implementation of laws), media and families. Each of these domains is important in the information ecosystem because each significantly affects the production of data and information about feminicide. These groupings also map roughly to the domains of oppression outlined in Patricia Hill Collins' matrix of domination introduced in Chapter 2. While that chapter provided a high-level analysis of different factors at play in producing missing data about feminicide in each domain, our interviews helped to elucidate activists' own analysis of factors that led to missing data in their locale.

Missing data: State laws

Many interviewees mentioned either narrowly framed legislation or the complete absence of legislation leading to the lack of state-produced information (see Table 4.1, first row). As Myrna stated, "In Canada, nothing is officially seen as a femicide because we don't have any legislation or any official recognition of femicide."18 When laws do exist, they may be narrowly formulated. A handful of laws only include intimate partner violence, and many exclude or make no provisions for transgender women, so cases that fall outside of these frameworks will not be recognized by the state as having a gender motivation nor be included in official counts, lists or statistics (see Table 1.1). Even when laws and official data do exist, activists mentioned that the absence of public records laws can inhibit the availability of information.

Activists counter these legislative hurdles by deliberately defining and counting femicide in a way that matches their own structural definition of the violence which exceeds the state's legal definitions. For example, Mumalá counts induced suicides – when a woman is driven to suicide by repeated domestic abuse – as femicides. This is a concept not currently outlined in Argentine law but it is recognized in El Salvador's feminicide law which Mumalá has drawn inspiration from. They also count deaths from clandestine abortions as femicides. They seek information on transgender killings and also on the premature deaths of trans people, which Mumalá attributes to gendered forms of social and economic violence which reduce trans life expectancy to half that of the cisgender population. By necessity, they must seek non-state sources of information to register such violence. But any such expansion is only done with careful consideration. For example, in the absence of adequate legal definitions for phenomena like transfeminicide, Mumalá relies on discussion, debate and consensus-building about individual cases through their WhatsApp group, what Betiana names as "the soul of the observatory."19 Through such collective deliberation, the group determines whether a case corresponds to their definition of feminicide, using their analysis of power, and therefore whether it should be counted.

Missing data: State implementation of laws

Likewise, in terms of the implementation of laws, groups mentioned a variety of factors leading to state oversight, bias and mismanagement of cases which affected downstream information systems (see Table 4.1, second row). For example, activists highlighted the lack of state resources to implement laws, investigate cases and/or to publish information. For ALDEA, among many others, they saw an acute absence of the state in rural areas of Ecuador, leading activists to speculate that cases in those areas are much higher than official counts. Some activists mentioned the frequent turnover in law enforcement officials, leading to the loss of institutional information about cases. As Juliana from Uma por Uma noted, "local police chiefs in the Pernambuco police change constantly which hinders the investigation process a lot. We saw the same case passed on to several local police chiefs and these local police chiefs did not have access to the amount of data that we had. So several times, for example, I passed the contact information from a family member to the chief, because he didn't have it. He didn't know that the guy had been released, that justice had released the suspect, the aggressor. So, this exchange, this inconsistency on the part of the police, was also something that we realized was a failure in the system."

Other activists surfaced questions of bias and training, noting how the public sector employees are not trained to recognize or investigate gender-related violence, leading to disregarding such factors in cases or to misclassify cases. State misclassification, especially of race (for Indigenous people), gender (for trans and gender-non-conforming people), and cause of death (for accidental deaths, induced suicides and police violence) was one of the most frequently mentioned drawbacks of official data. Even when the state may produce some kind of information on fatal gender violence, activists stated that there is often institutional fragmentation over which agency collects which cases, leading to multiple "counting" arms of the state which count different phenomena with different criteria. This is an observation corroborated by ILDA's work on their data standard in Latin America, which found that states lack both technical capacity and shared methodology to share feminicide data across divisions internally20 [CITE]. Finally, many activists noted that official information about feminicide is often not published in a timely and disaggregated way, and it occasionally shows signs of tampering, as was mentioned in the case of domestic violence data in Puerto Rico in Chapter 2.

Despite challenges acquiring official information, activists get resourceful to still mine what they can from diverse state sources. For example, many manually review state websites, attorney general pages, court records, state-based social media feeds or state-sponsored chat groups on WhatsApp or Telegram. These never yield "data" in the sense of systematized information in rows and columns, but activists can extract important details about individual cases for their databases. Activists based in diverse places – Guatemala, the US, and Canada – have resorted to visiting morgues and medical examiners to interview state employees and review specific cases together. For example, Sovereign Bodies Institute's aforementioned work with a medical examiner in Montana led to a case of accidental death being re-opened and re-classified as a murder. As Annita recounted the story to us, she described the impact of the collaboration: "it was very graphic and traumatic at first but it was also really empowering. And I was able to explain things to the family that I wouldn't have been able to explain otherwise."21

While it can be acquired creatively, information from the state may still be biased, misclassified, or unreliable, so activists have various strategies to triangulate it with non-state information. These include incorporating cases from other counterdata or citizen-led projects. For example, in the US, both Dawn and the African American Policy Forum periodically copy relevant cases from the Fatal Encounters database which documents police violence. Annita said she tried to do the same for Indigenous women, but Fatal Encounters had placed Native cases into the "other" racial category, illustrating how counterdata projects can replicate the same biases present in official data sources.

Groups develop novel and collaborative human networks of information-sharing in order to fill in gaps in official data. This is evidenced by Mumalá's federated monitoring structure where they are able to cover a vast geographic territory by having a handful of members responsible for discovering cases in each province. The Alianza in Ecuador works in a similar way where they have regionally-based groups as members of their coalition and those serve as a key source for discovering new cases as well as verifying information about cases in rural and remote areas. Many groups have developed relationships with feminist journalists and leverage those relationships to source information. And finally, once activists' work becomes well-known, like Dawn's, they begin to receive many crowdsourced tips about cases through email, messaging apps and social media.

Missing data: Media

By and large, the most prevalent source of non-state information for feminicide is the news media, as we see in the cases of Mumalá and Women Count USA whose observatories are built primarily around researching news articles (see Table 4.1, third row). Several times a week, members of Mumalá will type search queries such as "death of a woman Cordoba" into search engines to source recent news reports. They also plug those queries into Google Alerts so that they get notified when Google indexes new web pages that meet those criteria22. Because Mumala distributes monitoring across individual provinces, the geographic modifier - "Cordoba"- is meant to limit the search to reports from the province of Cordoba. But search engines don't always index the small regional news sites and local blogs that activists find most helpful in their research. So compañeras and compañeres from Mumalá find they still have to go to the website of each news outlet in that province and read through different sections where feminicides may be reported: police reports and society sections, mainly. Most groups do this digitally, going site-by-site and section-by-section, but a small number of groups still work in analog formats. Members of the Grupo Guatemalteco de Mujeres read and clip physical newspapers and Carmen Castelló (the retired social worker in Puerto Rico introduced in Chapter 2) watches the daily news on TV to discover new cases and then goes to the web, and in particular to hyperlocal news sites, to learn more about a case reported on television.

While all groups use news media as either a primary or secondary source for researching cases, all groups are also deeply critical of media reporting on feminicide and gender-related killing. It was variously called sensational, irresponsible, shameful, victim-blaming, misgendering, dehumanizing by design, stigmatizing, toxic, transphobic, lesbophobic, racist, xenophobic, demeaning, trauma porn, and poverty porn. Activists described how media often draw directly from police reports and quote law enforcement as an authoritative source, which leads to the transmission of bias – a kind of collusion between the state and the media23. In many cases, activists have to adopt the stigmatizing language used by the media into their search queries in order to retrieve the articles they need (i.e. googling "crime of passion"). However, activist groups generally reject the framing of the news articles, which tend to report on killings in ways that sensationalize the violence, depict it as an isolated incident and blame victims for their own deaths24. This was why Cuántas Más ran so many workshops, as discussed in Chapter 3, trying to improve journalists' understanding of gender-related violence. This is to say that media articles are useful for extracting specific fields needed for activist databases – such as victim age, method of death, or the relationship between perpetrator and victim – but they are toxic and harmful for the more important tasks: framing and analyzing the phenomenon. This is why groups' development and use of their own analysis of power, described in the prior chapter, is so important.

In response to harmful media narratives, activists reject dominant media framings of cases, use social media and work with organizations on the ground, and often directly with families, to verify important details about cases. Like Dawn, many activists seek out humanizing information such as photos or details about people's lives in order to give them some form of memory justice that the media has denied them. Where news media tend to over-represent the point of view of the state, activists strive to believe, support and follow the lead of victim's families and communities. For example, la Casa del Encuentro's observatory was founded in 2008 because the family of an Indigenous woman from a rural province in Argentina could not get the attention of either the judicial system or media and reached out to the feminist organization for support25.

Another built-in flaw in relying on media reports is that, in the words of Silvana, from the Brazilian observatory Neias, "the press are selective"26. All groups recognize this as an inherent limitation of trying using news articles to systematically detect cases of feminicide, but some groups experience it more acutely. Activists like Mumalá who monitor large geographic territories discussed how media don't cover those areas, or in some rural and remote places, simply don't exist. There are other regions where media do not report on any violence related to narcotrafficking due to fear of retaliation from organized crime networks. And there is the phenomenon of "Missing White Woman Syndrome", introduced in Chapter 2, which describes the disproportionate news coverage that white women receive when they go missing versus women of color in the US. Counterdata groups that monitor gendered violence at the intersection of race, indigeneity, and sexuality, such as MMIWG2, LGBTQ+ killings and Black women killed in police violence must find other methods to discover and document these cases because they are not considered newsworthy and not reported in the media. These groups develop extremely creative strategies to work around these limitations, such as the Sovereign Bodies Institute mining historical archives, doing FOIA requests, and developing partnerships with tribal enrollment offices. Gregory Bernstein, who worked at the African American Policy Forum on their database of Black women killed in police violence in the US, stated that such work requires "finding different, inventive ways of learning these stories." But at the same time, this work requires activists to acknowledge the incompleteness of their work; because the information is so challenging to find, Bernstein saus, "we know that there are so many that we are missing."27

Missing data: Families

While mentions of families and communities connected to missing data were less frequent, activists did note some situations which led to families not reporting cases and thus to resulting gaps in information (see Table 4.1, last row). First, while the pain of loss may lead some families to seek to make their cases public in order to secure justice, others handle their trauma by refusing to engage with either the state or media around the case. Numerous groups discussed how, as feminicide has grown in usage as a term, a stigma has been associated with it. This has led to wealthier families trying to suppress information about cases, particularly intimate partner feminicides where the perpetrator is another member of the family. And finally, in contexts of racialized state violence like the United States or widespread corruption such as El Salvador, families and communities have good reasons to not trust the authorities and may not report for fear of incurring retribution or further trauma.

Working with families can be an important source of information for activists to confirm details about a case, but activists are careful in how they engage with – and solicit information from – families. Some, like Sovereign Bodies Institute, do provide direct services and support to families, but Annita emphasized the importance of waiting for families to contact them directly: "we don't reach out to families directly until they're ready. That trauma is so severe that there are huge unintended consequences that can come from directly soliciting family." The majority of data activists do not work directly with families, but rather offer various forms of acompañamiento (accompanying) to families and family-centered advocacy groups28. This might mean providing space for their meetings, showing up for marches organized by family groups, providing data and information for family-led vigils, or other forms of support and solidarity. Through these relationships and partnerships, data activists and family groups often do end up sharing information with each other, sometimes even full case files, which can be important sources of information to fill in details about specific cases, especially when that information is not available from state or media sources.

Informatic resistance is contextual and relational

As a consequence of combining their power analysis with their researching practices, groups develop deep expertise in the flawed information ecosystems surrounding feminicide and gender-related killing. Incorporating an understanding of the ecosystem in which information is produced is part of what Lauren F. Klein and I meant when we outlined the data feminism principle consider context: "Data are not neutral or objective. They are the products of unequal social relations, and this context is essential for conducting accurate, ethical analysis." Feminicide data can never be collected, compiled, nor used at face value, because the contextual conditions of their production so deeply affect their quality.

Figure 4.3 MMIWG2 Jurisdiction Flowchart from the Sovereign Bodies Institute's MMIWG2 & MMIP Organizing Toolkit. SBI created the flowchart to help beginners and families understand the "jurisdictional mazes" produced by the current system, and which agencies should be taking responsibility for investigating which cases. Courtesy of Sovereign Bodies Institute.

Activist expertise includes not only creativity and skill in locating individual data points to enter into their spreadsheets, but deep familiarity with the political, historical, legal, cultural and geographic factors – the contextual factors – affecting the production, availability and reliability of information about the issue. We can see this evidenced by Sovereign Bodies Institute's MMIWG2 Jurisdiction Flowchart in figure 4.2 which they use to teach families and advocates about the legal geographies of a particular case. They published the flowchart in their organizing toolkit in 2020, and it walks the viewer through a series of questions in order to determine legal jurisdiction of a particular case in the US. Factors like whether the victim was taken over state lines, whether the crime happened on a highway, whether the crime happened on a reservation, and whether or not the perpetrator is a tribal citizen all shift which agency would be responsible for handling a case. Note that many of these details may not be known at the outset of a crime, thus the responsible agency can shift while a case is under investigation and information may be (often is) lost in the transfer. In-depth understanding of the legal landscape – as well as its shortcomings – has allowed Sovereign Bodies Institute to amass more complete and correct case data. In fact, their database has caught the attention of the federal government, who does not have comprehensive case information despite pressure from families and legal mandates to do so29. Recounts Annita, "the FBI keeps asking for it and we told them, we will not give you any data until you start adequately documenting indigeneity among victims of crime."30

Thus, while oppression may be organized to produce missing data and flawed information across the multiple domains of the state, media and families, creative informatic resistance is organized to respond in a multi-scalar way. Some informatic tactics of activists are designed to respond to failures of the state (e.g. counting with a more expansive definition of feminicide); and some to failures of the media (e.g. working in networks and partnerships to source unreported cases); and some to challenges of missing data in the familial domain (e.g. working directly with families to help them report violence and accompany them through the judicial process). This multiscalar resistance across domains of the information ecosystem demonstrates that counterdata groups have a deep understanding of the legal, administrative, narrative, and interpersonal factors that produce missing data in their context and have also crafted creative responses to research information in the midst of those challenges.

One important throughline across activist research tactics in all domains is their use not only of existing information sources but their creative production of coordinated, collective human relationships as a mechanism for case discovery, information sharing, and deliberative decision-making about cases. This resonates with data feminism's principle to embrace pluralism - to bring together multiple perspectives with priority given to local, collective and Indigenous ways of knowing. This is embodied by Mumalá's federated media monitoring structure, where different individuals are responsible for monitoring media in different Argentine provinces. Or in their collaborative method for deliberating, through WhatsApp, about which cases should be included in the database. It is also manifest in their relations with feminist journalists whom they can call on to try to get information from prosecutor offices. We can also see this evidenced in Sovereign Bodies Institute's work to build relationships with families to help them access services as well as with grassroots Indigenous groups across North America who then notify each other about new cases. There is counter-power in the development of these novel human-relational-informational configurations. And in contrast to hegemonic data science, these relations tend to be authentic, intimate and non-extractive31.

The emotional labor of researching violence

As I mentioned at the beginning of this chapter, a persistent theme across the researching phase of counterdata projects is the emotional labor required to research cases of violence as well as the mental and emotional toll of doing the work. Lara Andres, from Ahora que sí nos ven (Now that they see us), an observatory in Argentina says it like this: "The truth is that we can't spend all day, sitting down, reading about feminicides, because it does your head in. It does your head in, it makes you sick, so I think the most time I've spent registering feminicides, or reading news, would be two or three hours a day." Individuals talked about the psychological and emotional effects of this work on their well-being. Carmen found that watching the news to find cases early in the morning would leave her in a state of anxiety for the rest of the day, and Nerea discussed stories that she had read and registered that she couldn't stop thinking about. Some people find that it is not only about the graphic details of some cases, but also that the counting and aggregating becomes too much to bear. Paola Maldonado from the Alianza para el Monitoreo y Mapeo de los Feminicidios in Ecuador, reflected "...this work of marking how many women have been victims of femicide and placing those on a map is painful, it's terrible." And for Audrey Mugeni, seeing the aggregated numbers is tremendously difficult, "...when you go to the Excel sheet you’re like, oh no, I need to close my eyes. So it's so much harder to look at the Excel sheet than it is to just read the stories." Groups find various ways to navigate this emotional labor: taking breaks, using psychological support provided by the organization, doing group workshops on pleasure and joy32. Yet they also assert that the emotional labor involved in the research is itself an important form of care and witnessing – it is not to be considered an unfortunate byproduct of the work, it is itself a form of memory justice work. (I will return to this in Chapter 7 where we will see how our team, co-designing with activists, decided not to fully automate feminicide detection for precisely this reason.)

Given the range of emotional and affective responses that activists experience as they research and record each case of feminicide, and given the toll it can take on their well-being, it is curious that the output of this labor – typically the neatly arranged spreadsheet or database – barely reflects this turbulence. In April 2022, I had a conversation with Helena, my collaborator on the Data Against Feminicide project, about how challenging it is to "make labor visible" in relation to the research work that goes into feminicide data production. This is one of the principles of data feminism, drawing from feminist thinkers who describe how women's labor is frequently invisibilized and devalued. An antidote, then, should be to show and credit that labor. Yet in the case of feminicide research, this logic doesn't quite hold up.

When you visit the publicly available spreadsheet that Helena produces for the Feminicidio Uruguay project, it is straightforward. Almost painfully simple to read. Each row corresponds to a woman, to her violent death, and to related details surrounding the event: name, age, date, place, relation of the perpetrator to the woman. There is a matter-of-fact narrative in Column E which describes how her body was found, method of death, or whether she had been missing. Yet each row in this spreadsheet is the result of many hours of reading and research, triangulating details from multiple media articles, updating as new information surfaces about an investigation or as a case proceeds through the justice system. Furthermore, because Helena produces a map based on her database, she often spends still more hours using photos, Google Street View, and media articles to find the most precise place to geolocate a particular case33.

Figure 4.4 The open spreadsheet for Feminicidio Uruguay. Courtesy of Helena Suárez Val/Feminicidio Uruguay.

What is not visible in the spreadsheet is how Helena feels when she walks by one of the places of violence that she has geolocated on her map. One of them is about half a block from her apartment in Montevideo; she remembers the blood on the sidewalk every time she walks by. What is also not visible in the spreadsheet is her state of agitation when a case is initially reported in the press until the time that a perpetrator has been identified: "I would tell you that the day the alert arrives, I am involved in some way that whole day and probably a couple of days more to come…You kind of stay a little tense until there's some sort of resolution."34 Still, for some cases, locating the perpetrator may take much longer or may never happen. What is also not visible in the spreadsheet are the many hours of reading stories of brutal violence; the careful collection of each painstaking detail of what is known about a case; the way it feels to leave so many fields blank. Also not visible are the many stories needing to be read and deliberated that are not feminicide and the continual shock and horror that activists experience: "It's just never ceases to surprise and horrify me no matter how many stories you hear," states Julia Sharpe-Levine from the African American Policy Forum, speaking about their research on Black women killed by police violence in the US 35.

In both her capacity as an activist and as a scholar, Helena has been thinking deeply about data production work as care work. We wrote a short essay together about this following Maria Puig de la Bellacasa's formulation of care work as an ethical-political commitment to "neglected things". In the case of our project, those neglected things are feminicide both as a neglected political issue and also the neglected lives and memories of the individuals who are stolen by this violence. For activists, painstakingly researching cases is one small way to recuperate and repair such public neglect. Matters of care, for Puig de la Bellacasa, involve not only re-valuing the often invisibilized (and gendered) labor of maintenance and repair, but also disrupting the Western academic tendency to value distance, neutrality, objectivity and impartiality between researcher and research subject. Instead, care work weaves researcher and researched together into affective relationships, intimate relationships that produce new knowledge through proximity, emotion and connection36. This resonates with the data feminism principle to elevate emotion and embodiment. Feminicide counterdata activists are deeply intimate with their data points – the people and places whom the data represent – because they have spent so much time researching and triangulating sources, rejecting stigmatizing media narratives, finding photos, and correcting the errors of the state and the media. As a result, many activists are deeply protective of their data, as Annita evidenced in her resistance to sharing it with federal authorities in the U.S. For activists, paradoxically, these data are not data, never were data, and cannot be reduced to data. Yet their representation as data persists in the form of rows in databases or grids and columns of spreadsheets. The gridded orderliness hides the labor of its production. The spreadsheet remains deeply limited as a means of making the care work of feminicide data research visible – it is clean and cool, factual and to the point.

But is this a feature or a bug? Do activists want their spreadsheets to be messy records of emotion, to be testimonies to their rage, or their attachment or their sadness? Of aggregated encounters with narratives of violence? I will have to assume not since nobody that we interviewed documents feminicide in such a way. Rather, the erasure of the material and emotional labor behind feminicide data production feels more strategic and calculated. It functions as a kind of hack to hegemonic data production, where hack means the "clever or playful appropriation of existing technologies or infrastructures or bending the logic of a particular system beyond its intended purposes or restrictions to serve one’s personal, communal or activism goals."37 Presenting the results of a fraught and intimate and labor-intensive research process in the cool logic of the spreadsheet grid appropriates Western, colonial, patriarchal penchants for distance and quantification and deploys them to elevate and amplify the feminist concerns and the feminist rage seething just under the surface of the lined, light grey boxes38 . But it elevates and amplifies rage precisely by visually obscuring it. This troubles the data feminism principles make labor visible and elevate emotion and embodiment. It may not be desirable, in this case, to make labor visible but rather to hide that labor behind the strategic deployment of the rhetoric of objectivity. Likewise, it may not always be politically productive to elevate emotion when such emotion becomes the basis for discrimination – the familiar patriarchal approach of dismissing emotion as counter to reason or painting feminist concerns as too "personal". Moreover, some parties do not deserve our grief and our outrage, which are constantly leveraged in exploitative and voyeuristic ways by the media, leading to more than one activist characterizing media coverage of feminicide as "trauma porn". This leads me to ask the question – in the process of doing data science, when should one seek to make labor visible? In which contexts do we elevate emotion? And for whom?

Counterdata science vs. hegemonic data science

The information discovery tactics described in this chapter surface some key differences between counterdata science projects about feminicide and hegemonic data science projects in the researching stage. First, the labor of hegemonic data science is masculinized, overvalued, and overcompensated. The research group AI Now reported in 2018 that women comprise only 15 percent of AI research staff at Facebook and 10 percent at Google39. More broadly speaking, men dominate computer science occupations in the US, comprising almost three-quarters of the workforce, and gender balance in computing has been declining since the mid-1980s40. The median salary of a data scientist in the US in 2020 was $164,50041. In contrast, the work to research and monitor feminicide is almost the polar opposite. As we have seen, it is done almost exclusively by women. Counterdata work that centers on racialized feminicide, such as MMIWG2, is almost exclusively undertaken by women of color. Most research is mostly conducted in a volunteer capacity, meaning it is unwaged work. When projects originate from a nonprofit or from journalism, people may be paid for their time, but often these wages are precarious. Many nonprofits can secure one-time grants to support the work but have a hard time securing operational funds to sustain it, and grants often come with onerous reporting requirements that tax the capacity of small organizations. Thus, the labor of feminicide counterdata research is feminized, racialized, devalued and undercompensated. This is consistent with the feminist concept of reproductive labor – the care work that sustains, maintains and reproduces society42. In this case, instead of cleaning homes and raising children, activists are stewarding the memories of killed women, caring for wounded communities and working towards repair and justice. It is women who are doing the reproductive labor to clean up, informatically speaking, after the structural excesses and negligence that led to the gender-based violence in the first place.

But this stark divergence is not only related to feminicide data research. Not all parts of the conventional data science pipeline are masculinized and highly compensated. Melanie Feinberg, an information scientist, describes how data collection and classification work is perceived as "unskilled and mechanical." When she runs data collection assignments in her classes, the students imagine (erroneously!) that "there is nothing to be learned from the process of generating data, because data collection is the mere recording of objects speaking for themselves."43 Indeed, data collection, annotation, sorting and labeling is often outsourced from the Global North to the Global South where it is done for low wages, disproportionately by women and racialized people44. This goes even and especially for information tasks that resemble the work of feminicide data research in their emotional burden: content moderation on commercial platforms such as Facebook, YouTube, and the like. In these jobs, low-paid workers spend hours per day sifting through graphic, violent, racist, misogynist, exploitative content and labeling it according to platform policies45. In contrast, the high-status part of data science supposedly comes after the data production: collection, classification, acquisition and pre-processing. This led Nithya Sambasivan and colleagues to write a paper titled "Everyone wants to do the model work, not the data work" outlining the downstream harms of disinvesting in data quality46. In a similar vein, Michael Muller and Angelika Strohmeyer describe the many types of "forgetting" that happen along the data production pipeline, in part because of this stratification of data labor47.

Feminicide counterdata activists provide a compelling model for what it looks like to refuse this (gendered, racialized, colonial) stratification of data labor. For better or for worse, they are affectively connected to and intimate with their data due to the hours of time invested in researching each case. This labor engenders a deep expertise in the information ecosystem surrounding feminicide, as evidenced by activists' analysis of sources of missing data in Table 4.1. Not only do activists understand the flaws and limitations of their informatic context, they understand how their own data also inherit those limitations. For example, all groups mentioned that they know their databases are not complete since not all cases are reported in the media or publicized on social media. They have to thus be careful in regards to how they communicate about them since any statistics remain undercounts, and the most intersectionally marginalized populations are the ones who remain most undercounted.

As Annita from Sovereign Bodies Institute reported, "the data requires a really intimate relationship in order to make it workable." She scoffed at some of the data requests that she has received from outsiders: "the requests that we get are based on assumptions that the data is just kind of like this divining tool that anybody can just jump in and use and that somehow all of the mysteries of this crisis will be solved from their armchair with, like, casual exploration. And that's not the case. If it were, we would have fixed it already." The ignorance that Annita is challenging here is a direct product of hegemonic data science's undervaluing of data research and collection work – leading data scientists to believe that they could just explore "the data" (which data? Well, whichever are available and cheap), treat them as ground truth, make a model, and then VOILÀ! hidden wizardly insights about MMIWG2 are revealed48.

Finally, as I previously mentioned, many of the most creative tactics that counterdata science groups develop to source information about feminicide come not through novel sensors or computational techniques but via the production of novel and collective forms of human relationships. These relationships tend to be non-extractive and authentic – they are not about acquiring data and then generating capital (financial or social) from it. This represents a contrast with hegemonic data science which, to the extent that it develops novel forms of human relationships, tend to be in the model of what Paola Ricaurte calls data extractivism which operationalizes everything as a potential data source49. Driven by profit, modeled on colonial relations, hegemonic data science treats humans as interchangeable units of supply and demand (e.g. Uber drivers and their customers), as mechanical automatons (e.g. Amazon Mechanical Turk & microwork platforms), or as chauffeurs for sensors (e.g. Waze, Google Maps). This is not to say that feminicide groups' coalitions and human relationships are free of conflict (they are definitely not), but rather that the commitment to non-extractive human relations represents a form of epistemic disobedience, a form of resistance to the data extractivist regime.

This epistemic challenge is resonant with non-Western calls for emphasizing relationality and collective responsibility in data and artificial intelligence. For example, in his project advancing a decolonial approach to AI, Sabelo Mhlambi advocates for the use of Ubuntu ethics in AI, which reject the idea of an isolated, rational individual and instead focus on how a person comes into full personhood only by "fulfilling one’s social duties and responsibilities to others ."50 In the CARE principles developed out of scholarship on Indigenous Data Sovereignty, the R stands for Responsibility, by which the authors mean 1) Responsibility towards others 2) Responsibility to expand capacity and capability and 3) Responsibility to center Indigenous languages and worldviews51. These approaches represent fundamental epistemic challenges to those Western data regimes that center information capture, extraction and hoarding for the purposes of corporate and individual wealth accumulation.

Ultimately, in their research process, feminicide data activists choose to place care, context, and connection at the center of their data gathering process. This presents a profound challenge to the data acquisition process of hegemonic data science. As Feinberg states, "The real revolution in data labor will be in acknowledging that data collection should be celebrated for its skill and creativity."52 It's not only about re-valuing data collection and production because it's the right thing to do, but also because it challenges harmful data extractivist regimes and it results in more rigorous and precise data science. For example, feminicide data activists are profoundly aware of their data's limitations and quality issues whereas hegemonic data scientists are often stunningly ignorant or surprised by theirs53. This is especially true when the information ecosystem surrounding an issue is highly influenced by structural inequality, i.e. when there is rampant missing data, biased data, harmful and stereotypical information, or mis- and disinformation, all of which need to be assessed and evaluated for inclusion in a broader data set. Indeed, in all facets of researching – from data generation to data collection to data verification – hegemonic data scientists have much to learn from grassroots feminist data activists.


Researching is the second stage of a feminicide counterdata science project in which an individual or group seeks information about individual cases to add to their database. This can include sourcing existing datasets, mining state and media sources of information, verifying details and triangulating across sources. Such research either discovers new cases or adds information to existing cases in the database. In the researching stage, activists conjoin their analysis of power with their information seeking practices, leading them towards a skillful, on-the-ground understanding of the sources of missing data and biased information that permeate the feminicide information ecosystem. Consequently, they navigate informatic gaps, biases, and errors by employing a variety of creative strategies to source information from the state, the media and families. Activists often develop novel human-relational-informational configurations – using relationships of trust and solidarity to establish networks of information providers. Still, the work demands extensive emotional labor to sift through stories of violence and precisely document details of a human life and death.

Researching feminicide cases goes hand-in-hand with recording such details into spreadsheets and databases, the subject of the next chapter. Thus, while these stages are described separately, in practice they are tightly linked, with activists going back and forth between seeking information about a case and copying new details into structured fields and categories. The work is not only challenging because of the violence but also because there is no end in sight. Activists seek a world in which gender violence has been eliminated but women continue to be killed and cases continue to surface. In the face of state injustice and media stigmatization, researching cases is an informatic strategy to challenge such structural bias with care and rigor.

Mariel Garcia-Montes:

The whole book in general, but this chapter in particular pays a beautiful homage to the movement of feminicide data activists around the world. I am moved reading the way you document, explain, highlight the invaluable epistemic and political labor these counterdata scientists have undertaken.