Skip to main content
SearchLoginLogin or Signup

Chapter 5 – Recording

Published onNov 02, 2022
Chapter 5 – Recording

In 2017, a map and a manifesto went viral in Ecuador. Both were published by the Alianza para el Monitoreo y Mapeo de Feminicidios en Ecuador – which translates to the Alliance for Monitoring and Mapping Feminicides in Ecuador – a coalition of feminist groups from around the country 1. The map depicts feminicides by province in Ecuador with both raw counts and shaded colors over a period of seven months (figure 5.1a). The accompanying manifesto spelled out the Alianza's demands which included, first and foremost, that the state recognize that the violent death of a woman every 50 hours is a public emergency and prioritize the issue in public action and policy. Other demands included the legalization of abortion, a substantial shift in mining and extraction laws that had long put Indigenous and poor women at risk of violence, and the immediate creation of a national registry of information on violence against women and LGBTQ+ people. The manifesto even spelled out detailed requirements for such a national registry. The database, they insisted, must:

(1) specify gender, ethnicity, age and other demographic characteristics about both the victim and the perpetrator

(2) include geographic information so spatial patterns can be understood

(3) show patterns of the various forms of violence and

(4) be open and easily accessible to the general public2.

Infographic by the Alianza depicting data  in 2017. At the top right it highlights that 103 women have been victims of feminicide in Ecuador in 214 days. To the left is a choropleth map of the country in shades of pink and white. The highest category of feminicides has 20-21 cases and the lowest category has 1-2 cases. The province where Quito is located has 20 cases. On the bottom right there is a bar graph indicating the same. And some key statistical numbers on the top right. Hashtags #VIVANOSQUEREMOS, #NiUnaMenos and #FeminicidioDeUstado appeared on the bottom in the center.
Figure 5.1 The first map published by the Alianza – Alliance for Monitoring and Mapping Feminicides in Ecuador – in 2017. Courtesy of the Alianza para el Monitoreo y Mapeo de los Feminicidios en Ecuador.

This chapter is about exactly these details that the Alianza was demanding. It is about the process of recording feminicide and why the variables, categories and classifications used to document cases of feminicide matter, how activist data schemas are created, how they vary across counterdata groups, and how they are used and shared and harmonized across groups. A couple notes on the terminology used throughout this chapter: All groups have a data schema, which refers to the set of fields or variables that they collect about each case of feminicide, such as name, age, date, and so on. If you are thinking in terms of spreadsheets, the data schema consists of the names of all of the columns in the spreadsheet. One row of data corresponds to one case of feminicide. Activists enter data values or points into particular rows and particular columns. For example, if a victim were 38 years old then that would be the value for the age field in a particular row. Activists also have categories for cases – such as linked feminicide or transfeminicide – which they apply once a row of information is reasonably complete and they are able to classify a case with one or more categories. Such categories might be drawn from international standards, like the UN's model protocol on feminicide, or drawn from activists' own analysis, or a mix of both.

Recording, then, encompasses the transformation of unstructured data that activists find through research into structured data: cases organized into rows, values organized by columns, and lists of categories and classifications to apply to cases for analysis. Here there is a continuous back-and-forth relationship with the prior stage – researching – and activists are constantly moving between researching cases, verifying information and recording values.

The Alianza's 2017 map seemed to instantly be on everyone's mind. "It shocked the country," reports Paola Maldonado, president of ALDEA Foundation, one of the organizations in the Alianza. "There was no such map and it was the first time something like this had been released." In just a few days, she said, they had more than 30,000 visits to their website and numerous articles in the popular press. Behind the scenes, the spreadsheet that produced the map was simple - it consisted of just four columns: the date of the feminicide, the age of the victim, the name of the victim and the province where it happened. The Alianza sourced these cases primarily from the media, but they also drew from official data from INEC (Ecuador's national statistics agency) and from women's and feminist organizations that had been keeping handwritten ledgers of feminicide in certain provinces since the early 2010s.

In fact, it was the possibility of amplifying the existing monitoring labor of women's and feminist groups that had inspired the creation of the Alianza. Paola felt that this was work that by necessity had to be done in community. As they started the work, she said, "I realized that we had to generate a system and that for that we needed an alliance to be able to validate cases and not myself be the only one responsible for them. It seemed like a huge responsibility to me to issue a single figure, as if I were the only one in charge."3 The Alianza now includes seven organizations, including a national network of women's shelters, lawyers working at Ecuador's Commission on Human Rights, as well as regional women's organizations4. The group discusses, validates and classifies cases through multiple WhatsApp groups – a pattern of pluralistic deliberation and harmonization that we saw across groups. As a result of this systematizing, their data schema has expanded from four variables to sixty, including nationality, gender and ethnic identity, sexual assault, pregnancy status, whether the woman had been missing, information about the perpetrator, including whether he was a member of the military or police force, and many more.

The Alianza's more recent maps reflect the addition of these newer analytical variables. Like the map from 2017, the 2021 map of feminicides (figure 5.2) also plots spatial patterns, showing raw counts of feminicides per province in Ecuador along with using shaded colors. But this map is more complex - it also depicts transfeminicide separately, also by province, using proportional circles. The infographics surrounding the map show a breakdown of cases by age range, by month, and by type of weapon used. And various parts of the graphic connect feminicide cases to patterns of sexual abuse, disappearances, suicides, prior registered complaints, and organized crime. While the purpose may seem purely analytic – connecting these factors to the incidence of feminicide – the bottom left of the graphic shows the group's larger demands for MEMORY, JUSTICE AND REPARATION5. In fact, this first demand – memory – surfaced as a central theme for many data activists in the recording stage, connecting the work of recording feminicide to the memorialization and collective memory of lives lost to structural violence.

Infographic by the Alianza depicting data  in 2021. One of the highlighted statistics says that there have been 197 violent deaths of women for gender-based reasons, along with 8 transfeminicides and 67 feminicides perpetrated by organized crime. There is a choropleth map of the country in shades of pink and white in the center. Key figures appear on the top and the right.  On the bottom right there is a line chart and on the left there is a pie chart about weapon used in the killing, with guns comprising the largest category . Below the pie chart there is text ‘Memoria Justicia y Reparacion #VivasNosQueremos #Cade44horas #BastadeFemicidio’
Figure 5.2 Feminicides in Ecuador 2021. Courtesy of the Alianza para el Monitoreo y Mapeo de los Feminicidios en Ecuador.

Like all activists we spoke with, the Alianza's data schema is dynamic. It has evolved along with their analysis of power; along with their understanding of the problem of feminicide in Ecuador and globally; along with their participation in various networks and communities that discuss feminicide and data; and along with their goals for using data to challenge and prevent feminicide. For example, they added various intersectional categories of identity to their database as they came to understand how violence against certain subgroups was under-registered. Because of the misgendering of trans women in the press, says Paola, "these cases are almost completely invisibilized."6 The way fields and categories come into the data schema is through dialogue across the Alianza's membership where groups report back about patterns they are seeing. Observing these patterns led to the addition of method of death and a particular emphasis on firearms, which are illegal in Ecuador. Recounts Paola, "A high number of women were victims of femicide with firearms. So, the purpose that the registry of femicides has to serve is to be able to question and to be able to fight with public policy and say, if nobody should have guns, why are so many women victims of homicide with guns? Where are they? And what is the state doing about it?"7

After 2017, what the Alianza considers their "year of trial and error", they decided that it was important to align their data schema with national and international definitions and standards. This meant, for example, adding to their data schema some of the categories of femicide outlined by the UN Latin American Model Protocol in the hopes that their data could be used for comparisons both regionally and globally8. It also meant taking into account Ecuador's existing legal frameworks such as the inclusion of femicide in the nation's penal code in 2014 and the more recent 2018 Law to Prevent and Eradicate Violence Against Women9. Yet while the group seeks to align its data schema and categories with these standards, they deliberately exceed them in some cases. For example, Ecuador's law does not recognize femicides of trans women as femicides, while the Alianza does. The state has a separate law covering killings of children, while the Alianza counts all ages. The state does not recognize induced suicides as femicides, while the Alianza does. And if a perpetrator commits suicide, the case disappears from the judicial system and can never be named as a femicide, whereas the Alianza still considers it so. These divergences are intentional and strategic - designed to push the state to expand its conception of the problem. As the Alianza's visibility has grown, they have been invited into regular dialogue with the state agencies who are monitoring femicide. Reports Geraldina, "the idea is that, in these meetings, we invite them to see it [femicide] in a new way so that they will expand their criteria."10

Activists in Canada are familiar with the need to develop their own counts and classifications of femicide. There is no law defining or typifying femicide or feminicide in Canada, yet there have been activist efforts to register cases since at least the 1990s. In this decade, the Women's Memorial March started the annual practice of reading out names of missing and murdered Indigenous women in Vancouver's Downtown Eastside neighborhood (see Chapter 1), and the grassroots effort Woman Killing, Intimate Femicide in Ontario started their registry11. As a young volunteer and graduate student, Myrna Dawson worked on the latter project with sociologist Rosemary Gartner and a group of workers at a local women's shelter, who came together around their dismay and grief at the murders of women in their shelters. While the effort eventually dissipated, Myrna continued producing data about femicide in her academic research. In 2017, the UN special rapporteur on violence against women and girls called for femicide watches in every country following the global uprising sparked by #NiUnaMenos. At that point, Myrna took action: "I looked at the data that we had and realized if we were going to do this work in Canada that this was the moment in time."12

The Canadian Femicide Observatory for Justice and Accountability (CFOJA) is based at the University of Guelph and seeks to be a reliable source of primary information about justice and accountability for femicide victims in Canada. Like Paola, Myrna realized that she could not do the work alone as an academic: "I really wanted to have the voices of the people who were working in the sector – of people who had experiences with domestic violence and intimate partner violence." She assembled an expert leadership panel of over forty people, which seeks to represent the geographic and gender diversity of Canada, the perspectives of racialized women and especially Indigenous women, disabled women, and survivors. This form of pluralistic governance has created some major debates and disagreements but the observatory has managed to weather them to date.

Infographic from the 2018 report on femicides in Canada. It highlights in a headline that 148 women and girls were killed in Canada in 2018. Below this headline, on the top left there is a choropleth map of geographic concentrations of femicide, with the northern province of Nunavut having the highest rate of feminicide. On the bottom left there is a table indicating the age distribution of the victims, with the majority (40%) of women being between 18-34 years old, whereas the majority (32%) across the general population is 55 years old and over. Below this is a list of gender-based motives/indicators which includes ‘misogyny’, ‘sexual violence’, ‘coercive control’, ‘separation estrangement’ and ‘overkill'. On the top right there is an overview of key statistics. Under this there is a pie-chart indicating the race and ethnicity  of victims indicating that 36% of victims are Indigenous. And further below on the right is a bar chart indicating the relationship with male accused of perpetrating the crime. At the bottom on the right there is a set of percentages indicating relationship types of intimate femicide, such as 'spouse', 'dating', etc.
Figure 5.3 Infographic from the 2018 report on femicides in Canada produced by the Canadian Femicide Observatory for Justice and Accountability (CFOJA). Courtesy of the CFOJA.

CFOJA is an interesting outlier in terms of the sheer number of fields and categories they use to describe femicide in their database. While many counterdata groups, like the Alianza, started their monitoring with a handful of fields, CFOJA launched their effort with a data schema of almost two hundred variables (perhaps reflecting the project's academic origins). These include detailed information about the victim, the perpetrator, and their relationship, as well as many variables that attempt to capture the case's path through the justice system: charges laid, type of trial, type of plea, length of trial, defense's argument, type of sentence, length of sentence, date of deposition, name of judge presiding over case, and so on. This reflects CFOJA's mission to not only focus on the violent murders of women but also on monitoring how and whether justice was served by the state. One of Myrna's long term research interests, for example, is in studying "the intimacy discount" – the fact that perpetrators of intimate femicide nearly always receive lighter sentences13.

Despite the fact that Canada has no national legislation on femicide, the CFOJA still draws from international standards to determine both the variables that they track as well as the types and categories of femicide they use to classify cases in their database. Like the Alianza, they have used the UN's model protocol to identify gender- and sex-based motives in cases as well as to add variables recommended by the protocol – such as disability and sexual orientation – to their data schema. And like many counterdata groups, there are cases where the CFOJA will intentionally and directly contradict the state's ruling on a case. Myrna says this happens most often in cases related to Indigenous women where the state does not classify a death as a homicide but the family and community insist that it is. In these cases, the CFOJA expert panel discusses and sources information from whichever members are closest to the geographic and cultural communities in the case. In most of these deliberations, the CFOJA tends to side with families and communities. But this is precisely the point – to count and include those cases that the state is systematically ignoring and misclassifying. Says Myrna, "You can get really cynical realizing that our systems are just not working. But that also is a motivation to keep producing research that can help inform evidence-based decision making and help make sure that we have those measures in place so that no more women and girls are killed."14

Recording cases

Graphic about the ‘Recording’ stage in the feminicide counter data collection process. The panel  has a red background that fades into beige at the bottom. In the middle there is a globe facing the Americas with various feminicide data collection initiatives mapped on it. In the background there is a screenshot of the spreadsheet by Women Count USA with feminicide categories such as ‘ethnicity’, ‘relationship status’ and ‘case details’. On the top left there is text: 
‘Recording
Information extraction + classification
Activists transform unstructured datasets located in databases, spreadsheets and/or text documents. They classify cases according to diverse typologies. They manage data, including ethics, access and governance of the database.’
Figure 5.4 Recording is the third stage of a feminicide counterdata science project. Courtesy of the author. Graphic design by Melissa Q. Teng.

Recording is the stage of a counterdata science project which encompasses extracting and registering information, classifying cases, and managing data (which in turn includes issues of ethics, governance and access to the database). For feminicide counterdata projects the unit of analysis is a case of feminicide or gender-related killing and so one row in a dataset corresponds to information about one case of feminicide. In the process of recording cases, activists transform unstructured information into structured data, arranged into specific pairs of fields and values, and they apply categories to cases in order to detect patterns of violence.

With regard to the logistics of registering information, a large majority of the counterdata projects we interviewed use spreadsheet software (mainly Google Sheets and Excel) to record data on individual cases. The Alianza, for example, started by recording cases using Google Sheets and Excel but later moved to Kobo Toolbox, which is open source and has more robust data security measures. Several projects, including CFOJA, additionally keep individual case files in word processing documents where they can store free form notes and the full text of media articles about the case. CFOJA also uses SPSS, a statistical analysis software program, for recording variables and performing statistical analysis. Five of our interviewees use database management systems (DBMS) despite the added cost and technical knowledge required because they wanted to include multimedia files related to cases (like images) or build more complex relational structures and queries with their data.

In order to record cases in their databases, activists tend to follow a manual process of copy-pasting individual data points (such as victim's name or age) from news articles or other sources into the spreadsheet or database program. Some groups use a web form for this initial data entry and others enter the information directly into the spreadsheet. An individual case might be entered over the process of many recording sessions, because only a certain amount of information will be available at the outset; the information will be corrected and updated; and new details will inevitably emerge as a case proceeds through the justice system. Thus, while researching and recording are ostensibly separate stages of a counterdata science project (and separate chapters in this book), they unfold as a continuous back-and-forth process over time. Especially for those groups that track cases through the justice system, a single case can take multiple years to fully research and record before activists consider the information complete or at least "clarified" ("esclarecido" in Spanish). This is the word many activists used to describe cases where stable and verifiable information has come out and details are relatively certain.

Structuring data

But before individual data values are registered in discrete rows and columns, activists have to decide which fields to record, and how. Thus, recording also encompasses the process of determining which specific fields to use within activist spreadsheets and databases– that is, their data schema for how a feminicide will be registered. These range in complexity from the four fields that the Alianza started with to CFOJA's 180 variables that they attempt to collect about every case. The median is around 25 fields per counterdata project. In the case of both the Alianza and the CFOJA, their fields and categories are produced from a collaborative governance process that tries to incorporate grassroots perspectives across large territories, coordinate with other observatories, and align with emerging international standards such as the UN model protocol. In this, groups are operating in a transcalar (community + local + national + regional + global) way in how they define, systematize, and analyze feminicide in their contexts, a point we return to later in this chapter.

What are the similarities and differences in activists' data schemas? How do they map on to emerging data standards? As a preliminary answer to these questions, the Data Against Feminicide project mapped activist data schemas in relation to the feminicide data standard developed by ILDA, and we compared activist categories to those proposed by the UN Latin American Model Protocol. The ILDA data standard, in fact, draws from the UN's work, and was developed in collaboration with federal governments in Latin America (see Chapter 1). Called the Guide to protocolize processes of femicide identification for later registration, it proposes the collection of 65 variables about each case, with groups of variables that relate to the victim, the accused, their relationship, the event and place of the crime, and the legal process. For Silvana Fumega, the Director of Research for ILDA who led this process, "our aspiration is for there to be at least a minimum set of data that allows us to understand the situation of every country and for them to be comparable enough."15 That is to say, the standard is designed as a base or minimum.

Figure 5.5 shows some of the commonalities and divergences between activist data schema and the ILDA standard. In order to make this comparison, we asked interviewees if they would be willing to share their data schemas (not their data, just their list of fields) with us16. Eighteen groups assented, with the majority of groups (13/18) coming from Latin America. It's important to note that this is a preliminary, descriptive mapping to investigate how activist variables and categories align (or do not) with international standards but this cannot be considered representative of "all feminicide data activists."

All data schemas we reviewed contain identifying information about the murdered person, which most frequently includes name and age (figure 5.5a). Other frequently recorded fields included number of children, race/ethnicity, whether there had been prior complaints against the accused, and immigrant status of the victim. A number of fields suggested by the ILDA standard are very infrequent in activist data schemas, including place of birth, nationality, education and protection measures (i.e. whether the woman had a restraining order or the equivalent), likely because these are extremely difficult to source from press reports and public information. And activists frequently collect fields that are not mentioned in the standard. For example, around a third log whether a killed person was pregnant and about half registered whether the feminicide involved suicide. This follows research that shows strong links between pregnancy and intimate partner violence17. Other fields in activist data schemas but not in the standard, included race, housing situation, and socioeconomic status.

Almost all activist data schemas include information about the relationship between the victim and the accused, often integral to determining whether a murder constitutes feminicide (figure 5.5c). The majority of data schemas also include fields about the accused perpetrator (figure 5.5b), but activists on the whole include comparatively less information about the accused than about the victim. Three groups log no information about the accused. This is the subject of on-going conversations within activist circles around whether and how to "visibilize" (visibilizar) the alleged perpetrators of feminicides. These debates often surface more conceptual arguments around, for example, how to not only study the effects of harmful systems like patriarchy (such as feminicide) but to study the mechanisms for how those effects are produced and reproduced (such as toxic masculinity). Nevertheless, activists must navigate tricky ethical and legal territory in naming and potentially exposing the identities of individuals who have been accused but not condemned through the justice system. For example, Dawn, who publishes the Women Count USA database openly online, has a disclaimer message prior to viewing the database and then also in every row of her dataset that reads:

"ALL ACCUSED ARE ABSOLUTELY PRESUMED INNOCENT UNTIL CONVICTED IN A COURT OF LAW. INFORMATION ABOUT SUSPECTS OR PERPETRATORS IS OBTAINED FROM PUBLISHED NEWS SOURCES OR LAW ENFORCEMENT BULLETINS & CASE OUTCOMES WILL BE UPDATED AS TIME ALLOWS. PLEASE EMAIL CORRECTIONS TO [email protected]."

Almost all projects log numerous fields about the event and place of the crime itself (figures 5.5d & 5.5e). Commonly recorded information about the crime includes date, location, and method. Other fields include murder weapon and whether there were prior reports of abuse or restraining orders. Activist data schema have varying scales of geographic information about a feminicide (figure 5.5e). For those that create maps, relatively precise geographic locations are required, but this precision can be hard to come by from news reports and activists may have to develop alternate methods for sourcing place information. Helena, for example, reported that in some cases she has triangulated photos from the crime scene with photos from Google Street View in order to come up with approximate geographic coordinates for her database.

In relation to the legal process of the case, there are a handful of projects like CFOJA who focus on case tracking but by and large it is clear from figure 5.5e that the majority of activist data schema do not include robust information about the case as it proceeds through the justice system. This is not from lack of desire. Numerous groups told us that they aspire to track cases so that their databases may be not only a record of deaths but also records of justice served by the state, but few have the resources – time, money, and access to legal information – to be able to do so. Projects that do so successfully often have existing ties to the judicial system (like CFOJA) or else specialized legal knowledge on their team (like Observatorio Neias in Londrina, Brazil).

Two horizontal bar charts indicating how many data schemas used for categorizing by activists align with variables recommended by ILDA's data standard. The first chart concerns variables related to the victim in a feminicide. It lists variables such as ‘Name’ (18/18 activists collect the victim's name), ‘Age’ (18/18 activists collect the victim's age), ‘Children’ (11/18 activists collect this variable), at the top and ‘Sexual Orientation, (1/18 activists collect this variable),  ‘Place of Residency’ (1/18 activists collect this variable) and ‘Place of Birth’ (1/18 activists collect this variable) at the bottom. At the bottom of the first chart, there is a note that says that there fields collected by activists but not included in the ILDA data standard include: race, pregnancy, socioeconomic status, housing situation, marital status, and photo. The second chart concerns variables related to the ‘Accused/Perpetrator.’ It lists variables such as ‘Legal Situation’ (12/18 activists collect this variable), ‘Age’ (12/18 activists collect the age of the accused), ‘Name’ (11/18 activists collect the name of the accused) and several variables ILDA recommends collecting but no activists collect: ‘Gender’ (0 counts), ‘Nationality’ (0 counts) and ‘Address’ (0 counts) at the bottom.

Beyond the ILDA standard, many groups record fields that are not named in the standard and that no other group records, reflecting their different missions to interrogate economic, racialized and/or homophobic violence. For example, the Sovereign Bodies Institute has a variable to indicate whether an Indigenous woman was murdered within 50 miles of extractive industries because Indigenous communities and researchers have documented how these industries lead to a rise in sexual violence and trafficking18. Other fields – among many – collected by only one or two groups include history of substance abuse, mental illness, and whether the crime was perpetrated by police.

Finally, it is important to note that activist data schemas are dynamic. Activists add or modify fields as their understanding, dialogue with others, and political objectives evolve, which also means that they continually revisit and update existing cases. For example, Sovereign Bodies Institute has a policy to add any field that a family requests to their data schema. This is an important example of the transcalar approach I mentioned earlier, where community input can guide and shape the data recording, even when it means increasing activist labor. Annita described this dynamism: “we’re always adding new data points which unfortunately means we always have to go back and add those points for the 4,000 cases already in the system.”19

Classifying cases

Counterdata producers also develop their own categories for classifying cases – assigning types and subtypes to cases of feminicide and gender-related killing. Both the Alianza and CFOJA referenced how their classification strategies were influenced by the UN Latin American Model Protocol. As described in Chapter 2, the protocol outlines two broad categories for femicide – direct and indirect – as well as a typology that includes intimate femicide, non-intimate femicide, child femicide, racist femicide and the 15 other types pictured in figure 5.6a. As with the data schema mapping in the prior section, we (the Data Against Feminicide research team) were interested in understanding how activist categories for feminicide aligned with or diverged from the typology proposed by the UN model protocol.

A  bar chart indicating how many feminicide categories used by 18 activist groups correspond to the UN categories of feminicidal violence. It lists a mix of 26 categories on the left in descending order, with 13 UN-proposed categories indicated by the color gray, and 13 activist-proposed categories in purple. The top three categories, which are also all UN proposed categories, include ‘Child femicide’ (18/18 activists collect this information), ‘Intimate femicide’ (16/18 activists collect this information) and ‘Non-intimate femicide’ (16/18 activists collect this information). The bottom three categories, of which the latter two are UN proposed categories, include ‘Feminicide by overdose’ (2 counts), ‘Feminicide because of trafficking’ (1 count) and ‘Feminicide because of smuggling’ (1 count). At the bottom of the chart, there is a note that says ‘Categories only used by one group: Neo-liberal feminicidal violence, Corrective femicide, Gender terrorism, Feminicide in state custody.’
Figure 5.6 These charts demonstrate how activists use UN-proposed categories of feminicide and they also develop their own categories from observing patterns in their countries and contexts. The bars in beige show how many activist groups use feminicide categories proposed by the UN. The bars in purple show categories developed and used by activists that are not captured in the UN protocol. Courtesy of the author. Analysis by Angeles Martinez Cuba. Visualizations by Wonyoung So and Melissa Q. Teng.

While the majority of groups that we interviewed had heard of the UN model protocol and many had drawn some concepts and categories from it, no group directly imported its typology into their database for classifying cases. Some counterdata projects' classification schemes remained relatively simple. Groups like Cuántas Más, the Manuela Ramos Foundation and Uma por Uma simply flag whether the case is a feminicide or not. A case in their database may start in an unknown status and later be marked as feminicide once more information about it comes out in the press. Other activists have evolved more complex categorizations with sub-types of feminicide. Some of these align with the UN model protocol typology. For example, the majority of groups have either a category or variable to track intimate vs non-intimate feminicide as well as to distinguish feminicides of cis women from feminicides of trans women (transphobic femicide in the UN's categorization). Around half of activist projects track a victim's racial, ethnic or Indigenous identity in order to be able to name a killing as a racist femicide and one-third have some indicator of sexual orientation in order to be able to classify a lesbophobic femicide. The relative dearth of these categories in activist typologies may be because of the difficulty of obtaining such information from media reports; activists expressed discomfort with making guesses about race, ethnicity and sexual orientation without it being explicitly mentioned in the press or confirmed through another source. CFOJA, for example, has a field called "suspected racialized" where activists mark whether they have reason to believe that a woman is Indigenous or a person of color, and they can then follow up with their expert panel or reach out to local partner organizations to try to ascertain the victim's race. Yet they are not always able to verify such information, as is evident in figure 5.3 where, next to the pie chart about race, they have a statement that they believe an additional 4% of victims were Indigenous but could not verify that information. Finally, certain categories that appear in the UN protocol are used very infrequently or not at all by the activists we worked with: femicide because of trafficking, femicide because of smuggling, and femicide because of female genital mutilation20.

On the flipside, there are a number of common activist categories not included in the UN protocol that are used by at least a few of the groups that we interviewed (see figure 5.6b). The most common are Induced suicide feminicide – where a victim is driven to kill herself because of verbal or physical abuse – and Linked feminicide – in which a person the woman loves is killed in order to hurt her. A number of groups have categories or separate databases for classifying cases close to feminicide: Missing girls and women, Attempted feminicide, and Feminicide under investigation (denoting that it's unclear yet whether it is a feminicide). And there are a host of diverse activist categories that attempt to capture the relationship of a feminicide with organized crime and paramilitary activity: Gender terrorism, Femicide because of criminal economies, Femicide by organized crime/gangs, Femicide by armed men. These are different – and broader – than the UN categories relating to trafficking and smuggling, but all are trying to get at emerging patterns between organized crime and gender violence. Finally, there are a number of categories formulated and used only by one or two groups. For example, neoliberal feminicidal violence is a category from the Red Feminista Antimilitarista in Colombia (discussed previously in Chapter 3) which links neoliberal economic policy with gender violence.

What are these categories for? Why do groups classify cases? Most of the activists elaborated something similar to Estefanía's sentiment from Chapter 3, " it is the possibility of understanding what is happening to women in our country." Categorizing cases is intimately bound up with the resolving stage of groups' work: both their analysis of power – understanding the workings of structural inequality – and their theory of change – developing strategies to intervene in and challenge those dynamics using data. Categories enable special focuses for analysis, advocacy and communication. They permit downstream analysis of types of feminicide that may be less frequent quantitatively but take a different qualitative form than the majority of cases (e.g. transfeminicide). Groups may then highlight those numbers or those cases differently in data visualizations, infographics and reports. For example, the Alianza's map in figure 5.2 which notes the number of transfeminicides (8) and feminicides by organized crime (67), along with total feminicides (197). Additionally, activists may see patterns of violence emerging in their region and then seek to formalize them through a category so that they might advocate specifically around such cases. This was the case with categories like gender terrorisim and feminicide by overdose, both of which were categories under discussion (not fully formalized and incorporated yet) by a couple of groups at the time of our interviews21.

Neither the UN model protocol nor activists provide many categories for classifying indirect feminicides – those which are characterized as "passive" murders and include deaths from clandestine abortions, maternal deaths, and other forms of deaths due to neglect and lack of access to adequate health care, housing and social services. For example, the Black maternal mortality crisis in the US is an example of widespread indirect feminicide, where racialized access to food, housing, services and healthcare leads to the disproportionate and preventable deaths of Black mothers during or following childbirth22. Only two activist groups attempt to classify Obstetric feminicides – deaths caused by malpractice or violence in reproductive care. Mumalá, in Argentina, has a category for Social trans-travesticides – here the group tries to log the premature deaths of trans and travesti people due to social exclusion and lack of access to services and employment. A category that is emerging from the Observatorio de Equidad de Género in Puerto Rico, and elaborated in conversation with the Alianza in Ecuador is the aforementioned feminicide by overdose. This places responsibility on the state for the proliferation of women's deaths by drug overdoses due to, for example, the ready availability of opioids.

But this is not necessarily to say that there should be more UN categories or activist categories for indirect feminicides. Scholar Julia E. Monárrez Fragoso expressed hesitation about trying to create and document many different categories and subcategories of feminicide: "I think that when you open up a concept too much, the concept loses the force it has."23 In fact, this highlights some of the key limits and tensions in the concept of feminicide, and just how capacious it is or should be in order to accurately document gender-related killing. Should a death from a clandestine abortion be considered part of the same phenomenon as a death from an intimate partner or as a death from a travesti person unable to find gainful employment or as an abduction which ends in murder? These questions are not resolved in either the scholarly literature or activist practices surrounding feminicide. One of the ways that they are increasingly playing out is through global and grassroots typologies of gender-related violence – where some groups insist that a broad frame for feminicide is important and strive to count many different types of violence within that frame, while other groups focus more on counting what they can (which tend to be those cases of intimate feminicide which make it to news reports). The challenge for groups that seek to document cases with the more expansive framing of feminicide is that many types of indirect feminicide are never made public. From an informatic standpoint, for example, it will be nearly impossible to obtain information about clandestine abortions or maternal deaths or premature deaths of trans people, leading to an unknown amount of missing data24.

Managing data

Finally, the third core dimension of the Recording stage is managing data, which encompasses how counterdata producers store their data; how they handle ethical questions related to data privacy and data access; as well as how they navigate security risks – both to the database and to the activists themselves. Around half of the projects we worked with log their data in a single spreadsheet or database. The other half of the projects have multiple spreadsheets or databases. Not surprisingly, these are grouped and used in a variety of different ways. Carmen has a word processing document for missing women and another for feminicide cases. If a missing woman is found murdered, then she moves that case from one document to the other. Observatorio de Equidad de Género and Sovereign Bodies Institute also monitor missing people, not only murdered people, in separate databases. Annita describes how they maintain five different databases which cover different time periods, different geographies, and different gender identities, with a separate database focused on murders of two spirit people. Dawn groups cases into separate databases by year. Some activists also maintain separate spreadsheets for cases in progress and cases that have been fully recorded. For example, CFOJA has one database for metadata about the cases, and one database with the cases themselves, so that they can easily see which cases have been fully researched and recorded and which are still in need of attention.

Across all projects, activists voiced concerns about privacy. The abundance of such concerns is not surprising, since the work of these activists is motivated by two nearly opposing goals: first, to publicize the structural reality of feminicide, and second, to care for the individual women who were killed, along with the families and communities left behind. Only four out of 35 projects published their databases openly online, as information which may be accessed and used by anyone. Two of these expressed some ambivalence about this decision because of experiences with journalists and news outlets using their data inappropriately or without crediting them25. These individuals and groups reflected carefully about the publication of names and identifying information about both victims and alleged perpetrators and whether it could be re-traumatizing to families to see their loved ones in a spreadsheet (a discussion we will return to in Chapter 6). For Sovereign Bodies Institute, data sharing issues were particularly salient. As mentioned in the prior chapter, they had received, and declined, several requests from law enforcement offices in the United States and Canada to access their data. SBI protects their data to prevent uses that are voyeuristic, opportunistic and/or not in solidarity with the goals of Indigenous people and families – a rationale they explain in their MMIW Database Data Sharing Protocol26. The vast majority of groups simply do not share their disaggregated data and choose to publish aggregated reports, statistics and infographics.

Finally, various security risks – both physical and digital –were mentioned by around one quarter of the groups as a significant concern. Sovereign Bodies Institute partially protects their data by physically placing their server in a specific location. To secure their data infrastructure, the Alianza in Ecuador transitioned their database from Google Spreadsheets to Kobo. This shift was additionally due to the sensitivity of data they hold related to other work such as their Atlas of Rural Women that monitors threats to Indigenous land defenders27. While the majority of activists did not discuss threats to their physical security, Colombia Diversa has faced physical threats and has worked hard to secure their digital systems after facing several hacking attempts. And María Salguero, for a time, went by the pseudonym "Princesa" because she was under scrutiny from narcotrafficking groups for logging cases related to the drug trade28. Yet sometimes security threats are internal. As detailed in the prior chapter, several participants left Mumalá's observatory over political differences and took the group's single copy of the database with them. It was in response to this data loss that they instituted their federated research and recording structure, where members in each province work on their own separate copy of the database to ensure redundancy.

The practical politics of columns and categories

While there is significant variation in activist practices of recording, some interesting patterns emerge across them. Counterdata activists are intimately attuned to the politics of variables, categories and classifications within their spreadsheets and databases – both what they afford and what they foreclose. Activists recognize that they are engaged in a process of "deciding what will be visible within the system (and of course what will thus then be invisible)."29 While activist typologies of gender-related violence are often informed by legal definitions in their countries, as well as the UN model protocol as we saw in figure 5.6, they often intentionally exceed these definitions and standards. In the recording stage, activists use counting, classifications and categories as a way to challenge power - to contest the state's narrow or absent legal definitions of gender-related violence, to uplift what communities and families are naming as feminicide and structural violence, and to gain insight into emerging patterns of violence which are not named and not captured in laws or international standards.

The idea that columns and categories are political is old hat. Literature on the sociology of classification literature has long described how official counting, undertaken by the state, encodes a particular institutional agenda30. Aryn Martin and Michael Lynch name the concept of numero-politics and outline the ways that counting is both enumeration and also, at the same time, classification, because each time you count something, you are also judging that thing to be a member of the class of things that you are counting31. To record a death as a feminicide, then, is to count and to classify at the same time. But it is one thing to describe and explain, in academic literature, the fact that all counting is political and another to explicitly deploy that knowledge as a practical political tool. Counterdata activists operationalize numero-politics in the service of amplifying subjugated knowledge – forms of knowledge that have been excluded from mainstream institutions. Families, communities, women's and feminist groups know what is happening and are witness to the structural patterns and interpersonal effects of feminicide, but often the media, the law and the judicial system continue to treat cases as individual, isolated events. Says Geraldina from the Alianza about the group's mission: "One of the most important things of this alliance is to give voice to the women who say that it is femicide. We fight. I fight here at the national level...I have fought with the prosecutor's office many times because of the issue of the cases, because for them they are not femicides."32 Thus the work of recording and classifying cases becomes, in the words of Patricia Hill Collins, a "resistant knowledge project" – an endeavor to understand and challenge the unequal systems of power that result in subordination and exclusion33. And while sociologists may have long known that counting and categorizing is political, the institutions that typically produce and use numbers have a tendency to forget this fact until confronted with evidence to the contrary34. Thus, activists wield institutions' penchant for naturalizing and reifying and depoliticizing numbers against them.

This politics of counting and classifying unfolds through the rows and columns of the spreadsheet, what Helena Suárez Val names as a data frame. She asserts that it is important to understand how "different entities and relations are put in the frame through specific arrangements of data."35 Suárez Val's point is that activists are deliberate in their choices of fields and categories as a way to contest media and state framings of feminicide. Data schema, in other words, are playing both a rhetorical and representational role, directing attention towards some variables (say, pregnancy status or induced suicide or connection to narcotrafficking) and foreclosing others (say, marital status or what the victim was wearing, both of which support a victim-blaming frame). We see this in the way that some counterdata projects use variables in their data schema in an aspirational way, which is to say they include columns in their spreadsheets which are nearly impossible to obtain from media reports about the event. These include, for example, variables such as education level, socioeconomic status, mental health history, and sexual orientation. Activists know that incorporating these fields into a data schema will produce many rows of missing data or hard to verify data. Says Betiana from Mumalá, "we do not consider the data on the socioeconomic and educational situation of both victims and perpetrators to be solid data for publication."36 Thus they do not include these fields in a downstream analysis for data quality reasons, yet formalizing these variables into their schema becomes a way to assert – using the data frame – that it matters to track the connection between socioeconomic status and feminicidal violence and to demand that the state monitor such factors as well. As elaborated in the opening of this chapter, the Alianza's first manifesto dedicated a whole paragraph to outlining demands about which variables should be collected in their proposed national registry. They wanted the state to collect gender, ethnicity, and age of the victim and the perpetrator; to log geographic information; and to disaggregate patterns of violent behavior. Counterdata activists seek, through columns and categories, to not only fill in gaps where official data are absent, but also to assert what, how, and which factors "count" in a case of feminicide.37

But deliberately exceeding state and international standards for what counts as a case of feminicide and how each case should be elaborated and systematized does not mean that anything goes and everything counts. For example, the data journalism group Agencia Presentes saw reports circulating on Facebook about a trans woman who died by being set on fire in a particular neighborhood in Buenos Aires and they were preparing to include the case in their map of LGBT+ hate crimes38. But after following up with local shopkeepers, police and sex workers, the report turned out to be a rumor posted to social media by a well-meaning but wrong activist39. While most groups do not have the resources to interview people about every case (projects led by individuals in particular do not), all activists try to draw firm lines around their definition of feminicide and their standards for determining that from available information.

In order to define and classify feminicide, counterdata activists construct a data infrastructure – collective, dynamic, social and technical – around their work. This resonates with the data feminism principle embrace pluralism which states that the most complete knowledge comes from synthesizing multiple perspectives, with priority given to local, Indigenous, and experiential ways of knowing. Activists build on legal and international standards with participatory, dialogue-based processes that are transcalar. This includes dialogue and reflection about individual cases amongst members of the project and any advisors: e.g. CFOJA taps its advisory board for case knowledge and the Alianza's member organizations deliberate on cases through WhatsApp. It also includes dialogue with other activists in which they share their typologies and influence each other. For example, Carmen shifted her classification from asesinato (murder) to feminicidio based on conversations with the Observatorio de Equidad de Género in Puerto Rico. Dawn began recording information on Indigenous identity based on her friendship with Annita from SBI.

Thus, the recording stage of a counterdata project is not only about the production of columns and categories and counts, but also represents a form of infrastructuring. Following work by Susan Star and colleagues, this means the active production of the socio-technical context and set of relations that lend credence and meaning to the spreadsheets, that make space for deliberating on categories, that ask fundamental questions around what is and should be counted40. Infrastructuring, in this sense, is the production of sociotechnical relations around activist data. For example, a number of the organizations we interviewed are in regular dialogue with each other because they participate together in events like the ones we have produced through the Data Against Feminicide project or networks like the Latin American and Caribbean Committee for the Defense of Women's Rights (CLADEM), or the Interamerican Anti-Femicide Network (RIAF) or the Latin American Network Against Gender Violence41. This last group is a network of 35+ organizations that monitor feminicide by producing data (and includes many of the organizations we interviewed). Facilitated by MundoSur, a nonprofit based in Argentina, the members participate in technical capacity building sessions as well as group discussions around definitions and methods. Ignacio Piana, who leads data analysis for MundoSur, used the term "harmonize" to describe their work. He explains, "We thought that if we came with a methodology or a form of standardization already preconceived by us and tried in some way to impose it on the organizations – imagine! Thirty-five organizations with their particular struggles and some with decades of experience in this, it was very likely that it would fail right from the start because it was coming to impose something that perhaps was not consistent with what they had been doing. So, we said, we are going to co-construct it from the organizations, we are going to co-construct a methodology, we are going to co-construct definitions as a first step."42 While groups in the network share their data monthly to produce the regional map in figure 5.9 with aggregated statistics, the goal is not to impose a "standard" definition of feminicide nor a fixed set of categories, but rather to encourage alignment and articulation in order to create a space of ongoing dialogue.

This method is fundamentally different from the typical process of producing global indicators of gender-related violence. In her book The Seductions of Quantification, Sally Engle Merry describes how global indicators to measure violence against women are produced through a long and multifaceted process, involving contestation between diverse schools of thought. The approach may be pluralistic, but it often ends up as a debate amongst experts who have competing approaches: between the technocrats from the national statistics offices versus those from the criminal justice department, or between bureaucrats focused on gender equality versus those focused on human rights. Whose vision prevails often has to do with which agency has more resources and implementation authority, and the resulting international categories "are often quite different from the way violence is experienced in the everyday lives of women."43 At the end of the process, a set of standard indicators is published and circulated at the global scale.

In contrast, feminicide data activists scale via harmonization over standardization. This process of harmonization unfolds over time, nurtured by community, dialogue and relationships which provide the infrastructuring tissue. Activists are not only producing their own data about feminicide, they are also producing their own processes of data sharing and data integration, but on their own terms, respecting differences, local context and multiplicity. This infrastructuring work contributes to the dynamism of activist data schemas – we saw how the Alianza in Ecuador began by collecting four variables about each case and now collects more than 60. These shifted as the Alianza learned more about the issue, noted subregistries of specific groups such as trans women and Afro-ecuadorian and Indigenous women, joined the Latin American Network Against Gender Violence, and discussed patterns with activists in Ecuador and in other countries. This process of dynamic learning, sharing and diffusion of fields and categories resonates with rethinking binaries and hierarchies – this is the data feminism principle that asks us challenge the gender binary, along with other systems of counting and classification that perpetuate oppression. Far from seeing columns and categories as fixed or natural, activists are continually engaging in reflexive practice to add, remove or align their variables and categories with those of others.

Yet pluralistic dialogue about definitions and categories is not free of conflict. Myrna described how, as they got started, the expert panel of the CFOJA was divided on two major issues. One was whether the terms "prostitution" or "sex work" should be used to discuss femicide that occurs in this context. This relates to longstanding tensions in feminism around whether prostitution should be "abolished" (the abolitionist argument) or whether it should be legalized and made safer (the sex worker argument)44 [CITE]. In the end, the category for this violence was named "Femicide in the context of what is referred to as sex work or prostitution."45 The second issue for CFOJA's expert panel was whether the murders of trans women should be included in the observatory. Ultimately, says Myrna, "We decided that there was no research on this, that we were going to include transfemicides because there's evolving definitions of femicide. Some definitions focus on females whereas others focus on women, including trans women."46 Like a handful of other groups, the observatory decided to document all cases of trans killing, including both trans women and trans men, to better understand the contexts and circumstances surrounding violence experienced by transgender communities. That said, they do not include the murders of trans men in aggregate counts of femicide like those seen in figure 5.3 since that would be misgendering. While the group moved through these tensions, their process highlights the challenges of pluralistic governance and the ways in which acts of classification can surface longstanding political debates.

The database is a place, the spreadsheet is a memorial

In May 2022, our project team had the chance to speak at length with Julia E. Monárrez Fragoso who undertook one of the earliest academic studies about feminicide in Ciudad Juárez between 1993 and 200547. Much like the counterdata work described thus far in this book, this study involved assembling her own database from press reports, activist archives, and official data sources. She developed variables and categories, including the category of systemic sexual femicide, used today in the UN's model protocol. She has written two books on the topic and continues monitoring feminicide in Juárez using the same database, even though her academic research has moved on48. In the interview, we asked Monárrez her thoughts about the now widespread use of data activism to challenge feminicide, and she responded: "It seems to me a very important task in the sense that it is ultimately a counter-hegemonic memory against a state that does not deliver justice and imposes its own memory. These will be the central memories, because while people do take the official statistics into account, they also turn to the organizations that keep the list, the count."49

Monárrez' focus on counter-hegemonic memory as one of the primary contributions of these activist databases was striking to me and resonates with the Alianza's infographic in figure 5.2 in which their first demand was MEMORY. Some – not all but a significant proportion of counterdata producers – consider the database itself as a place of remembering, witnessing and caring for people, even in their death. This shouldn't be confused with the use of data to undertake public memory work, such as vigils and art installations that use data to represent absent bodies. We will discuss these further in the next chapter, Refusing and using data. Rather, for a number of counterdata producers, the spreadsheet itself functions as a memorial and the work of recording cases as memory work.

This surfaced in several ways in our conversations with activists. First, just as activists are attuned to the practical politics of categories and columns, they are also attuned to the power of such rhetorical subtleties as the naming and ordering of fields in a database or spreadsheet. It matters, for example, how women are described by recording their names in a field titled “Victim” versus a field titled “Name”. Helena uses "name of the woman." Dawn Wilcox of Women Count USA genders her database fields by using a possessive pronoun: "her name", "her age", "her occupation" and, for the alleged perpetrator, "his name", "his age", "his occupation." Why? A "victim" is defined by the event of her death whereas a person has a name and a community and a life which precede the event that ended it. Likewise, an alleged perpetrator is also defined by the violent event of victimizing another. In her in-depth analysis of feminicide databases as data frames – as objects with representational, communicative and narrative properties – Suárez Val asks, "Ontologically (and politically), this characterisation potentially has essentializing effects: were the persons in the frame always already victims and victimizers?"50 She argues that the victim-victimizer binary works to displace attention from the social dimensions of pervasive gender inequality; instead we see an isolated event perpetrated by a pathologized individual, on someone who was potentially "always already" a victim. Figure 5.7 illustrates this binary as manifested in the dataset about domestic homicides of women published by the Uruguayan government. While most government data schemas and many of the activist data schemas set up this victim-victimizer binary, Suárez Val invites us to rethink binaries and hierarchies for representational and political reasons. In other words, how we name single columns in a spreadsheet can communicate a whole worldview about the nature of feminicide and the lives of the people involved.

Diagram by Helena Suárez Val, critiquing the binary ontology of victim-victimizer that emerges from the official dataset "Domestic Homicides of Women" published by the Uruguayan government. The diagram tries to map relationships between different variables in the Uruguayan government's data model and group them in relationship to major categories like“victim”, “victimiser”, “event deed” and “weapon”.
Figure 5.8 Helena Suárez Val critiques the binary ontology of victim-victimizer that emerges from the official dataset "Domestic Homicides of Women" published by the Uruguayan government. Source: her thesis in English. Black & white.

Likewise, activists also make key design decisions about which variables to record in their spreadsheets not only for downstream analysis but also for reasons of narrative and representational justice. Dawn described how she began Women Count USA logging how many children a woman had, but later removed that category: “I think it’s important to remember that these women mattered not because they were mothers, not because they’re somebody’s sister, daughter, but because they had value simply because of their own worth.”51 This message is a core part of Dawn's narrative around her project and resonates with the data feminism principle elevate emotion and embodiment – a way to foreground the fact that rows in a dataset about feminicide are people. For example, Dawn uses an image with large text front and center on the project's website as well as for her business card. It reads:

"She is someone's wife daughter sister mother."

The relations are crossed out so that it simply reads "She is someone." In this case, in contrast to the prior section about activists tracking variables like socioeconomic status, it is not so much that Dawn is demanding that the state not track orphaned children, but rather that linking a woman's value to her status as a mother is dehumanizing and misogynistic, even when it is "only" in a spreadsheet. In contrast, it is important to note that a number of other groups do track the number of children left behind by a feminicide. For example, aggregated statistics showing that 197 children were orphaned in 2021 are presented in figure 5.3 by the Alianza in Ecuador. But the Alianza is using this category in the sense of reparative justice, that is to say, to track and demand reparations from the state to these families. In contrast, Dawn is using the category as representation and memory, that is to say, to think about what ideas about a woman's humanity are communicated within the narrative space of the spreadsheet itself. Both uses are intentional, both are political, but the locus of the political demand is placed differently.

The work of recording cases can constitute ritualistic, even spiritual work for some activists. Helena noted how there was a monthly rhythm to logging cases, and if there seemed to be fewer cases, she became alarmed and started scouring the media for cases she may have missed. Annita from Sovereign Bodies Institute described the work of recording cases as “spiritual” and stressed the importance of culturally grounded data production and care practices, drawing inspiration from Indigenous epistemologies. Accordingly, activists seek to protect their databases and spreadsheets from being manipulated and used in ways that disrespect the lives represented therein. For example, when Helena and I were co-teaching the online course Data and Feminicide, I proposed using her spreadsheet for an activity where learners would make a pivot table and a data visualization52. Helena pushed back, feeling strongly that the rows of data in her spreadsheet – really the lives described by the spreadsheet – were not meant as playthings for experimentation, and so we chose a different dataset for the exercise.

Screenshot of database, listing categories “date” “her name” “her photo” “her age” “her race” “about her” “city” “state” and “relationship”.
Figure 5.9 Dawn Wilcox from Women Count USA spends many hours seeking photos of killed women for her database. Courtesy of Dawn Wilcox / Women Count USA: Femicide Accountability Project.

Activist databases, then, are not abstract spaces of latent "raw" data waiting for activation through statistical analysis. Instead, they are a form of memorial, a sacred space suffused with grounded connection to the embodied lives and people who are partially represented there. In this sense, counterdata activists are also archivists and memorialists, engaged in acts of remembering that take place in databases and spreadsheets, where they link the recording of individual human lives and deaths with collective social memory and responsibility. In this, activists resist what Christine Bold and co-authors call active forgetting – the fading away of individual women's lives into obscurity as well as the repression of gendered violence as a structural pattern and a problem in aggregate53. Perhaps the sharpest example of this approach is Dawn's. She can spend hours searching for a photo for each killed woman in her database, specifically looking for photos that are not stigmatizing, not mugshots, and not dehumanizing (see figure 5.8). When a photo that meets her criteria is low-quality she will spend time, sometimes hours, retouching it before publishing it in her database. This embodies what Michelle Caswell and Marika Cifor have named, in archival practices, as a feminist ethics of care54. Posited as a form of radical empathy, the authors argue that such an ethics of care stands in contrast to archives of human rights violations that robustly document harms solely in order to seek legal redress. For Dawn, her database is a space for honoring individual lives and asserting the vast scope of the crisis. The spreadsheet, then, has value in its own right as a memorial, not only when it is activated in some future data analysis process or in some future case for legal redress. Through these small acts of care and witnessing and crafting, activists cultivate their databases into sites of active remembering, disobedient archives that refuse to submit to the normalization of gender violence.

Counterdata science vs. hegemonic data science

The work to record cases – to transform unstructured information into structured data, to document feminicides into rows and columns, to categorize cases – surfaces some key differences between counterdata science projects about feminicide and hegemonic data science projects. First, unlike hegemonic data scientists, activists have no pretense that their work to log cases and categories and variables about feminicide is neutral or objective (key values for hegemonic data science). Rather than thinking about data as "information collected, stored and presented without interest," activists are leveraging a key feature of data capture: "to collect, store, retrieve, analyze, and present data through various methods means to bring those objects and subjects that data speaks of into being" (emphasis mine).55 This is what Ruppert and co-authors speak of as data politics. Hegemonic data science, however, remains largely ignorant of such data politics because refusing the political dimension of data has largely worked, thus far, to mask the fact that it is undergirding and exacerbating deeply unequal systems of domination.

This kind of ignorance is blisteringly explicated by philosopher Charles C. Mills, who proposed that white ignorance is essential to the maintenance of white supremacy. Mills writes, "the Racial Contract prescribes for its signatories an inverted epistemology, an epistemology of ignorance, a particular pattern of localized and global cognitive dysfunctions (which are psychologically and socially functional), producing the ironic outcome that whites will in general be unable to understand the world they themselves have created.”56 White people, in other words, are ignorant about the methods and impacts of white supremacy, and this epistemological gap is, in fact, central to the maintenance of white supremacy itself. It constitutes a kind of plausible deniability about the harm caused by such a system. In the case of hegemonic data science, the ignorance – and the plausible deniability that ensues – comes from the inability or the unwillingness to admit that data have politics, that data have political effects and that these mostly have to do with concentrating wealth and power in the hands of those who already have wealth and power. Dan Bouk and colleagues frame this collective ignorance around "forgetting": "We forget that official numbers have to be made even when things are going well."57 In other words, even though institutional data are actively produced, those same institutions tend to forget the design and measurement decisions they made and take their numbers at face value, as 1:1 descriptors of the world. Likewise, Sally Merry discusses the process of naturalization that happens when creating indicators about gender-based violence: "Once decisions about categorization and aggregation are made the categories may come to seem objective and natural, while the power exercised in creating them disappears."58

But is ignorance always "a bad thing"? In recording cases of feminicide into spreadsheets and databases, activists exploit this hegemonic ignorance of data politics and bend it towards their advantage. If numbers are neutral and data in spreadsheets are 1:1 representations of reality, then activists are simply producing feminicide facts – and who can deny the objective facts? This is a use of ignorance that feminist philosopher Alison Bailey names as strategic ignorance: "a way of expediently working with a dominant group’s tendency to see wrongly. It is a form of knowing that uses dominant misconceptions as a basis for active creative responses to oppression."59 Feminicide data activists thus appropriate some of the same formal language as hegemonic data science: numbers, systematic collection, categorization and classification, spreadsheets, databases, statistics. In adopting this strategy, activists borrow the (outsized) credibility and epistemological authority of the quantitative disciplines to birth the worlds that they purport to measure.

This is not to say that activists in any way distort the truth - indeed, all groups are earnest and rigorous in their quest to record accurate, verifiable information about every individual case in their databases. Rather, the point is that hegemonic data science prizes neutrality and (wrongly) sees quantitative measures as the surest route to expunging politics and bias from knowledge. So if data are simply facts without politics, a door is open to leverage them to boost activists' structural (political!) framings of feminicide to travel further and faster, influencing the actors that groups care about the most: the state and the media.

Another key difference between hegemonic data science and feminicide data activism in the recording stage has to do with the underlying conception of the spreadsheet or database that is being mobilized in each domain. For hegemonic data science, the dataset is a disembodied space of abstraction; It is "inert representation"60; The neutral record of facts; "Raw" or latent information awaiting to be brought to life through sophisticated analysis, which then yields "business intelligence" downstream in the pipeline. In contrast, as we have seen in this chapter, for activists the spreadsheet or database is an embodied space of rich and relational connections to the world; an aggregation of prayers; a product of collective deliberation; a refusal to forget individuals; an insistence on visibilizing systemic violence. Thus, activists do strategically leverage the epistemic authority that accrues to them for using "data-driven" methods but they are not so much participating in the surveillance capitalism regime of "data as property"61 but rather proposing and modeling an alternative vision of data as countermemory and data as megaphone. Following Monárrez' emphasis on activist data production as memory work, the countermemory function serves activist struggles to "keep the memory of hidden, everyday, and private violence fresh, public, and continuous."62 The megaphone function amplifies the voices of grassroots activists and women leaders. Paola from the Alianza describes this function, "I realized that there was the possibility of generating frontline, primary source information to be able to report these feminicides as soon as possible and that this allows us to give voice and enhance the voice of the women leading these organizations."63 Both uses fulfill the function outlined by Ruppert and colleagues in which "data is generative of new forms of power relations and politics at different and interconnected scales."64

Through this data as megaphone function, feminicide data activism is concerned with scale. But it is in a fundamentally different way than hegemonic data science thinks about scale. When I was a student at the MIT Media Lab, quotes from founders like Nicholas Negroponte would circulate throughout the building, showing up at student demos and faculty presentations. Adages like "It's more about trying to do things big, because if it's not big, it's not worth your time."65 Stated back in 1995, the Jurassic years of the Internet, scalability is now the unquestioned value that underlies virtually all of tech development today. In their paper "Against Scale," Alex Hanna and Tina Parks critique the very concept, not to mention the value, of scalability - the idea that a system can expand without having to change itself in substantive ways or to rethink its constitutive elements66. They link scale to standardization, classification and colonial violence. What is scalability if not a modern-day reboot of the centralizing, extractive colonial violence that led to the massive transfer of global wealth into the hands of white European individuals, companies and institutions? In her work on theorizing "nonscalability", Anna Lowenhaupt Tsing asserts that we need to pay attention to "the mounting pile of ruins that scalability leaves behind."67 Ironically, or maybe not ironically, it is scale thinking which is directly responsible for creating the Americas that tolerates feminicide in the first place – scalability's pile of ruins normalizes gender and racial violence in the name of wealth accumulation. Documentary filmmaker and MMIW campaigner Rain describes how he is often asked at screenings, "When did this violence begin?" and the answer he always gives is "at first contact"– the moment that Europeans began colonizing Turtle Island68.

Feminicide data activists have a different starting point for scale. What if the purpose of scale is not to concentrate wealth in the hands of our billionaire overlords? What if the purpose of scale is also not to generate and circulate expert technocratic judgment, as Merry details in her account of how global gender violence indicators get produced? What if the purpose of scale is to amplify the voices, power and knowledge of the people closest to the harms enacted by colonial, venture-capital scale? Feminicide counterdata work scales on its own terms and in its own way. Activists scale through building infrastructure - technical and relational and dialogical - through geographically-dispersed expert panels, through membership networks such as the Latin American Network Against Gender Violence and also through WhatsApp groups and encuentros69. At the same time, they remain rooted in their local contexts and see their role as a megaphone – helping to propagate grassroots voices across these broader networks. Activists harmonize definitions and categories and counting practices through such infrastructures but the system tolerates divergence and multiplicity, hence the proliferation of the many activist categories that depart from the UN model protocol. In other words, multiplicity is not a problem and standardization is not the answer.

Conclusion

Recording is the stage of a feminicide counterdata project when activists transform unstructured information into structured data comprised of fields and values, cases and categories. In the recording stage, activists additionally classify cases according to diverse typologies and they manage data, including data access, privacy and security measures. Recording is the stage which most highlights how data – all data, and especially data about feminicide – are intentionally produced. Data are not facts waiting to be found, they are actively crafted into rows and columns, following processes of careful deliberation and consultation. Likewise, data schemas and categories – the systems by which activists collect and label information – are dynamic. They shift and change as activists learn from each other, observe patterns of violence in their region, harmonize with international standards, and respond to family needs.

Being at the heart of a counterdata project, recording has close ties to the other stages of work. In their process of crafting data, activists move fluidly between researching and recording cases – reading news articles or searching social media, and then logging small details into specific fields, and then returning to research again for missing values. The categories activists develop and apply to their cases relate directly to their analysis of power developed in the resolving stage of a counterdata project. Those categories then enable the downstream analysis and disaggregation of subtypes of feminicide. While activists themselves know better than anyone else that data are produced – not found – they leverage the legitimacy and apparent objectivity of numbers and classifications to amplify the voices of grassroots organizations and to shift public memory. These strategic functions of data – as megaphone and as countermemory – surface especially in the next chapter on data communication and circulation in which activists simultaneously use data and refuse data.

Comments
74
?
Susana Galan:

“scale thinking is directly responsible”

?
Susana Galan:

Delete

?
Susana Galan:

Add names

?
Susana Galan:

Uncapitalize

?
Susana Galan:

Uncapitalize

?
Susana Galan:

Uncapitalize

?
Susana Galan:

Perhaps “care about” is not the best term in reference to the state and the media… maybe “target”?

?
Susana Galan:

Add names?

?
Susana Galan:

Mention by name?

?
Susana Galan:

It may be worth making more explicit how the latter differ from the former (for example, through the use of explicit depictions of violence that may at times resemble, in image and content, sensationalist news accounts)

?
Susana Galan:

Is there a reason not to mention them by name?

?
Susana Galan:

I would perhaps delete the reference to spiritual work here, as it is not addressed in the example that follows and works better when introduced in the next sentence.

?
Susana Galan:

Perhaps ”subjectivity” in line with the previous comment. After all, a mother is still a human being. And yet, reducing a woman to its role as a mother means recognizing/valuing her only in relation to an other (in this case, the child)

?
Susana Galan:

Perhaps “objectifying”?

?
Susana Galan:

Add comma

?
Susana Galan:

Add comma

?
Susana Galan:

Here it may be worth noting (for a more general audience) that, alongside the discussion around prostitution/sex work, the question of the inclusion/exclusion of trans women is another long-standing tension in feminism

?
Susana Galan:

Add “to”

?
Susana Galan:

“capacity-building”

?
Susana Galan:

This part addresses the boundaries of activist groups’ definition of feminicide, which is a very interesting point (especially the idea that not anything goes and everything counts). The example, however, reduces this question to ascertaining whether a feminicide account is fake or not, thus limiting the potential of the discussion

?
Susana Galan:

Add comma

?
Susana Galan:

“deviating it from”?

?
Susana Galan:

Delete

?
Susana Galan:

“two-spirit”

?
Susana Galan:

For those organizations who use these expanded categories of feminicide, what are their sources? Do they rely on family, friends, activist groups?

?
Susana Galan:

According to figure 5.6, this category was only used by one group. Shouldn't it be discussed in the next sentence?

?
Susana Galan:

Some of these categories do not appear in figure 5.6. How many activist groups did mobilize them?

?
Susana Galan:

Just “figure 5.6”? Could be added “in purple” for clarification

?
Susana Galan:

Is this akin to the category “violencia vicaria”, which is used, i.e., in the Spanish law (mostly to refer to children)?

?
Susana Galan:

No need to capitalize

?
Susana Galan:

No need to capitalize

?
Susana Galan:

Also the category in figure 5.6 is “feminicide by induced suicide”

?
Susana Galan:

The explanation of this category only appears later in this chapter. Would it be possible to introduce it here instead?

?
Susana Galan:

“. Given this lack of information, activists expressed discomfort”

?
Susana Galan:

“figure 5.6” or does this refer to a different figure?

?
Susana Galan:

Does this refer to figure 5.5 or is there another figure centered on geographic information?

?
Susana Galan:

“Figure 5.5”?

?
Susana Galan:

“the importance of not only studying … but also the mechanisms that produce and reproduce those effects”

?
Susana Galan:

Add comma, or delete it at the end of this clause

?
Susana Galan:

Are these fields listed in order of frequency? Compared to the rest, number of children seems less relevant for the analysis (yet, perhaps there is a rationale behind it that could be spelled out).

?
Susana Galan:

What is the purpose of the quotation marks? They don't seem to be needed, except if it is a quote, in which case a reference should be added.

?
Susana Galan:

“to”

?
Susana Galan:

“free-form”

?
Susana Galan:

“word-processing”

?
Susana Galan:

Add comma

?
Susana Galan:

“both … and”

?
Susana Galan:

“long -term”

?
Susana Galan:

Could you expand on the questions that sparked the debates and disagreements? Was it over substantial (i.e. around questions of intersectionality) or technical issues?

?
Susana Galan:

Add comma

?
Susana Galan:

Add comma

?
Susana Galan:

Add comma

?
Susana Galan:

“with”

?
Susana Galan:

Is this a process that took place after the publication of the 2017 map?

?
Susana Galan:

How many organizations did it include when the 2017 map was published?

?
Susana Galan:

This return to the Alianza's case seems a bit abrupt. Would it be possible to conclude this part and then go on to explain what the chapter is about and clarify the terminology?

?
Susana Galan:

“as well as about”

?
Susana Galan:

Add comma

?
James Scott-Brown:

could just write “A significant proportion of…“

?
James Scott-Brown:

Other groups describe this violence as "Feminicide, stigmatized occupations" so that it could include strippers, escorts and others who may not actually engage in sexual acts.

AIUI, “sex worker” is a broader term than “prostitute”, and already includes “strippers, escorts and others who may not actually engage in sexual acts”.

The usage of “sex worker”/”prostitute” also does not perfectly align with favouring abolition/legalization: e.g., the English Collective of Prostitutes campaigns for decriminalization.

?
James Scott-Brown:

this placeholder can probably be removed

?
James Scott-Brown:

“many rows containing missing values“ or “many incomplete rows”

or “many rows with values that are missing or hard to verify“

(“rows of missing data“ sounds like it means “missing rows“)

?
James Scott-Brown:

a slightly unfortunate choice of name, given its use as a data type in R

?
James Scott-Brown:

it might be helpful to separate the discussion of digital “security risks” into Confidentiality (e.g., unauthorized access to details of land defenders), Integrity (e.g., malicious insertion of false data, such as by a hoax), and Availability (e.g., data loss due to people possessing the data leaving an organization)

?
James Scott-Brown:

“publish only aggregated reports, statistics and infographics“

?
James Scott-Brown:

“categorization for cases“ ?

”categories” and “classifying” feels redundant

?
James Scott-Brown:

“convicted“ ?

?
James Scott-Brown:

“by which“ ?

?
James Scott-Brown:

“but also to”

?
James Scott-Brown:

“Silvana Fumega, the Director of Research for ILDA, who led this process“

or

”Silvana Fumega, who led this process as Director of Research for ILDA ”

?
James Scott-Brown:

you could say that it is a choropleth map

?
James Scott-Brown:

consider a paragraph break here

?
James Scott-Brown:

consider italicising value/field/row in this sentence

?
James Scott-Brown:

I think that deleting “ Activists enter data values or points into particular rows and particular columns. For example,“ might make the paragraph clearer

?
James Scott-Brown:

“for the age field (column)“

?
James Scott-Brown:

“, and how they are used, shared, and harmonized“ would flow better