‘The tears don’t give you funding’: data neocolonialism in development in the Global South

Abstract This paper examines the knowledge politics and cultures that shape data relations between international aid organisations and Global South public institutions, taking African libraries as an example. International organisations increasingly rely on data from the Global South, purportedly as a resource for development, which has raised valid concerns about the emergence of new practices of data colonialism. One proposed solution is to expand the capacity of Global South institutions to control their own data processes, so they can likewise control the politico-economic relationships that draw on their data. A pan-African library organisation representing 34 countries is exploring this possibility though a multiyear research project to increase library capacity to use data to partner with development aid organisations. However, this work revealed that data colonialism precedes practices of value extraction. In focus groups, a survey of library systems and interviews with aid organisations, aspects of the data cycle are epistemically framed by aid organisations to undercut Global South control, and subtle neocolonial mechanisms encourage libraries to shape their own data cultures according to desires of aid organisations. This underscores the need to expand data neocolonialism as a frame for confronting epistemic injustice by highlighting Western rationalities embedded in data relations.


Introduction
The old adage 'knowledge is power' could be updated for our globalised economy to say that 'data are power' . The practices surrounding the collection and use of data have become a key site of production for both knowledge and the power that comes with it. This paper considers the knowledge politics surrounding how international aid organisations -often based in the Global North -use data from the Global South for the purposes of development. International development 1 is increasingly data-driven, and international aid organisations commonly rely on data from the Global South as a resource for development (Mann 2018;Taylor and Broeders 2015). Public institutions involved in development efforts, such as African library systems, increasingly view the collection and sharing of data about themselves as a potential tool in getting aid organisations to recognise them as credible partners, and therefore as potential recipients for funding. However, this arrangement has raised concerns about the emergence of new practices of extractive data colonialism, or data neocolonialism, which expose both data and their human subjects to ongoing exploitation in the service of development organisation growth. While current resistance towards these dynamics emphasises local or national control over data, such as data sovereignty, these framings do not account for the more subtle, epistemic ways aid organisations undermine Global South control over data. These processes mirror phenomena of digital neocolonialism (Mouton and Burns 2021) and warrant critical attention in the wider pursuit of epistemic justice.
This article is situated within critical data studies (Dalton and Thatcher 2014) and, accordingly, questions the emergent data relations (Couldry and Mejias 2019) between international aid organisations and public institutions of the Global South. Specifically, this paper examines these dynamics in the context of emerging relationships between African library systems and international aid organisations based in the Global North. 2 It does so by re-tracing the findings of a multiyear research effort that sought to increase the capacity of library systems to collect and share data to attract development funding and support from international aid organisations. That project worked with library systems to develop strategies to increase their data capacity, so that they could use data to better achieve their own development goals. The project developed strategies and tools to support our partners in making visible their own data, and published papers describing data strategies for local development efforts (eg Lynch et al. 2022a(eg Lynch et al. , 2022b(eg Lynch et al. , 2021a(eg Lynch et al. , 2021bYoung et al. 2021aYoung et al. , 2021b. As this work progressed, the project team identified moments where they felt that their 'data capacity building' goals lay in tension with the actual political empowerment of library systems. We found that the data requirements suggested by international aid organisations had subtle neocolonial implications despite being framed by objective, technical discourses. This paper steps back from the original goals of our project and questions the ways our capacity-building efforts may have simultaneously reproduced neocolonial relations. We ask, In what ways did technical suggestions for building data capacity within African library systems produce openings for data neocolonialism? We find that data neocolonialism operated via recommendations related to two key aspects of the data cycle, data validation and data sharing, which exercise epistemic power and encourage libraries to shape their own data cultures according to the desires of international aid organisations. Notably, these practices precede the use of data by aid organisations, and therefore point to ways in which data colonialism precedes the extraction of value from data. This hints at subtle aspects of data neocolonialism that remain undertheorised within the existing literature. While the rationale behind these practices can be taken as purely technical or common sense, we take a closer look at how the epistemic framing benefits aid organisations, undergirded by politico-economic structures of neoliberalism that open Global South data to exploitation.

Development, data and neocolonialism
From its inception, development funding has been linked to the expansion of colonial power. Originally employed as a mechanism for improving Global South infrastructure to facilitate extraction of resources, development funding took on a more geopolitical dimension post-World War II as the uS and Soviet union leveraged financial and technical assistance to compete for world influence (Kanbur 2006). Scholars have critiqued the concept of development as an 'historically produced discourse' (Escobar 1995, 6) that valorises ethnocentric, Enlightenment-era rationalities and the implementation of economic 'modernisation' , including through neoliberal free market policies that benefit formerly colonial powers (Escobar 1995). By normalising these capitalist values (Young 2016) and the 'White gaze' that undergirds them (Pailey 2020), development rhetoric and practice maintain and expand neocolonial power.
These processes had considerable implications for public institutions of the Global South, including the African libraries highlighted in this paper. The era immediately following the fall of colonialism in Africa saw a remarkable influx of development funding for libraries, driven by the belief that libraries could 'provide information to development agents and agencies' and support the education of the masses in new democracies (Mostert 2001, n.p.). However, shifts in the 1970s ushered in neoliberal reforms to development funding, including structural adjustment policies which discouraged funding for public institutions such as libraries (Bouri 1994). International aid organisations, funded by foreign governments and corporate and private donors, have rushed to fill the gaps in state funding according to their own disparate agendas (Young 2016, 53). There is now a much higher expectation of neoliberal competition for funds on the part of Global South organisations (Lynch et al. 2022b). Many public institutions have thereby been locked in a new cycle of dependency, but one in which they feel the need for visibility to garner the support of international aid organisations and private funders (eg Biruk 2018; Halkort 2019).
The use of data has become a common approach for attracting the attention of Global North funders, reflecting a broader shift within the field of international development to become more data-driven (Lynch et al. 2022b;Taylor and Broeders 2015). Broad frameworks like the united Nations Sustainable Development Goals (SDGs) have sought to make development more easily measurable, spurring data collection efforts. Smaller initiatives such as Open Data for Development (OD4D) maintain that issues associated with poverty in low-and middle-income countries (LMICs) are simply the result of 'information gaps' , and that non-governmental organisations and private corporations can solve these issues cost-effectively if given the right data (Mann 2018;Taylor and Broeders 2015). International aid organisations based in the Global North are increasingly reliant on Global South data to drive their projects and decisions. There have been calls for Global South organisations to increase their capacity to provide these data (uN 2015). However, this reliance has raised concerns about the emergence of new practices of extraction that mirror processes of colonialism, highlighting the enduring entanglement of development and colonialism (Couldry and Mejias 2019;Thatcher, O'Sullivan, and Mahmoudi 2016). Couldry and Mejias (2019, 337) characterise data colonialism as an emergent combination of 'the predatory extractive practices of historical colonialism with the abstract quantification methods of computing' such that individuals are exploited through data for profit. Social media platforms and other Internet-based services, often employed on increasingly ubiquitous mobile devices, encourage users to constantly capture data about their daily lives and then give those data away to corporations and international aid organisations. Processes of datafication -or transformation of lived experience into data -are often so opaque or imperceptible that users do not even realise that they are occurring. Mouton and Burns (2021) argue that this subtlety is such a defining feature of data extraction processes that data neocolonialism offers a more accurate label. In contrast to the visible political violence that characterises colonialism, neocolonialism describes the political, economic and socio-cultural structures that continued to normalise the dominance of former colonial powers even after colonies had achieved political independence (Nkrumah 1966). Mouton and Burns (2021, 3) argue that neocolonialism better describes the 'more indirect, more diffuse, perhaps subtler than outright coercion' processes that characterise extractive data practices. Couldry and Mejias (2019, 337) argue for an expanded analysis of emerging data relations, or the 'new types of human relations which enable the extraction of data for commodification' . These relations are shaped by broader epistemic, socio-cultural and politico-economic dynamics. Critical scholars remind us that data are themselves socially constructed and shaped by local contexts of knowledge making (Biruk 2018; Dalton and Thatcher 2014;Lynch et al. 2021a). Data bring with them the cultures that surround them -what Biruk (2018, 5) refers to as data culture, or the 'practices, rhetoric, and relations informed by epistemic conventions that underlie what … good clean data [are] supposed to be' . These data cultures are implicitly built into digital platforms used to collect and share data (Young 2019). Although Western data cultures can be an imprecise term, scholars (eg Burns 2014; Thatcher, O'Sullivan, and Mahmoudi 2016) have illuminated how beliefs and practices around data often align with Eurocentric, Enlightenment-era philosophies such as Cartesian-based measures of accuracy, or units of information that are visible and easily categorised and quantified. These rationalities stand in contrast to other ways of knowing.
Politico-economic discourses are often employed to invisibilise the cultural biases of data collection, and instead represent data collection as producing universal benefit. Couldry and Mejias (2019) describe how data extraction relies on subtle development discourses and epistemologies. Data are framed as a naturally occurring and neutral resource that are available to be appropriated by organisations 'for supposed benefit to wider society' (Couldry and Mejias 2019, 339-340). While the data have value, the process of their extraction is constructed as value-less, invisibilised as 'just sharing' , or naturalised as necessary in order to translate social life into 'quantifiable' data of value. Any labour that Global South individuals and organisations put into data collection is often invisibilised, and any risks of disclosure are overlooked or justified by a need for visibility (Biruk 2018;Halkort 2019;Mann 2018;Omanga and Mainye 2019). Furthermore, these data are ideally open, '"unlocked" from organisations based in Africa and provided to experts based in advanced economies' who then analyse them and generate profit (Mann 2018, 9). This highlights a complex, but often subtle, set of discourses and practices that seek to motivate Global South organisations to produce and share data, even when those data are framed in ways that primarily benefit the goals of organisations situated in the Global North. This results in hierarchical data relations that are neocolonial in nature.
In opposition to these dynamics, a 'decolonial turn' within academic studies of data and technology is increasingly exploring approaches to expand the capacity of Global South communities and organisations to control how their data are produced, owned and used (eg Couldry and Mejias 2021; Milan and Trere 2019). Outside of academia, politicians in the Global South have similarly attempted to bolster national control over data. Vinit Goenka, an Indian politician, popularised the term data sovereignty in 2014 as an explicit attempt to counter new forms of data colonialism (Vila Seoane 2021). The term data sovereignty has since been adopted by other marginalised groups, including many Indigenous Peoples (eg Carroll et al. 2020;Kukutai and Taylor 2016;Te Mana Raraunga n.d.), pursuing similar levels of control over their data. In this context, Snipp (2016) argues that data sovereignty is defined by the power of Indigenous Peoples to (1) define who is counted within data about their communities; (2) ensure that data about their communities reflect their own interests and priorities; and (3) determine who has access to Indigenous data. He argues that these factors are critical to support data decolonisation and, more broadly, Indigenous self-determination.
Despite increasing popularity of the concept, more research is needed to understand the ways in which data sovereignty is delimited by complex interactions between local and external data cultures. Mouton and Burns (2021) note that these resistance strategies often rely too heavily on the choices of informed individuals to 'opt out' of data tracking processes set up by external actors and platforms. This leaves intact the neocolonial relations that structure these common data relations and platforms, and often forces individuals to opt out of the benefits of data collection. National efforts have similarly had to negotiate contradictory effects of data sovereignty. In India, Goenka's efforts have had mixed long-term results, spurring legislation that has been attacked for having negative impacts for business and Internet freedom (Vila Seoane 2021). Wainwright (2008) argues that these tensions are intrinsic characteristics of any project that attempts to harness development for local empowerment. He believes that development projects are defined by aporias, or impasses that are nearly impossible to resolve. He describes aporias in the context of countermapping projects in Belize and Nicaragua, in which Indigenous communities have created their own decolonial maps to assert territorial control over their lands and development (Wainwright and Bryan 2009). He concludes that these maps rework, rather than reverse, existing power relations because they are unable to disrupt the broader economic and governance systems that perpetuate colonialism. Countermapping simultaneously, and contradictorily, advances Indigenous rights to land even as it solidifies the politico-economic systems that denied Indigenous connections to the land in the first place. Nevertheless, Wainwright (2008) believes that these aporias may provide important launching points for more fundamental critiques of the larger politico-economic systems that drive colonial outcomes.
Our own project expresses many of these same aporias. On the one hand, it has been motivated by a desire expressed by library systems to expand their data capacity. On the other hand, we have found that the data preferences held by aid organisations subtly but powerfully undercut the political control that library systems might have over their data. These organisations consistently framed seemingly technical aspects of the data cycle -including characteristics of data validation and sharing -in ways that encourage Global South institutions to open data to outside control. This has produced tensions between the 'data capacity building' and 'political empowerment' components of the project -precisely what Wainwright might describe as an aporia. Libraries broadly want, and even need, the material benefits resulting from development aid, and yet the steps they must take to attract that aid subject them to new forms of colonialism. In this case, though, it is the technical practices of data collection -rather than mapping -that support that aporia. This paper identifies the technical mechanisms that help to set up this aporia. We argue that these mechanisms are employed as forms of neocolonial power to encourage African libraries to shape their own data cultures according to the desires of aid organisations. This exposes limits to data capacity building and sovereignty as standalone tools for resisting data neocolonialism. By identifying these mechanisms, we hope to provide opportunities to see through their technical nature so that scholars and practitioners can engage in the deeper work to transform the colonial values that undergird them.

Methodology
This project reflects on and critically analyses a collaborative project called Advancing Library Visibility in Africa (ALVA). This project emerged from a series of stakeholder meetings and workshops that involved representatives from library systems from across the African continent. These meetings also involved organisations that had a history of partnering with library stakeholders, including the uS university-based research group the university of Washington (uW) Technology & Social Change Group (TASCHA). One consensus emerging from these meetings was that there was a need for stronger continent-wide advocacy to support library systems to advance the development of the countries and communities that they serve. These meetings led to the creation of a pan-African library support organisation, African Library & Information Associations & Institutions (AfLIA), that currently represents 34 member countries across the continent (see Figure 1). An additional set of stakeholder workshops identified a need to develop the capacity of AfLIA, and African library systems generally, to collect and use data in their advocacy efforts. Based on this need, AfLIA and TASCHA collaboratively developed a proposal that was submitted to, and then funded by, the Bill & Melinda Gates Foundation.
The goal of the resulting project is to increase the capacity of African library systems to use data to attract and maintain partnerships with development aid organisations, to increase the visibility of libraries as contributors to development. To achieve those goals, AfLIA and TASCHA carried out a series of activities designed to (1) describe the existing data collection practices and capacity of library systems across the African continent, (2) identify the types of data that development aid organisations find persuasive when making partnership or funding decisions, and (3) develop tools and practices that can overcome any gaps between the types of data libraries are able to collect and the types of data aid organisations want. These activities were approved by uW's Human Subjects Division, and they followed appropriate informed consent and other ethical procedures. The activities are detailed below.
In February 2018, the research team conducted focus groups as part of a three-day workshop in Accra, Ghana, focused on participants' experiences with and knowledge of data within their library systems ). The participants were 28 stakeholders in the library sectors of 16 countries across Africa. Participants were selected through purposive sampling as experts to represent the library systems of their countries and included members of national library systems, national library associations, and organisations identified by the research team as influential within the library sectors of their countries. Focus groups consisted of four or five individuals organised by the research team for a mix of regional perspectives and genders within each group. The groups participated in activities including written responses to questions posed during a group activity; written responses to worksheets that asked questions about participant experiences with data; transcripts of focus group discussions; and written observation notes taken by the research team during focus group discussions. All sessions were conducted in English, with French and Portuguese interpretation also offered. Written responses were primarily in English, and those written in French or Portuguese were translated. Responses from focus group activities are marked with FG and the group number below (eg FG1).
Based on this workshop, the project team developed an initial prototype of a mapbased data collection platform Young et al. 2021a). This prototype was designed to collect only the most basic information about libraries, including their location, organisational information, and contact information. It was designed to enable the crowdsourcing of this information from African libraries themselves, and it was tested for usability with 120 participants from one of AfLIA's library training programmes. The project has utilised Champions in 27 countries to support librarians through the crowdsourcing process. In October 2019, these Champions were invited to participate in a workshop held in Accra to discuss their experiences, which has shaped the research team's general understanding of the constraints and possibilities in this work. Results of this workshop are not directly referenced in this paper's discussion but are included here to provide extra context.
Between September 2019 and January 2020, the team conducted interviews with 20 development practitioners based in North America (13), Europe (one), and Africa (six) who work for well-known organisations that fund or implement projects in Africa . Our inclusion criteria for these practitioners were left open to account for the varying definitions and debates around the term 'development' and to elicit perspectives across a wide range of actors . Participants were solicited through a combination of purposive and snowball sampling. Their organisations comprise bilateral, multilateral and non-governmental organisations including private foundations and faithbased organisations. Interviews were semi-structured, and questions focused on the participant's professional background, their organisation and how it selects local partners, their perceptions of libraries, and what types of data could be effective in shifting perceptions. Interviews were transcribed, anonymised, and analysed for initial themes by two researchers to triangulate findings. Responses from interviews are marked with an I and the participant number below (eg I1).
From June to November 2020, the team also conducted a survey of 29 public library systems 3 in 19 African countries (Lynch et al. 2021c). The aim was to gain a broad understanding of the data life cycles within these library systems, from data collection to use and analysis. Participants were invited to respond to the survey based on their roles within African national or provincial/federal/state libraries. These individuals, however, were permitted to delegate responses to the survey to other members of their staff. This ensured that the individuals with appropriate institutional knowledge responded to survey questions. This means that each survey response could be taken to reflect the data practices of a network of libraries that the national or provincial/federal/state libraries directly manage. The survey was administered electronically and had five sections asking about what data are collected by the library system; how the data are used by the library system; how data are shared; how data are involved in decision-making; and the experience of data collection. Respondents chose to take the survey in Arabic, English, French, or Portuguese. For analysis, open-ended responses in Arabic, French, or Portuguese were translated to English, and closed-ended questions were analysed using basic descriptive statistics. Responses from the survey are marked with an S and the response number below (eg S1).
Across these activities, the research team used an expert sampling method to include broad representation of library systems from across the continent. AfLIA led the identification of participants, with care taken to include representation of all African regions (southern, eastern, western and central Africa), the major colonial languages (English, French, Portuguese and Arabic -although we were unable to represent Arabic at all events), and national library system types (centralised and decentralised) within each activity. Countries involved across these activities included Botswana, Burundi, Burkina Faso, Cameroon, Côte D'Ivoire, Egypt, eSwatini, Ethiopia, Ghana, Guinea Bissau, Kenya, Lesotho, Madagascar, Malawi, Mali, Mauritius, Namibia, Nigeria, Senegal, Seychelles, Sierra Leone, South Africa, Tanzania, uganda, Zambia and Zimbabwe. The present analysis is an attempt to draw these activities into conversation with one another. This analysis was sparked by noticing emerging lines of tension between our studies of library systems and aid organisations which seemed to form around issues of data validation and data sharing. Data were analysed inductively and recursively using a grounded theory approach, and the researchers triangulated insights across the studies (Glaser and Strauss 1967). As research questions, we ask, In what ways did technical suggestions for building data capacity within African library systems produce openings for data neocolonialism? This reflective analysis, and the writing of this paper, was performed collaboratively by the same AfLIA and TASCHA researchers that carried out the activities described above. The following section details our findings and discusses their implications for conceptualising data neocolonialism in the Global South.

Data practices as a site of power
These studies elucidate lines of tension between the data cultures of international aid organisations and African library systems within two aspects of the data cycle: (1) data validation and (2) data sharing. Analysis shows how particular data epistemologies are weaponised by aid organisations to produce forms of control that raise concerns about fundamental extractive rationalities in the international development field. This highlights subtle epistemic mechanisms in play as Global South institutions shape their own data cultures according to the desires of international aid organisations.

Data validation
Practices and standards of data validation are a key mechanism by which international aid organisations shift control over Global South data from local governmental structures to aid organisations themselves. Aid organisations wield statistical measures of validity to leverage control over funding and extract value from the data of their partners such as African libraries.
Our research suggests that data validity in African library system data cultures has an integral social component whereby a data set's integrity is established via hierarchy or relationships. For example, multiple focus groups discussed the importance of acquiring data from someone of 'official' or 'mandated' status, which establishes the data's credibility (FG2,4,5). One participant discussed a case in which the head of a library contacted them retroactively to say that data given to the participant by the staff of their library were incorrect; the participant reflected, 'Maybe if it was reported directly, [the head of library] would have doctored it a little, massaged it, but the staff just put it out […] the lesson we learned was, you know, you need to be sure who you are dealing with, who is giving the information' (FG5). This anecdote illustrates social notions of credibility and the significance of the relationship between the data and the person giving them in order for the data to be deemed acceptable. Other responses suggested collective measures of validity, as when joint responsibility for data quality is given to 'The informants of the group and community leaders' as well as 'The Council and Traditional rulers and librarians' (FG4). Survey responses echoed the theme of hierarchy as a prevalent form of data validation. Most respondents (59%) described the data cycle of their library system as hierarchical in nature, whereby data is collected by library branches on a regular schedule and passed up to a higher authority who compiles, analyses and submits the data to higher levels, such as a board or ministry. When asked to describe their process of data verification, many respondents mentioned gaining the approval of a higher authority -for example, a senior manager who 'verifies whether the information on the report is correct' before sending it to the library director who 'verifies all reports received' before sending to the national department (S26). These findings across our studies point to relational forms of validity as key to establishing the integrity of data in context. They also implicitly confer power to the social or organisational structure in which data are produced, since it is those structures that mark data as valuable.
These social approaches contrast strongly with the preferences of development aid organisations. Interviews with development practitioners widely emphasised the importance of statistical measures for data to demonstrate accountability. When asked what types of data aid organisations would need to consider African libraries as potential partners, practitioners commonly expressed a desire for quantitative data on library usage and user demographics. While qualitative data were also seen as useful, 'the quantitative allows [development organisations] to sell and to demonstrate undeniable success' (I5) to their own funders. Data are often made valid through randomised control trials (I1) or analysed in terms of economic returns (I5, 11, 12), such as 'If I invest in the library, I will reduce inequality by X percent or increase jobs by X percent or improve livelihoods by X percent' (I11). More importantly, quantitative data are also used to assess the integrity of partners themselves through monitoring and evaluation (M&E). Many interviews emphasised how 'rigorous' and 'strict' (I13) and 'serious on reporting these metrics' (I19) organisations are for the purposes of partner accountability, with 'checks and balances […] to check up on you if you try to lie' (I15). Accountability was a major theme, with the desire and expectation that a partner monitor themselves with 'statistics' , often to appease specifically 'Western' donors (I19) to demonstrate that they 'can do the work' (I17) and 'deliver quality work on time' (I14). Interviewees stressed that their organisations 'really have to be numbers focused' for 'reaching indicators' (I13) in order to satisfy donors such as 'ex-tech funders' (I11) who emphasise accountability through data verified through statistical means. In this way, a focus on numbers and accountability furthers the interests of funders who want protect themselves from risk.
Herein lies a clash over what is viewed as valid and useful data, with implications for the power dynamics between aid organisations and African library systems. The social validation approach adopted by libraries ensures that African organisations -most often, national library systems -have the authority to control what data have value and, therefore, the power to impact decision-making processes. The emphasis placed on statistical validity by aid organisations severs that relationship between the library organisation and the validity of data. This is often framed as a purely epistemic issue -statistical validity is grounded in Western scientific principles that ensure data better represent truth. However, this epistemic claim also has the effect of increasing the power of Global North donors, by emphasising quantitative data that easily feeds into existing neoliberal logics of accountability. Interestingly, some development organisation interviewees conceded that the emphasis on quantitative data could get in the way of developing an accurate understanding of local realities, but that the preferences of the donors overrode those concerns. Participant I13, for instance, said that their local team 'thinks more qualitatively about the impacts of their [local] partners' , but ultimately they 'have to be numbers focused and so we do put some of that onto the grantees as well because we really are reaching our indicators through our partners' . Other participants similarly lamented the emphasis on quantitative data, but chose to 'blame the donors' (I5) without making material changes in their data validation practices. These discussions show how statistical validity is described as an epistemic mechanism, even as it is used primarily to transfer political power to donors.
ultimately, though, African stakeholders expressed the need to conform to these preferences, indicating that they were aware that numerical data had become integral to their own financial survival. As asserted in one focus group, 'Statistics are really vital to our existence … they don't fund anything emotional. You need evidence' , or more bluntly, 'The tears don't give you funding' (FG5). While social measures of data integrity may be important locally, they are no match for a statistically valid randomised control trial in the eyes of a development organisation looking to fund a reliable investment. Interviews further emphasised how numerical data and assessments allow a partner to establish a 'track record' (I14, 15) which makes aid organisations more likely to partner with them. This benefit incentivises public institutions such as libraries to adopt the accompanying 'habits, investments, and standards' (Biruk 2018, 3). As a result, it is the institutions or individuals within them who are responsible for navigating the logistical and social tensions to adapt to the desires of aid organisations, including added labour of data collection, analysis and reporting (Omanga and Mainye 2019). African libraries are already struggling to collect data of high integrity, due to a combination of social and logistical challenges exacerbated by digital divides , limiting any potential capacity for resistance. The responsibility placed on institutions to perpetually strive to bridge Global North expectations and Global South realities thus extracts value from institutions in the form of their own labour to monitor themselves and protect Global North investments. The framing of data validity, or 'practices, rhetoric, and relations informed by epistemic conventions' (Biruk 2018, 5) that wield statistics as a form of neoliberal control, further cedes power to aid organisations and their statistical regimes of truth, an epistemological harbinger of neocolonialism, discussed below.

Data sharing
The data sharing practices and norms championed by development aid organisations are an additional mechanism that shifts control over data from Global South institutions to aid organisations. While aid organisations advocate for data openness and sharing, African library stakeholders are reticent to embrace these norms due to political and social risks associated with open data. Nevertheless, we found that the libraries feel bound by the rhetoric of visibility to share their data to access international support under neoliberal funding schemes.
Development practitioners valorised the sharing of data as mutually beneficial for African libraries and aid organisations. Interviewees emphasised how sharing data, particularly quantitative data, allows a potential partner to demonstrate accountability and thus increase their attractiveness to aid organisations. Data help partners in marketing, or 'putting [themselves] out there' (I10) to 'flash those skills, expertise, and potential in front of folks and shops like us' (I14), stressed as necessary for any partner looking for funding. This rhetoric of visibility is also embedded in our own project, ALVA, rooted in the assumption that visibility to international aid organisations is both good and necessary for the sustainability of African libraries, and that data are the key to achieving this visibility. Many interviewees took this one step further to suggest that their organisations would be interested in African libraries as a source of 'data for decision-making' (I18) in development. They were interested in partnering with libraries to acquire data that is otherwise unobtainable for them, or 'data that we would not otherwise have access to or know existed' (I18). These include unique, localised data that are 'different and/or higher quality' (I20) than data from other partners such as national governments, or data gained through access to 'marginalised' and thus 'targeted' populations (I13) such as women or youth gathered at the library. In this way, data shared by partners helps aid organisations achieve their own goals and expand their reach. Ethics of data sharing went largely unmentioned in interviews, with the only exceptions being 'digital privacy' mentioned in passing (I11) and concern over 'larger funders' having 'agreements' with mobile service providers to supply geospatial data that are 'managed and led by the [Global] North over the [Global] South' , resulting in a power imbalance (I20). In this case, however, the interviewee used this example to advocate for making such data 'popular' and widely available to regular citizens or government bodies, again emphasising open data sharing as an ideal.
African library stakeholders, however, are hesitant to widely and publicly share data. Survey responses suggest that a majority of library systems do not provide data openly. Only 34% of respondents indicated that data is uploaded to open portals or websites, and only 45% make data available to the public. Instead, data is more available to official entities such as ministries (86% of respondents) and library staff (76% of respondents). When asked about relative risks and benefits their library system would incur from sharing data it collects more widely, open-ended survey responses revealed a rich discussion around attitudes towards open data. Perceived benefits include those for libraries, with 'increasing visibility' (S27) and partnerships as main themes. For example, one respondent explains, 'One of the benefits is greater visibility of the [library system]. Other potential donors who are willing to support the [library system] would not have problems accessing data. Potential partners would use the information to support the [library system]' (S28). Respondents also saw benefits for library users, including increased awareness and usage of services to 'build a better image of the library' (S21) among the public, and enhanced ability to advocate for funding to expand library services (S6).
However, stakeholders also expressed that sharing data comes with real risks for African library systems, offering key insight into local data cultures and emerging issues of privacy. Survey responses revealed fear of data misuse, if data should fall into 'the wrong hands and [be] use [d] for the wrong purpose' (S10) such as 'hacking' (S26), 'discrediting' by 'opponents' (S11), or 'abuse of data by some entities for selfish gain' (S22). Others fear 'misinterpretation' (S1, 22) of data if they are made public, or 'interpretation of the data out of context to harm the [library] and create confusion in the minds of the public in general, and more particularly our [stakeholders]' (S21). These responses and others highlight the political risks of using data for visibility as well; in one potent example, a respondent states, 'Resources are very scarce. If we are able to attract help from an organization that is politically affiliated to an opposition party, the ruling party may blacklist us' (S9). This sentiment was echoed in focus groups where participants spoke of 'political barriers' to funding, in the sense that libraries need to be 'impartial' and unbiased towards a particular political group in order to secure support (FG2). Additionally, privacy was a recurring concern in making library data widely available, with mentions of 'confidentiality' (S11, 12) and 'privacy breaches' (S11). Respondents also acknowledged the impact of these issues on library patrons, with the awareness that 'as cybercrime rates soar, sharing data can potentially lead to devastating repercussions' (S7). These comments illustrate why institutions such as African library systems may be reticent about embracing open data policies and sharing data with outside organisations, given the complexities of navigating both local and global concerns.
The push to share data with international aid organisations and the pull of privacy concerns within African library systems creates another line of tension. How can public institutions and populations in the Global South gain financial support without losing control of how their data are used? There is a degree of disclosure in making institutional data available to organisations where there is not yet a relationship of trust, for the purposes of attracting potential partnerships. This disclosure is uncritically expected of institutions as a form of marketing and maintaining partnerships, but local data cultures and multilayered attitudes towards data sharing contour the complexities of what constitutes fair use. However, the rhetoric of visibility is also baked into advocacy efforts. Even if a Global South institution can maintain a form of ownership over their data, they are stuck in a bind where they must share those data to get support, and this unspoken mandate raises tangible concerns about how the data will be used. To borrow from Halkort's (2019, 322-323) discussion of refugees' data sharing, this dynamic intimates a dependency which has 'locked [institutions] in an environment of compulsory data sharing in which their ability to raise their voice and influence critical decisions remains inextricably bound to the imparting of vital knowledge that may eventually be used against them' . The consequences of data misuse may be disproportionate for the socially and politically vulnerable, and as our research affirms, 'efforts to make African societies more legible to corporations and humanitarian bodies also make opposition groups more visible to regimes' (Mann 2018, 5). These concerns about the risks of visibility may be shared among public institutions in the Global South, where neoliberal austerity measures leave public institutions particularly susceptible to budget shortfalls and political whims. As institutions such as African libraries seek support internationally, they are left to contend with the local consequences of data disclosure. Development organisation interest in local, exclusive data, particularly in relation to marginalised people, also raises the spectre of data extraction to benefit aid organisations and, thus, underlying neocolonial patterns of resource exploitation.

Data neocolonialism in development
Two forms of data practices recommended for libraries by our project -data validation and data sharing -can enact mechanisms of control whereby international aid organisations undermine Global South control of data and open them to potential exploitation. By looking more closely at these practices, this analysis adds heft to existing descriptions of the extractive rationalities of data neocolonialism by illustrating their embeddedness in contemporary international development. However, it also extends thinking about epistemic injustice by highlighting the Western epistemologies which undergird these data practices.
Overall, the data practices emphasised to us by development aid organisations point to fundamental extractive rationalities which mirror the extraction of 'raw' resources under colonialism and treat data as a resource for development organisation gain (Couldry and Mejias 2019;Thatcher, O'Sullivan, and Mahmoudi 2016). These data represent larger technocratic efforts to quantify social life to derive value for 'supposed benefit to wider society' (Couldry and Mejias 2019, 339) -in this case, funding for development. These practices serve aid organisations by extracting data that benefit these same organisations in various ways, including reassuring Global North-based donors, reaching larger numbers of Global South individuals and (higher value) target populations, and building a 'track record' of metrics, all of which increase their chances in competitive funding cycles and spur continuous growth. These benefits reflect the underlying belief and practice of data as a valuable resource (Thatcher, O'Sullivan, and Mahmoudi 2016) which is used to bolster the economic expansion of the Global North-dominated development industry led by former colonial powers (Escobar 1995). While most aid organisations seem to treat data as 'just there' for the taking (Couldry and Mejias 2019, 339), our research shows that this is not the case in practice. In the African library system context, data are embedded in relationships, and their extraction requires the local labour (and local risk) of frontline library staff to engage in collection, validation and sharing. This work has echoes of neoliberal responsibilisation, 4 in that international aid organisations are attempting to inculcate library staff into self-monitoring, a burden 'put onto the grantees' (I13) so that the development organisation can achieve success through their partners. There is cruel irony in the fact that the same neoliberal rhetoric that emboldened aid organisations to structurally dismantle funding for public institutions now expects those institutions to prove themselves worthy of funding, in the wake of decades of chronic underinvestment. This exploitation of Global South data and labour is in effect accumulation by dispossession (Thatcher, O'Sullivan, and Mahmoudi 2016). While institutions such as African library systems have undoubtedly benefitted from development funding in the past, this money often has strings attached and may not align with local agendas (Omanga and Mainye 2019). Furthermore, the labour takes a disproportionate toll on vulnerable populations who are subjects of and/or collectors of data, leaving them feeling exploited or unfairly compensated (Biruk 2018).
These data practices are informed by the epistemologies of the development aid organisations that promote them, mirroring neocolonial processes by which the socio-cultural values of former colonial powers continue to normalise their dominance (Nkrumah 1966). The ways in which African library systems are effectively compelled to shift their data cultures accordingly raises concerns about epistemic injustice as a more subtle yet sinister aspect of neocolonial power enacted through data (Mouton and Burns 2021). Through data validation and data sharing practices, the dominance of former colonial powers continues in terms of their power to determine what is true and valid. Data sharing practices reinforce Western ideals of data that are visible, quantifiable and declared 'accurate' according to Cartesian measures that emphasise objective distance (Burns 2014;Thatcher, O'Sullivan, and Mahmoudi 2016), as opposed to relational forms of validity or deliberate opacity to protect data subjects or sharing institutions from harm, as highlighted by African library systems. Presenting these approaches as mutually beneficial and universally applicable depoliticises the epistemic power in play. A critical data studies lens pushes back against the idea of data as inherently objective or neutral (Dalton and Thatcher 2014) in favour of an approach that considers the 'social lives of numbers' (Biruk 2018, 3) and how they can be used to exercise power. Likewise, a critical data studies approach moves beyond discourses of development as technical support to critically consider the values underlying such practices. Biruk (2018) reminds us that colonial attitudes and motives have fuelled data collection and practice in sub-Saharan Africa (23) with an enduring link between measuring or counting populations and 'those who seek to govern/control them' (20). Pailey (2020) sees this emphasis on monitoring as a manifestation of the unacknowledged Whiteness in development. When data practices are uncritically promoted and taken up, the question of whose truth matters is no longer on the table, and, in fact, is rendered irrelevant by the sense of urgency and scarcity that permeates the contemporary, globalised economy. The precarity experienced by public institutions is undergirded by the political economy of a neoliberal system or, as expressed by Halkort (2019, 232), 'the increasing pressure to maintain visibility and attention for fear of one day being rendered out of existence and invisible […] with no other choice but to submit to the daily and ongoing pressures to be mapped, calculated, and recorded' . Discourses of accountability and visibility serve to wrest power away from Global South institutions, in favour of submitting to evaluation by development organisations as a means of survival (Mann 2018). As a result, the epistemic power wielded by development organisations is reinforced by political and economic structures that continue to feed neocolonial power and epistemic injustice (Ndlovu-Gatsheni 2021;Nkrumah 1966).
Naturally, we have struggled with these critical reflections on our own work. Should we continue to provide recommendations that may simultaneously provide libraries with material benefits and produce new forms of data colonialism? In response to our own concerns, we are trying to offer stakeholders both sides of the story. We want them to have the information they need to better understand the benefits and risks of pursuing international aid, so that they have greater capacity not only to choose how to engage development organisations but also to choose whether they wish to do so in the first place. In many cases, librarians seem to have the same mixed feelings that our research team holds. In some instances they have agentively adopted colonial discourses related to data practices, while in others they are resisting those discourses. For example, we recent offered a webinar series in which we discussed project findings with key African library stakeholders (Lynch et al. 2021c). Webinar discussions yielded responses that affirmed many librarians felt responsibilised to attract the attention of international aid organisations with data. Webinar attendees stressed the need for mindset change or 'fighting spirit' , asserting that extra labour is simply necessary for libraries to gain access to development funding (AfLIA 2021a). We would emphasise that we are not criticising this stance -visibility for development funding can and does bring material transformation for the communities that these libraries serve. However, panellists and attendees also confirmed that they are worried about some of these trends. Privacy, for example, is an ongoing concern in data collection and usage for African libraries, with new national policies such as South Africa's Protection of Personal Information (POPI) Act changing local attitudes towards personal data (AfLIA 2021b). Panellists voiced the need to work with aid organisations to navigate these new and emerging concerns around data sharing. These responses highlight once again the aporiac nature of development. As African library systems seek more visibility and support from aid organisations, they may simultaneously be exposing their institutions and their patrons to mechanisms of control that remain ominously opaque.
This raises substantive questions for us around what it means to empower Indigenous and Global South communities to collect data that serve their interests and priorities. Our project highlights a politico-economic landscape in which Global South institutions are presented with a binary choice of either (1) remaining invisible to funders, such that they cannot attract the funding they need to survive or (2) collecting data that might allow them to attract funding while also further integrating them into systems of global neocolonialism. This can hardly be described as a choice that encourages any meaningful level of self-determination. Moreover, it is unclear how the many technical approaches to advancing data sovereignty -from expanding the data capacities of local communities to passing legislation that encourages data ownership and legal control -would help to address the structural conditions or disciplinary forms of power that shape this choice for Global South institutions. While data sovereignty offers Global South institutions the option to retain agentive power over their data collection and usage, the actual choices available to them remain shaped by the economic imperatives of neoliberalism and the afterlives of colonial power.
Public institutions remain an important threshold between their communities and private, international and corporate interests, and making decisions informed by critical analyses of data relations and their systemic implications is vital. We view this insight as increasingly important for our own work and partnerships with these institutions, meaning that our role as researchers must be to not only build data capacity but also to build the capacity of institutions to understand the political economy into which these data feed. As Adriansen and Madsen (2019, n.p.) argue in the context of their work on African higher education, 'we begin the process of decolonization by raising questions, because when involved in capacity building projects, the researcher might not have the power to change the dominant structures but does have the power to challenge and ask critical questions' . In raising the questions explored in this paper, we hope we can help African librarians make more strategic decisions about when and how to deploy data in pursuing their own goals. We also hope our research provides a model for others for drawing on data neocolonialism as a frame to engage emergent cross-cultural norms around data, beyond data sovereignty and towards a more integrated, epistemic approach to uprooting neocoloniality in data relations in international development.

Conclusion
As our research has shown, data relations between development institutions and public institutions of the Global South are at a crossroads, rooted in a colonial past and currently structured to ensure a neocolonial future. Data practices are a site of neocolonial power where development aid organisations are able to undercut local control over data and open them to exploitation via data validation and data sharing. While these practices are presented as norms and reinforced by neoliberal pushes for accountability and visibility, public institutions must be attentive to the subtle epistemic power that these data practices leverage. This analysis both affirms and extends data neocolonialism as a frame for critically analysing data relations in international development, with a needed focus on new ways to ensure epistemic justice for Global South institutions and their patrons. As Couldry and Mejias (2019, 345) note, neocoloniality 'can only be opposed effectively if it is attacked at its core: the underlying rationality that enables continuous appropriation to seem natural, necessary and somehow an enhancement of, not a violence to, human development' . Only the active dismantling of seemingly common-sense truths will open the path to alternative futures.