Presented at the American Historical Association Conference, 5-8 January 2023, Philadelphia, PA.

Abstract: During the 1830 French conquest of Algeria, witnesses watched, aghast, as officers looted administrative buildings, and soldiers used official Ottoman documents to light their pipes. In the following years, the plunder continued until most of the official records from Algiers and neighboring cities were either lost, stolen, or destroyed. The case of Algeria, a palimpsest of overlapping Berber, Roman, Arab, Ottoman, and French legacies, highlights the problematic nature of colonized archives and the question of what decolonizing data means in such a context. In this talk, I extend arguments for archival decolonization to the age of big data by offering a definition of restorative data justice and presenting two methods by which we may begin this work. The few extant fragments of knowledge from Ottoman Algeria emerge from French and Algerian chronicles of the governors, travel narratives, and consular records. Through close reading, structured notes, and context-informed classification, I reconstructed data sets on the governors of Ottoman Algeria (1518-1837) for prosopographical study. This reconstruction does not simply reconstitute imperial classification schemas but rather seeks to describe these men and women (in the social network study) with categories that they themselves would have likely employed. Similarly, we can use data mining approaches, which by their very definition are extractive, to address voids in the archive. Through text mining to identify named and unnamed entities and social network analysis to illustrate and study their relationships, unnamed women’s spectral presence may be recovered and represented despite their absence in the archival record. Hand-in-hand, these techniques allow us to reassemble data lost in the violence of colonial conquest and to resurrect the stories, if not the voices, of men and women long silenced. In this way, we subvert colonial weapons of quantification and convert them into tools for restitution.

Speaker’s Notes

My notes for this talk follow. Please remember that these are speaking prompts only and not a polished essay, but they may prove a useful addition to the slides shown above.

Introduction

During the 1830 French conquest of Algeria, witnesses watched, aghast, as officers looted administrative buildings, and soldiers used official Ottoman documents to light their pipes. In the following years, the plunder continued until most of the official records from Algiers and neighboring cities were either lost, stolen, or destroyed. The case of Algeria, a palimpsest of overlapping Berber, Roman, Arab, Ottoman, and French legacies, highlights the problematic nature of colonial and colonized archives and the question of what decolonizing data means in such a context.

The few extant fragments of knowledge from Algeria’s Regency period emerge from French and Algerian chronicles of the governors, travel narratives, diplomatic correspondence, a few surviving Ottoman registers, and commercial records from the French coral concessions. Through close reading, structured notes, and developing a custom, context-specific classification schema, I reconstructed data sets on the governors of Ottoman Algeria (1518-1837) for prosopographical study. This reconstruction does not simply reconstitute imperial ontologies but rather seeks to describe these men and women (in the social network study) with categories that they themselves would have likely employed. Historical data set (re)construction is one way we can begin to address voids in the archive. Similarly, through text mining to identify named and unnamed entities and social network analysis to illustrate and study their relationships, unnamed women’s spectral presence may be recovered and represented despite their absence in the archival record. Hand-in-hand, these techniques allow us to reassemble data lost in the violence of colonial conquest and to resurrect the stories, if not the voices, of men and women long silenced. In this way, we subvert colonial weapons of quantification and convert them into tools for restitution. 

This is (what I am calling) restorative data justice, a new theoretical framework I propose for our consideration today. I present restorative data justice as a response to both common historical challenges we face when undertaking studies of marginalized populations, especially using colonial archives, and it is a response to this present moment, this era of capitalistic datafication and increasingly urgent calls for social change and justice. I argue that this concept of restorative data justice may serve as a bridge between historical studies and work on current data cultures. It offers one way to redress the problematic past of colonial knowledge production and its legacy in information structures and systems still at work today in the present age of capitalistic surveillance and widespread data misuse.  As I present this concept in greater detail, you will notice that it describes processes in which many of us are already engaged but frames them in a way that will be legible to non-historians with the hope that this framework can do work both in and outside of the historical context. Just as importantly, this framework addresses our current reality of doing History in neo-liberal educational institutions where we need to justify the value of our research and teaching, even the value of our subject. It translates historians’ modes of thought and practice in ways that demonstrate how they may also be applied to understand pressing issues in the era of big data. 

Colonial & Colonizing Archives

Sources related to colonized territories and people often have problematic provenances, and those from and about Ottoman Algeria are no exception. The first challenge is that such sources are almost exclusively written by the colonizers, and the second is that the archive is often fragmentary, replete with gaps and silences in extant collections.  In the case of Algeria, nineteenth-century French colonizers destroyed many of the Ottoman records and removed the few that remained. 

“European colonial cultures are data cultures, and data is a raw material for empire,” contends postcolonial digital humanities scholar, Roopika Risam. Thus, the erasure of an empire’s data is an erasure of evidence of a previous imperial presence; that was precisely the goal of the French troops who destroyed Ottoman records when they invaded Algiers. The silencing of Ottoman records could not erase that past, however.  In spite of French administrators’ desire to remove all traces of preceding rulers in order to draw a clear line of connection between ancient Rome and their own government in the North African territory, the built environment and mixed ethnicity descendants of Ottoman-Algerian families persisted and testified to a past the French wanted to forget.

Decolonizing Data

With this background in mind, I will briefly define data justice, as it is being used among Information, Media, and Communication Studies scholars today and then present a definition for restorative data justice within and for Historical studies. 

• Data justice: Concept, framework, approach, practice
  • As a concept, data justice emerges from critical data studies, Information Science, Media Studies, and brings together critiques and historical context from the Humanities and Social Sciences.
  • As a framework, data justice “examines data issues in the context of historical legacies, existing power dynamics, ideology, and social practices” and highlights how data systems and infrastructure perpetuate or create varying and often unequal effects across different communities, with particular attention paid to communities that have been marginalized in the past and/or present. (Quote from @dencik2022, 3)
  • This approach is radical, in that it seeks to destabilize “the appearance of natural orders, … reveal contemporary power relations, and considers how they could be different.” (@dencik2022, 2)
  • As a practice, data justice, fosters and operationalizes the idea that data can be used to “challenge dominant understandings of the world, (re)create conditions of possibility for counter-imaginaries and advance social justice claims” (@dencik2022, 4)
• Data justice, as a concept, while acknowledging that data injustices are rooted in specific, historic contexts, generally problematizes present data gathering and use practices. In this talk, I link these conversations to our work as historians and present “restorative data justice” as an extension and application of critical data studies to historical research in and on Indigenous communities and the Global South. 
• Restorative data justice
  • Data justice “as a way to promote a reparation to the cognitive injustice that fails to recognize non-mainstream ways of knowing the world through data.” (Data Justice, 2)
  • Process: 
    • (1) thinking of & working with our sources as data or generating data from them specifically on/about people who have been marginalized, 
    • (2) structuring the ontology of that data based on their own epistemologies and not those of the colonizing or otherwise oppressive forces, and 
    • (3) conducting analyses of that data in ethically and historically responsible ways that highlight the experiences and roles, if not the voices, of those who have been marginalized, restoring them to their historical context, restoring information about their lives to a general, if not specific, archive of knowledge not, as the colonial archive once did, to control and dominant, but to resist imperial narratives of the past. 
  • In the historical context, this means returning actors to the narrative through data gathering and analysis, and, in my own work, re/creating Ottoman registers, at least of the governors that have been lost, and identifying women whether or not they are named in the sources to bring their actions, experiences, even their very existence, to light, to literally restore them in/to historical memory even though most are absent from the archive and have long been forgotten by their community. This effort is about restoring people to the historical record, recreating a historical record that has since been lost or destroyed, and outlining the limits of what is or can be known about this specific past. Through this work we notice the spectral presence of those long-since dead, as well as the shadow of records of their lives that may have once existed and those that only ever existed in our imaginary archive. As such, this framework and the methods I will share are evidence of the productive nature of imagined records and even the impossible archive of the imaginary, defined by Archival Studies scholars Anne Gilliland and Michelle Caswell. 
  • This approach addresses the critiques that scholars, such as Jessica Marie Johnson, among others have raised. 
    2018 article, “Markup Bodies: Black [Life] Studies and Slavery [Death] Studies”:
    • “Databases, for example, reinscribe enslaved Africans’ biometrics as users transfer the racial nomenclature of the time period … into the present and encode skin color, hair texture, height, weight, age, and gender in new digital forms, replicating the surveilling actions of slave owners and slave traders.” (59-60)
    • “Data without an accompanying humanistic analysis – an exploration of the world of the enslaved from their own perspective – served to further obscure the social and political realities of black diasporic life under slavery.” (61)

2 Methods

{Using existing methods within a new framework.Prosopography is one solution; developing new methods to work with partial information in fragmentary sources, such as using social network analysis to study unnamed people through their positions within a socio-political network, is another. Affordances? Limitations? Problematic due to quantification and apparent objectivity, as well as apparent completeness.}

Data mining is an algorithmic approach to extract patterns found within large bodies of information. It lies at the intersection of statistics, statistical/machine learning for prediction, and artificial intelligence. My own process was much more organic, although I am working toward an algorithm to automatically identify, extract and model relationships between named and unnamed people in unstructured text. Until that algorithm is complete, however, my process consists of close-reading documents, structuring notes in spreadsheets, and then creating categories based on the raw text to describe each of the variables on which I have taken notes. 

Let me share an example to make this concrete: I was interested in governors’ prior leadership experiences, fates at the end of their terms in Constantine, Algeria, how they changed over time, and if these fates were correlated with a man’s ethnicity or kinship connections. This required identifying information related to six variables: previous positions, dates in office as governor (and tenure length), fate, ethnicity, and relationships. For the qualitative variables, I first noted information on each governor as it appeared in my sources and included a separate column for citations. Then, I needed to simplify the classification strategy so I could more easily identify patterns and test hypotheses. This necessarily requires some abstraction, but I kept the original notes for reference in addition to the more simplified categories. In this way, I created a data set to replace and augment the lost or destroyed Ottoman registers. There are several key differences between this data set and the originals of course—the language in which it is encoded (English) and the variables included. Ottoman records for this period of Algerian history were written in Ottoman Turkish and included details about the value and finances of each province, including governors’ payments, and registers of Ottoman officials often included details about each person’s rank, military unit, point of debarkation, and location in the imperial province, but few records allow us to trace a man’s movement from one position to another once relocated in Algeria. 

I will briefly present prosopography as one way to (re)generate information about long-forgotten individuals. It is worth noting that although the men included in this prosopographical study were among the elite in their own time, with the French incursion into Algeria in 1830, their stories were quickly overwritten, and the French colonial period has dominated Algerian historiography ever since. I also highlight this first step of creating a data set about these elite men as a way to show how we can use such work as an intermediary step to uncover lives even more marginalized, both in their own time and in History, as I used the governors’ chronicles to reveal the lives and relationships of Algerian women during the 300 years of Ottoman suzerainty.

Method 1: Prosopography

In order to resurrect the long-forgotten governors of early modern (Ottoman) Constantine, Algeria, I combed French and Arabic language sources to identify each governor, their terms in office, ethnicity, marital relations, previous positions, and fates. However, to study these men, the most relevant method is prosopography, or the study of a collection of biographies, but this method places the researcher in the uncomfortable position of quantifying their lives in some of the same ways colonial states have employed to surveil and control colonized subjects. However, to leave them in obscurity, as historian Mary Elizabeth Perry observed, would be “to collude in [their] silencing and in the writing of hegemonic history—that is, studies of the past that promote and preserve the interests of those in power.” Therefore, to counterbalance the quantification of their lives, I pair qualitative analysis and narrative history with data visualization and statistics to describe individuals’ experiences, contextualize their stories, and identify larger historical trends.

Method 2: Social Network Analysis

[SLIDE] This table shows the first eight individuals in the data set, along with their assigned identification number, gender, ethnicity, tribal affiliation, and the manner in which they were identified.

A person’s familial relationships determined their names, which then served as calling cards, indicating a person’s place in society. For instance, “Hadj Ahmed Bey Ibn Muhammed Chérif” was a man who had completed the pilgrimage to Mecca (hadj), the namesake grandson of a former Ottoman governor (Ahmed), and the son (ibn) of a high-ranking Turkish Ottoman official who served as the right hand of several provincial governors (Mohammed Chérif). Similarly, “Oum Hani bint Redjeb Bey,” was the daughter (bint) of a former governor (Redjeb Bey) and both a literal and figurative mother (oum), as the head of the tribe into which she married. Therefore, the relationships that appear in the network signify the ways in which each person would have identified themselves, even if their first names have been lost to history.

Kinship connections can be meaningfully investigated quantitatively using social network analysis measures. Betweenness centrality scores, for example, are particularly informative because they highlight the individuals who served as essential bridges between people, family units, and socio-political cliques. In addition, a close analysis of several of the families represented in the graph corroborate our quantitative findings.

Much like the royal women in Istanbul, Algerian women and their mixed-ethnicity daughters, became guardians of Ottoman sovereignty through their various roles as brides, wives, mothers, and counsellors for their husbands.  Marriages between Ottoman officials and local women served as one of the surest avenues to power for ambitious men. Such marriages were a seal of approval for officials newly assigned to the eastern province, and provided Ottoman administrators access to local political and military networks. What is more, with their knowledge of local customs, hierarchies, and tribal relations, locally-born women became advisors and translators for their Ottoman husbands. Through their relationship with imperial administrators, Algerian women served as caretakers of Ottoman sovereignty at the very edge of empire.

The kinship connections of unnamed Algerian women provide more information about their lives than one might assume at first glance. By studying their positionality within these networks and to whom they were related, we begin to understand the lives, and roles of both named and unnamed women. As the daughters and wives of the political elite, these women were also significant political actors in their own time, even if their names are lost to modern scholars. Through marriage to Ottoman administrators, they connected Algeria to the Ottoman Empire, legitimized their husbands’ authority in the province, and bore children who embodied the tensions of empire – neither fully Turkish, nor fully Algerian. Both Algerian women and their mixed-ethnicity daughters were key links in the chain that bound Algeria to the Ottoman Empire.

Social network analysis of Ottoman Algerian society complicates the masculine picture of Ottoman imperial governance and suggests that women were key mediators between the imperial administration and the local population.  Previous scholarship has only examined male roles in household formation and political ascendance, but the data gathered and examined in the social networks above show how marriages between local women and imperial administrators were essential to a man’s path to the governorship and to the maintenance of Ottoman sovereignty.  

In this way, we are beginning to restore Algerian women to the narrative of Ottoman North Africa whether or not they are named in the sources or the archive more broadly. This is not a perfect solution, but such an imaginary ideal does not and cannot exist. This is not truly reparative, but it is a step in the right direction. What is more, the way we have represented women in the data sets and the social network graphs mirrors the way they would have understood and described themselves. This is the essential difference with this type of data mining and analysis – it does not replicate imperial ways of knowing but seeks to counter them through historically accurate representations of the actors as they would have described themselves. In what ways might this serve as model for repairing harms in the construction of data sets today?

Panel Summary Comments

This panel features four papers that explore the relationship between data and colonialism from multiple geographic perspectives. Ashley Sanders Garcia’s paper extends the work of archival decolonization in the age of big data by offering a definition of restorative data justice and presenting two methods by which we may begin the endeavor. Through an exploration of case studies of Algeria, a palimpsest of overlapping Berber, Roman, Arab, Ottoman, and French legacies, Garcia reconstructs data lost in the violence of colonial conquest to resurrect the stories, if not the voices, of men and women long silenced. Turning to the British Empire, Roopika Risam’s paper examines how sources including travelogues, land surveys, blue books, and gazetteers offer insight on how the British built and managed their empire. She argues that these sources demonstrate that data was a colonial technology, as essential as navigation, the slave ship, and the gun. Considering the relationship between data and colonialism for Indigenous peoples, Jennifer Guiliano’s paper explores digital history projects that focus on the use of colonial data in the North American context. She focuses on the ways in which digital history methods can serve as extractive processes that continue the project of American colonization. This presentation explores the tensions that exist between recovery efforts that seek to enrich our knowledge of Native peoples and their communities and the struggle of digital historians to reckon with using data that was, and continues, to exert violence on Indigenous communities. Christy Hyman’s paper will focus on the demands of geoprocessing workflows when spatially referenced primary source data animates the history of slavery. Each panelist will speak for 10 minutes, reserving the rest of the time for a discussion among the speakers and audience in an effort to probe the connections between these geographically distributed studies of data and colonialism. This panel will be of interest to a broad audience of students, historians, archivists, and digital humanists, as it engages with questions of the archive, evidence, as well as how postcolonial theory shapes our methodological practices.