Beyond extraction: Data strategies from the Global South

Abstract

This article draws upon a desk-based review and expert interviews with practitioners in the Global South to understand the diverse forms of data mediation that have become increasingly visible in the wake of the global coronavirus disease-19 pandemic. In contrast to accounts that frame the Global South solely as a site for the extraction of data and cheap, unskilled digital labor, we explore alternative accounts of the ways in which individuals and organizations in the Global South are asserting their role as active mediators of data who carve out spaces for value creation that are meaningful in their local and national contexts. From data collection and “refining” to the analysis of data for local needs and markets, these forms of data mediation demonstrate some of the changing dynamics of data practices globally and reflect the necessity of more nuanced analyses of value and power within and across regions.

Introduction

The coronavirus disease (COVID)-19 pandemic has played a significant role in transforming the global digital economy. Disruptions from lockdowns and border closures to glitches in global supply chains brought traditional economic practices to a halt. They also forced changes to the ways that goods (Gonzaga, 2021; Hawkins and Rossiter, 2022; Lin, 2022; Mezzadra and Neilson, 2022) and services are provisioned across different communities (e.g. Cook et al., 2021; Flore et al., 2021; Hastings et al., 2021; Lupton, 2020; Matthewman and Huppatz, 2020). These shifts often involved the increased reliance on technologies of digital intermediation in our lives from digital payment systems to delivery services (Squire, 2022; Starkey et al., 2021; Watson et al., 2021; Qiu, 2022; Teräs et al., 2020).

As Milan et al. (2021) have demonstrated in their edited volume on COVID-19 in the Global South, the pandemic involved a reorganization of people, goods, and flows of money in ways that often reinforced global inequities. India’s strict and sudden lockdown brought many industries such as construction to a halt (Breman, 2020). Delivery and transport platforms expanded as newly minted drivers and delivery people worked to supply groceries, medicine, and other goods to private homes (Rani and Dhir, 2020). Closed borders in the Pacific nation of Fiji led to redundancies in the tourism industry, resulting in a reduction of access to currency needed to purchase goods. In response, a group of Fijians created a Facebook page to barter food, goods and services (Finau et al., 2021). ¹ In the Philippines, digital payments increased due to restrictions on mobility, limited access to cash and fears about the transmission of COVID through paper and associated services (Acopiado et al., 2022). Marshall et al. (2021) note that families in the Caribbean region developed strategies such as “eating less preferred foods, skipping meals or reducing food intake” (pg. 25). COVID negatively impacted food security, altering the everyday coping strategies in households globally.

In this article, we suggest that other digital processes, practices, and dynamics—many of which were well underway before the global pandemic—became visible during this time. ² The simultaneous slowing down and cessation of once mundane face-to-face activities, and their widespread augmentation through modes of digital intermediation that the pandemic has brought about, made the politics and dynamics of digital practices newly visible. To examine this dynamic we draw upon a desk-based review of automation practices in the Global South (2021–2022) and our own long-term engagement with two of the three regions; Horst has carried out ethnographic research in Jamaica and other parts of the Caribbean for over 20 years (e.g. Horst, 2006, 2008, 2014) and Sargent has worked in India since 2011 (e.g. Sargent, 2019, 2020, 2021). We followed up our desk-based review with expert interviews to understand how differently positioned actors in the Global South are using ADM technologies in projects of value creation. Based upon these analyses, we focus in this article upon three processes of data mediation—specialization, localization and modularization—where organizations in the Global South actively mediate data for their own ends. We conclude by reflecting upon the ways in which such shifts nuance accounts of the power relations shaping the digital economy in the Global South.

It should be stated at the outset that the three case studies are not necessarily representative of the range of data practices across these diverse regional and national contexts. Nor do the case studies provide full accounts of the social dynamics and politics of each type of data mediation in situ. Rather they should be seen as partial accounts that, taken together, suggest initial frameworks for attending to the heterogeneity of data relations that make up the global political economy of data. They do so, in part, by foregrounding the experiences and perspectives of those people involved in the day-to-day engagements with data. With this stated, we certainly acknowledge that the disruptions and reorientations of digital economic practices we discuss across the Global South take place within larger political-economic relations characterized by the extraction of data and inexpensive unskilled digital labor that benefits consumers, companies and organizations in the Global North (Anwar and Graham, 2022; Couldry and Mejias, 2019b; Irani, 2015; Madianou, 2021; Ricaurte, 2019). They also rely upon digital technologies that, despite the origins of the minerals or production processes, may benefit individuals and corporations in the Global North. Nonetheless, the pandemic made evident a more complex picture of the political economy of data and technology, one that warrants greater scholarly attention in our understanding of data practices (Siles, 2023; Soriano et al., 2021).

Data extractivism

If the ongoing global pandemic has made the infrastructures of global commerce more visible through their breakdown, it has also illuminated the flows of data that constitute contemporary global connection. As has been widely noted, technology companies responsible for managing and maintaining these infrastructures saw huge profits as those who were able began to work and socialize online. The impact of these changes involved not only the direct use of digital infrastructures and platforms but also resulted in profits derived from the massive amounts of data extracted from these activities. By increasing dependence on digital platforms, the pandemic raised the stakes for understanding the economies of data that automated decision-making technologies rely upon and create.

One way in which data economies have been understood is through the lens of extractivism. The concept of extractivism emerges from scholarship and activism in Latin America (Riofrancos, 2017) that has critiqued the practices of appropriating natural resources (from mining to mono-crop agribusiness) and the social, political, and cultural logics that have emerged to legitimate and support them. Critical data scholars have taken up this concept to analyze the increasing economic value that is accumulated through the appropriation, aggregation, and interpretation of data (Chagnon et al., 2022; Cohen, 2018; Morozov, 2017). As Mezzadra and Neilson (2017) argue, the framework of extractivism usefully draws attention to the productive role of frontiers in contemporary capitalism. Data, like natural resources, are treated as cheap inputs to capitalist production. As with the extraction of natural resources, powerful discourses that frame pieces of our material and social environments as “just there” form a fundamental first step in legitimating the appropriation of these environments for private accumulation (Sadowski, 2019).

There are several ways in which this relationship manifests. Some accounts of digital extractivism have focused on forms of online mining, such as in cryptocurrencies, where digital operations result in an informational good with financial value (Mezzadra and Neilson, 2017). Other accounts have stressed the irreducibly material nature of digital technologies, relying as they do on massive data servers which, like other processes of extraction, lead to ecological damages (Taffel, 2021). For our purposes, we are most concerned with approaches that use the lens of extractivism to understand the collection of ever-increasing amounts of data about populations (Cohen, 2018) but also environments (Ricaurte, 2019) as financially valuable inputs in algorithmic models. We build on this work by drawing attention to nuanced ways in which data are used in diverse projects of value creation in what we broadly conceive as the Global South. Through a discussion of the ways data practices were shaped by and through the pandemic, we seek to further nuance analyses of the global political economy of data. Our aim then is not to refute the critical insights provided by studies of data extractivism but rather to draw attention to the diverse relations of power that undergird, reproduce, and transform data economies.

From data extraction to data mediation

Scholarship focused upon data extraction has made a crucial intervention by drawing attention to the economics of data and, by implication, the potential for the records created by mobile sensors, communication platforms, and other digital infrastructures to be approached as a seemingly simple and naturally occurring resource. Popular discourses of data as the “new oil” (see Taffel, 2021) imply that data precede the political and economic machinations of big technology companies. Here data are framed as simply an unintended by-product of our increasingly digitally mediated activities or our “digital footprint.” Yet, as scholars of digital extractivism argue, there is violence inherent in the seemingly innocuous collection of “digital footprints” (Ricaurte, 2019). Couldry and Mejias (2019b) treat this dynamic as a new form of colonialism that reproduces geopolitical inequalities of older forms of colonial domination. Here, technologies of algorithmic prediction and surveillance replay violent forms of appropriation that harnessed the lands and peoples of what is now the Global South for capitalist profit (Couldry and Mejias, 2019a).

In our recent review of automated decision-making (Horst, et al., Forthcoming), extractivist dynamics frame many of the ways in which the Global South is integrated into the global digital economy. Throughout our research, we encountered alternative forms of relationships and data economies that move beyond any straightforward relation of data extractor and extracted. Interviews with a series of stakeholders revealed how the creation of value was being developed through three processes of data mediation, what we have termed specialization, localization, and modularization. Our aim here is not simply to argue that these processes are more complicated than existing academic models suggest, indeed this is always the case, but to begin to point to processes that provide an alternative picture of global data economies. We name these processes in the hopes that they may help draw attention to lived data mediations as they are practiced in other parts of the world. They are thus not fully formed analytic concepts in their own right, but rather conceptual foundations for approaching the global political economy of data from the emplaced practices and strategies of the Global South. Moreover, they should not be taken as autonomous strategies of resistance, although some forms of mediation may lean in the direction of producing greater autonomy. All of the practices we outline rely on technology, methods, infrastructures, and funding that come from elsewhere, often locations in the Global North. Rather our named strategies center upon what practitioners in the Global South are making with data as they are collected, stored, analyzed, and mobilized. Specialization involves a process whereby data are collected and analyzed for specific projects or tasks. It often involves defined skills not only around the use of technology but also the analysis and interpretation of the data gathered. However, unlike modularization or localization, the data are handed over to clients once an analysis is complete. It is often an approach used in response to local or time-delimited issues and therefore forgoes processes such as “storing” data for accumulation or reuse. Like specialization, localization is often focused on its purpose and occurs where data are collected for use within a circumscribed local economy. By contrast, localization often involves the accumulation and reuse of data as a critical element in the creation of value within local circuits (Maurer, 2012). Modularization is a mediation process where data are refined into a standardized product for a larger and often global market. While data are accumulated its collection is often outsourced as the focus is on creating a data product that will become infrastructural within global data economies. Modularization aims to create not simply data products but standardized instruments of data analysis which can be embedded in a wide range of applications. While these terms are, of course, not exhaustive of the complex ways in which data are mobilized and mediates social life, we foreground these to draw attention to the complexity of these dynamics by highlighting the shapes and logics of three different practices. These forms of mediation became more prominent during the pandemic as companies were forced to explore different strategies in the face of lockdowns and economic disruption. But these forms of mediation were in practice well before the pandemic and are significant not only as responses to the pandemic but also as frameworks for nuancing our analyses of the political dynamics of global digital economies. Moreover, these forms of mediation continue to increase in significance in the current phase of the pandemic as different organizations in the Global South develop their own technology, data ecologies, and systems alongside and even outside of Global North frameworks.

Specialization

Our first case study, from Flying Labs Uganda, is an example of specialization. Part of a broader Flying Labs Network established by humanitarian Patrick Meier, the network was developed to build locally led knowledge hubs across 32 countries in Africa, Asia, Latin America and the Pacific. One of the key initiatives is to build on local expertise in drones, robotics, data, and artificial intelligence (AI) to train a broader community through collaborations aimed at developing humanitarian, health, development, and environmental solutions. Indeed, Flying Labs Uganda’s aim is to improve livelihoods through harnessing location technology and robotic technologies “for good.”

The critical literature on drones globally highlights the complexities of using drones in humanitarian responses (Crawford and Finn, 2015; Madianou, 2021; Madianou et al., 2016; Taylor and Broeders, 2015) and their use in war as methods of surveillance (Richardson, 2018, 2019). Flying Labs Uganda largely prioritizes the use of drone technologies for local disaster responses, community development, and refugee activity. For example, the Flying Labs Uganda team implemented drones as part of a disaster response strategy in the aftermath of the Bushika landslides in 2019, which destroyed hundreds of homes. Drones were used to view the area and to strategize how those affected might be evacuated by identifying access points. Through predictive analytics, the technology also enabled the monitoring and evaluation of the area to facilitate the relocation of displaced residents and collected image data and mapping data have also been used in areas where there are large numbers of refugees. Drones were also used to identify areas close to refugee settlements and evaluate if the surrounding landscape is suitable for developing infrastructure such as schools. In addition, because one of the major sources of income and “the backbone of the economy” is agriculture, drones hold promise for their potential to provide valuable data for crop analysis.

During the first few months of COVID-19, Flying Labs Uganda analyzed data to support local organizations to coordinate the COVID response in refugee settlements. As the Director of Uganda Flying Labs described, one of the organizations,

used their images . . . to see how best they can respond, . . . such as Covid response where the clinics are the access points. Are the roads wide enough to evacuate our refugees if it is becoming critical for them to get oxygen in the hospital? Are there water points to serve the community so that they can do constant washing? How best can they educate the community? Those were images which were getting all that information. How close are their homesteads in the markets? How crowded is the area?

Interview 31 July 2021

As COVID-19 progressed, it became difficult to implement some of their plans. Indeed, the increasing number of lockdowns during the peak of the pandemic forced Flying Labs Uganda to take a break from implementing some of their newly planned projects, such as delivering HIV medication in remote regions and expanding to robots and technologies to deliver health care and supplies. In its place, the team in Uganda focused upon training staff in the specialist knowledge required not only to collect the data via drones and other forms of automation, but also to properly analyze and comprehend the data to make it usable to different stakeholders. Such trainings did not rely on or utilize datasets made public by previous clients. Many of these trainings took place online over Zoom. In effect, COVID made their specialist knowledge—and role as a mediator between public and private sectors—much more visible. Note here that this role turns on the specialized skill of data analysis and is limited to funded projects. They do not leverage or seek to extend this optimization through additional analysis of data or a marketable data product as we see in the example of localization.

One of the interesting features of this context is that Uganda’s automated decision-making initiatives have been largely driven by the national government. Yet, one of the key challenges is that the availability and deployment of drone technologies outpace the development of consensus around protocols and policies for the use of drones and the data generated. As the Director of Flying Labs Uganda, described, “When we use drones there are so many regulations and some loops which we have to go through, and that is what is slowing down the whole process of drones being effectively used” (Interview 31 July 2021). The Director further outlined a scenario of training agriculturalists and noted that such training requires approval from the Ministry of Agriculture and the Ministry of Defense and the Aviation Authority to validate licenses, set flight paths and identify aeronautical conflicts. The Communication Commission then needs to approve flight paths to ensure drones’ radio transmission does not disrupt the workings of existing infrastructure.

The current landscape of cooperation between organizations like Flying Labs Uganda and government agencies has been built on support by the Minister of Information and Communications Technology to grow innovation but remains subject to challenges of disrupted funding. As the Director of Flying Labs Uganda observed,

The private sector come with a big mind . . . but then the government seems to say, “Okay, go ahead and implement it.” It still comes back to us to implement . . . The government are the policymakers. So the private sector goes back and takes the idea to implement on behalf of the government. That is what actually becomes the pride of Uganda because the private sector is the one which implements start-ups and many other sectors are really into it, and the government supports through policies. So they kind of complement each other.

Interview 31 July, 2021

However, because the use of drones for development-oriented initiatives requires private sector investment, data from the initiatives designed for public goods ultimately are returned to the organizations, company, or entities that funded the request; data are owned by the people who commission the research rather than the Flying Labs or the broader community. Combined with the lack of funding available for storing and managing these data on servers and the need to comply with government regulations around defense and what is effectively viewed as within the national interest, data from drones and the analysis that an organization like Flying Labs Uganda collects cannot be leveraged for other initiatives and activities. This, in contrast to the case study on localization, means that Flying Labs Uganda lends its skills and expertise not only in collecting and analyzing data, but also to navigate the regulatory infrastructures required to operate drones in Uganda. Even in the more general training that they offer which can draw on data made public by previous clients, these data are made valuable as part of providing a service; namely training a corporate workforce, government employees, or private individuals. They have developed a specialist service and skill set that is desired by government entities and nongovernmental organizations (NGOs) working in the region.

Localization

Our next case study of data mediation is focused upon localization and a company, Stone Technologies, founded and based in the Caribbean nation of Jamaica. Stone Technologies is a private company providing AI and machine learning (ML) Services for local civic and consumer organizations in Jamaica. Owned and operated by Jamaican Matthew Stone, Stone Technologies is composed of a small number of staff members many of whom graduated from the Department of Computing at the University of the West Indies (UWI) which has provided support to students to develop and market their technology innovations.

Although there are a range of solutions that the company provides, the company has developed expertise in operating and analyzing drones. Like Flying Labs Uganda, Stone Technologies uses drones to monitor particular places and activities. For example, one of the company’s biggest contracts was with the Jamaica Public Service (JPS) Company, the entity that provides electricity throughout the country, including to densely populated urban areas where illegal electricity use (e.g. informal electricity networks) emerges, and checking meters and other aspects of measuring usage can be challenging for human meter readers (Horst and Miller, 2006; Horst, 2008; see von Schnitzler, 2008 for an example in South Africa). JPS began working with Stone Technologies to use drones to automatically read numbers from electric meters, both analog and digital, enabling access for difficult to reach locations and saving staff time and resources. In these examples, Another Stone Technologies drone-based initiative is a university campus-based security mobile app designed to be used by staff and students. Once the app is triggered, the drone will prompt a (human) security officer to respond. It also offers video footage of the incident for use by the police, which they have designed to act as a deterrent toward an assailant.

While on the surface these activities look similar to those of the drone services provided by Flying Labs Uganda, there are key differences in the way that data are used. Flying Labs Uganda offered a particular data collected for a project as part of the service offered. In contrast Stone Technologies used the data it collected to optimize operations and build up local talent and expertise to address local (e.g. Jamaican) needs and concerns such as contributing to the reduction in theft and violence in specific sectors. As the founder and chief executive officer Matthew Stone asserted, “we need to stop being consumers [of technology]. We need to design our own solutions” (Interview 9 March 2022). While Stone was not averse to investments coming in from outside of Jamaica to assist with scaling the business, there is a clear vision to address local problems through local circuits. Scaling up means designing technological solutions and reaching out to other Caribbean countries. This also includes using the data collected for particular projects and partnerships to further optimize the company’s operations as they work to grow their company into other areas. As the company website proclaims, “We build smart solutions using that data, combined with the latest technologies in AI and Machine Learning, to provide key business insights and analytics. We also build automated solutions that can help companies streamline their efficiency and increase revenue” (stonetechjm.com/about/). Their aim is to not only build ML/AI technologies and analyses but also to “realize the potential gains from the data that they collect.” Unlike the case of specialization that focuses on the provision of bespoke analytical services, here data are mobilized for the creation of distinctly local value chains. Collected, stored, and reused within the region, data become a key resource in a local digital economy.

By committing to a localized, context-focused approach, Stone Technologies builds upon and leverages existing relationships and opportunities within the region’s tech-preneur landscape. As an example, the company’s livestock-drone product emerged from a collaborative process with PreeLabs Ltd, another UWI incubated startup, working in the energy conservation and management sector. Through UWI connections, Stone discovered PreeLabs already had established relationships within the farming sector through their Pree Smart Farm System, a remote farm management and automation system. With Pree Labs already working on motion sensors within their farm system, Stone Technologies, through collaboration, could leapfrog design steps to successfully bring the drone-based technology to market.

Further highlighting this local-first approach, Stone Technologies has turned to local agencies, such as the Jamaica Computer Society and Jamaica Business Development Corporation (JBDC), to help nurture the company’s growth. While the Society has provided opportunities for practical information exchange with the country’s information technology professionals, JBDC acts more as a business incubator, which has enabled Stone to gain access to experienced Jamaican businesspeople, associations and research communities. Although only a recent tech-startup, Stone Technologies is already seeking to “pay-forward” this local-first approach through his creation of the Jamaica Artificial Intelligence Association (JAII). With the pause to product development brought on by the pandemic, community building has taken an increasing share of Stone’s time. A key goal of JAII has been to bring together other Jamaican tech-preneurs and also provide pathways for UWI technology graduates into the industry, as an important step to “develop the whole field of automation or AI machine learning . . . in the country” (Interview 9 March 2022). Zoom technology during the pandemic has enabled these relationships to grow and flourish. As Stone poignantly described in our interview these kinds of communication tools have served to bring “everything . . . online” (ibid.). While video conferencing software has, as with elsewhere, served to increase young people’s engagement with a variety of activities (e.g. Cirucci, 2023; Mandache et al., 2021; Shaw et al., 2022), Stone signaled how Zoom could bolster the participation of UWI students in Jamaica’s AI community. With these types of communication technologies enabling access to JAII, he hopes this will “help those younger people, to kind of guide them what’s the best path to learning AI and having some impact in the field” (Interview 9 March 2022). Thus, these technologies, despite their ability to shrink global communication pathways, are mobilized by Stone Technologies to build and maintain a local network of AI experts and resources by extending and reinforcing pre-existing community relationships.

Modularization

Rather than bypassing the extraction of value from data (Flying Labs) or circulating it locally (Stone Technologies) Blue Sky Analytics presents us with a data practice—modularization—that capitalizes on the untapped value of data but does so by refining data into standardized products that operate within clients’ data analysis systems. The products that Blue Sky sells take on value precisely because they can seamlessly fit into a variety of systems. Founded in 2018 by a brother and sister in New Delhi, India Blue Sky Analytics describes itself as a “geospatial data intelligence company” (https://blueskyhq.io/). By leveraging satellite imagery collected by other organizations, the company creates Application Programming Interfaces (APIs) that provide clients with accurate data about the environment. These APIs allow companies to include relevant environmental data in risk assessments and, Blue Sky hopes, determine how they might lessen their environmental impact. To this end, the company has made some of its climate data products freely available on its website and it is involved in a number of initiatives to address climate change and environmental sustainability.

The pandemic has impacted Blue Sky in two ways. First, it increased interest in climate tech by providing something of a preview of what widespread disaster and disruption to global supply chains might look like. As the Blue Sky employee we interviewed noted, companies had become much more interested in risk analysis and specifically modeling and preparing for widespread disruptions after the outbreak of Covid-19 and its attendant chaos. Second, the pandemic made Blue Sky more visible because of its practice of data mediation. While forms of data mediation that relied on local creation and accumulation of data (e.g. Stone Technologies) experienced pauses in production, modularization with its focus on data refinement was well suited to surviving the disruptive conditions of the pandemic. In this sense, the pandemic brought to light the way profits could be gleaned from the refashioning of existing data. For Blue Sky Analytics, this focus on refining as opposed to collecting data was something that set them apart not only from other tech firms but specifically from other climate tech firms focused on sending up new satellites.

While modularization involves the repackaging and refinement of existing data, it is also aimed at a global market. Blue Sky Analytics is a corporation with ambitious goals. The company’s CEO and Co-founder Abhilasha Purwar described the company’s mission as becoming the “Bloomberg of environmental data” (Balachandran, 2020). The comparison with Bloomberg highlights the company’s aspirations for its products to become global and infrastructural. Like Bloomberg’s financial data, Blue Sky seeks to supply climate data to companies around the world, enabling them to make more environmentally conscious decisions. More than this, they seek to make their climate analytics into a standardized piece of business infrastructure, much like Bloomberg’s financial software is for finance companies. By cultivating relationships with data collectors—from NASA to the European Space Agency—Blue Sky is able to aggregate large amounts of geospatial data that it then transforms into customized datasets on particular environmental risks (e.g. forest fires) or conditions (e.g. air quality). Through this process, Blue Sky creates value through the refinement of data, transforming it into a product that can become valuable infrastructure for global companies.

The case of Blue Sky does not fit easily into the framework of data extractivism with its focus on flows of value from extracted to extractor. They do not provide a source of data or cheap labor for their clients, although they rely on cheaper labor costs in India. The data that Blue Sky Analytics primarily relies on are collected by space agencies—namely, NASA and ESA—and is essentially free. As one employee of Blue Sky noted, this model sets the company apart from other climate tech companies. She observed that “everyone’s trying to collect more data, so more satellites are going up into space. But there’s already so much data collected that it’s just sitting and it’s not analysed” (Interview 14 January 2022). Certainly, Blue Sky operates within a larger context in which data are being extracted as a resource for producing products and profits, but this tells us relatively little about their own activities since they neither extract the data nor market directly to consumers. Rather their data mediation practice aims to create new analytic infrastructures to embed environmental analysis into a wide range of corporate processes. Profit here still turns on the processing of data, but it is precisely at this stage of processing that Blue Sky operates.

The same employee explained that Blue Sky sees itself as creating a “data refinery,” an image that is repeated on their website (see Cohen, 2018 for a discussion of data refineries dealing with consumer data). Building from the notion of data as a natural resource—the new oil—the image of the company as creating a data refinery positions them not as providers, or even brokers, of “raw” data inputs but rather as purveyors of refined data products. The larger political economy in which Blue Sky operates is certainly widely understood in terms of data extraction, yet in emphasizing their work as building a refinery, Blue Sky draws attention to their specific position within this economy as neither extractor nor extracted, but rather as processors and builders. As the Blue Sky employee explained,

So, in the market, in the ecosystem, there are people building environmental datasets, or maybe even more use case specific, like there is a lot of remote sensing satellite imagery that is being analysed for particular say agriculture, precision farming, those kinds of things. But the approach that most companies have been using is to start from scratch, whenever they have to build a new data set. And when we were building our first product, air quality data, we were like a lot of components in the process to bringing it where it is today are actually common.

Interview 14 January, 2022

She went on to explain that by creating a platform to handle the processes that are common to creating different environmental datasets—fetching raw data, analysis, and so on—the company has been able to drastically reduce the time it takes to create specialized datasets and applications. The image of Blue Sky as creating and operating a data refinery speaks to the company’s trajectory toward taking up an infrastructural position in data economies.

Blue Sky’s position within the Global South is crucial to its success in becoming a provider of refined data products. Part of what enables Blue Sky to take advantage of this position is the larger geopolitics of labor and development. After demonstrating a new product called Space Time, a platform for visualizing air quality data, the employee remarked on the importance of the company’s location in India.

One of the main things is that skilled capital is so affordable. So, the kind of platform I just showed you, we built in a year-and-a-half, right. The amount of the volume of the data we’ve analysed in a year-and-a-half to get here would have taken a lot more capital if it was done anywhere else in the world.

Interview, 14 January 2022

Here, the availability of cheap and skilled labor was directly responsible for the platform’s short development time. Here we can see how data mediation strategies emerge out of longer histories. India has long been a site of cheaply available skilled labor, specifically in the tech sector. While this labor has largely been exploited by companies based in the Global North it is increasingly enabling small Indian start-ups to operate within the global economy. Notably, Blue Sky has effectively used its connections to centers in the Global North, pitching for venture capital funding at MIT Solve in 2019. ³ In this sense, Blue Sky Analytics demonstrates how Global South actors can leverage geopolitical inequalities in their favor to redistribute, if not reshape, data economies. Unlike the case of global labor platforms which often rely on digital labor from places like India, the case of Blue Sky suggests that Indian companies are able to be competitive in delivering key digital infrastructure precisely by taking advantage of lower labor costs.

Data mediation: beyond the binary

The process of unequal interconnection that accounts of data extractivism highlight captures a central dynamic of capitalist accumulation through data extraction. It draws attention to the political dynamics of big data economies. For example, studies of digital labor in the Global South have brought critical attention to the precarious human labor that undergirds AI solutions, from data cleaning, to training, to even posing as ADM systems by conducting manual analyses for companies purporting to supply automated services (Anwar and Graham, 2022; Ekbia and Nardi, 2017; Shestakofsky, 2017). These analyses draw attention to the uneven terms on which different regions of the world are incorporated into the digital economy and also challenge the illusion that automation, especially where it is powered by AI, simply supplants human labor (Munn, 2022). Yet this focus on extractivism tends to flatten the diverse relations through which data are put to use. The Global South figures primarily as a site of data extraction, often even less protected than areas in the global north with nominal data protection regulations, or as a source of cheap digital labor.

Our case studies nuance this framing of the Global South by highlighting the complex power relations that shape diverse projects of value creation through data and help us move beyond a singular focus on data extractivism without losing sight of the political stakes of automated technologies. In thinking about what is beyond extraction, our goal is not to discount the important interventions of theorizing data extractivism. Rather, we seek to sharpen this analytic in two directions. First, we seek to bring attention to the power relations that structure economies of data in ways other than straightforward extraction. That is, we focus on the political-economic relations that are produced after a given swatch of human activity has been captured in a digital database. When collection, analysis, and interpretation occur under the auspices of large firms, it is easy to collapse these steps into unidirectional flows of data and value. Yet, as our examples show, there are complex ways in which actors in the Global South have taken advantage of positions that rely on processing data after or aside from its extraction. Second, our case studies draw attention to the power dynamics of Global South actors that exist outside of relationships of data extractor and extracted. The organizations we explore do not easily conform to the position of providers of raw data or cheap labor, rather they offer examples of some of the diverse data mediations that make up political economies of data. These are not unambiguously alternative forms. The technology, funding, and even training that undergird these strategies of data mediation come from elsewhere and are caught up in global political economies of technology and trade. But what our cases do illustrate is precisely a hybridized space of contestation in which data can be mobilized for heterogeneous ends. From the specialized services of Flying Labs in which data are produced and used for creating purpose-built solutions to the localized circular economies of data that Stone Technologies maintains, to Blue Sky Analytics’ vision of becoming a provider of modular data products sold to a global market, these Global South actors are neither extractor nor extracted in any straightforward sense. That these positions are still entangled in the data economies of global capitalism is part of their significance. Our cases do not obviate the politics of data and automated technologies. Rather, they suggest a more nuanced and multifaceted political economy emerging from the increasing use of such technologies.

One way in which our case studies provide greater insight into accounts of data extractivism is in presenting diverse approaches to the use and ownership of data. Perhaps most counterintuitive in this regard is Flying Labs Uganda. Unlike the models of big technology firms which seemed to be premised on the accumulation of ever-greater amounts of data, Flying Labs only collects data for specific projects and presents it to the organizations that have commissioned them as part of the work provided. Data here are bespoke in the sense that it is collected and analyzed for specific purposes rather than as a valuable end in itself. Importantly, this strategy of data use grows out of the political and economic context; government policies complicate data collection and lack of capital constrains access to large data storage infrastructures. Stone Technologies contrasts with this strategy in that the company collects and retains data for the express purpose of optimizing its operations and innovations. On the surface, this may seem more similar to data extractivism, yet here data’s value emerges in a tight local circuit. While data are certainly an asset, they are one that is being used to grow local data economies. Blue Sky Analytics demonstrates a data strategy that is global from the sourcing of data from European and US satellites to the sale of data products to companies across the globe. In this, Blue Sky cannot be seen simply as a supplier of data or even of cheap labor. Rather, the company has ingeniously found a source of virtually free data produced in the Global North, which it uses to create what might become a piece of digital infrastructure owned and operated from the Global South. Taken together then our cases demonstrate that the ways in which data are being used (e.g. as bespoke service, as local resource, as refinable material) in the Global South go well beyond being the object of extraction, as important as this use is.

Narratives of data extractivism implicate a complex relationship to place, simultaneously bound to and free from it. On one hand, data function as a placeless extracted resource, anonymized, usable, and tradable across hemispheres. At the same time, the Global South operates as a key site for extraction and supporting the machination of data for exploitation by ADM technologies. Our case studies indicate specific ways that place informs relationships to data. With Uganda’s Flying Labs, “place” is liminal. Place draws different parties together, companies, NGOs, and other organizations on one side in need of localized data, and local ADM practitioners on the other side carrying out data extraction. Place is not concerned with leveraging data and its transfer to localized capital markets through new value-added goods. In contrast, Stone Technologies’ approach to data occurs, at least in the near-term, as a fixed space. Data extractivism operates within a decidedly local terrain. It draws upon nearby experience, expertise and support to meet local needs. It operates within a data economy, witnessing sharing, cooperation, and a willingness to uplift others through leveraging existing relationships and incubating new ones. For the moment at least, “local is king” (if to borrow a term currently applied to the flourishing African digital platform sector, Harrisberg and Farouk, 2022). Blue Sky Analytics, on the other hand, maintains a clear, outward looking worldview to place. Their data mediation practices function to create globally circulating data products. In this sense they are selling back finished products to companies often based in the Global North where much of their raw materials are sourced from. Modularization sets the dynamics of data colonialism askew albeit in ways that might leverage postcolonial inequalities within India (Thorat, 2021).

As another piece of soft power as India moves closer to global superpower status. What these case studies demonstrate in respect of place is that the Global South is far from monolith. Local governance and regulation (or lack thereof), local intention, and national aspirations all appear as important determinants of contextual relationships to data.

Conclusion

The twinned pandemic effects of increasing technological penetration and visibility have illuminated many of the invisible forms of data mediation in the Global South. The analytic of extraction that is often prevalent in discussions of data in the Global South draws attention to geopolitical hierarchies in the global political economy of data. However, it also tends to efface diverse data practices in the Global South. Throughout this article, we explore the ways in which the COVID pandemic has enabled us to see relationships around data that complicate the trope of extraction. By looking at different forms of data practices in Uganda, Jamaica and India, we demonstrate that the Global South is more than a site of cheap labor for the production or analysis of data; it is also a site of contestation that refracts and reroutes the lines of accumulation on which global data economies rest.

Our three case studies demonstrate the need to focus on how ADM technologies are put to use in diverse social, historical, political, and economic contexts. From the emergence of platform labor arrangements that inform algorithmically managed beauty work in India (Raval and Pal, 2019), to digital contract work in the Philippines (Soriano et al., 2021), ADM systems are reworked through imaginaries, of labor, data, or profit. Of course, these re-workings are always partial, constrained, and tenuous, but to fully understand the forms of exploitation and accumulation that shape contemporary data economies, such negotiations cannot be ignored. The critiques associated with extraction challenge prevalent popular discourses that trumpet the developmental promises of ADM in the Global South (Arora, 2016) while also calling into question the universalism of accounts based on the effects of ADM in the Global North (Milan and Treré, 2020). While the dynamic of extraction is certainly an important feature of the emerging global data economy, it is not the only dynamic, nor can it account for the diverse routes through which data are being made valuable in different places across the globe. Centering our accounts on the uses of data foregrounds the diverse social practices that undergird data economies. The aim here is not simply to counter meta-narratives with local specificity. Rather, taking inspiration from feminist critiques of capitalism (Gibson-Graham, 2006; Tsing, 2015) and work that acknowledges the different capitalisms (Miller, 1997), our aim is to provide a different account of the political economy of data and the place of the Global South in this configuration.

The processes of specialization, localization, and modularization are a demonstration of what such a different account might look like. Although there are examples of these in other regions of the world, they are by no means exhaustive in that they do not attempt to cover the total diversity of data practices and they are also not complete accounts. Our examples provide evidence of the complexities of the emerging political economy of data and identify potential starting points for an alternative account of the dynamics between the Global North and the Global South as well as spaces across and in between the Global South. They also offer a different perspective on flows of data, people, information, and technology and incite scholars to explore what, for example, it would mean to trace the global flows of data and value from the offices of Blue Sky Analytics and its production of modular data products that companies across the globe are coming to rely on as climate instabilities wreak increasing economic havoc. How might accounts of global interconnection and disenfranchisement need to be reworked if we traced the circulations of data and value that Stone Technologies puts in motion, fueling a resolutely local mode of accumulation? And how might starting with the specialized practices of Flying Labs force us to rethink that seemingly unerring link between data and accumulation, offering us a glimpse of a rather different relationship between data and value? These practices are not radical alternatives to digital capitalism and its economies of data; rather, they inhabit this context. Such alternatives offer fresh starting points for narrating these economies and the ways that they are affecting different places and people around the world. These are important not because they move away from questions of power and inequality but precisely because they allow us to begin to account for the ways in which inequalities emerge from the diverse practices that create, shape, and transform economies of data.