1 Introduction
During the months from December to April, a period of the year that is also known as the dry season, temperatures in the northern Ugandan city of Gulu often rise above forty degrees Celsius. On the campus of Gulu University, individual seminars are then moved outdoors. The cool shade of dense mango and acacia trees is often the only option for lecturers and students to escape the stifling heat of the university buildings covered with iron roofs. This kind of improvised teaching practice formed the contrasting background to the colourfully decorated and air-conditioned containers of Samasource. As I had learned shortly before, Samasource is a San Francisco-based company that specialises in producing accurate data by combining machine assisted annotation (MAA) with human validation for Machine Learning (ML) and Artificial Intelligence (AI). After passing the guarded doors, it felt as though I was entering a spaceship that was about to take off because of the constant whirring of the air-conditioner amplified by the noise of approximately 150 computer ventilators. The environment was intensified by the tightly closing windows of the containers, which not only muffled noise from outside, but also ensured that cool air would not escape the two hundred square metre room. âWe couldnât work here without A/C!,â explained William, the manager, âWith windows open, tonnes of dust would enter and quickly destroy the PC s. But also, excess heat from the desktops would make it unbearable. And of course, we all know that cooled bodies perform better â¦â1
The managerâs quick reflections on both the climatic peculiarities, and the insight that cooled bodies would achieve higher productivity than uncooled ones, again underlined the extreme contrast not only to the rest of the campus architecture, but also to the predominantly rural character of the rest of
Based on this ethnographic experience, in the following contribution I will elaborate on some conditions that I believe are central for the successful operation and general functioning of this comparatively young form of data factory work. I will argue that the (infra)structured connection to the World Wide Web and the access it facilitates to digital labour is based on a number of different but interacting disconnections. This chapter delineates the emerging socialtechnical tensions and asks which disconnections, points of access, and circulations of digital labour these tensions are based upon. Before offering further ethnographic details, the paper situates the conceptual discussions around âmicro-workâ by critically linking it to a broader debate centred around the concept of the digital divide (see Selwyn and Facer 2010, Selwyn 2004).
In the first empirical part of this paper, I will analyse the organisational effects and causes of the disconnections of Samasource operations from local realities, with emphasis on the infrastructural and material aspects. In the second part, examples will be presented, indicating how an extremely selective allocation of and access to digital resources represent a central operational condition for Samasource. In this context, software-supported control and standardisation regimes form an essential prerequisite for decoupling individual activities from relying on local knowledge and working practices. In the third and final part, a contrast is made between wage labour, which is based on strictly standardised measuring of labour, and the specific requirements of local labour cultures that are characteristic of low resource settings.
The aim of this paper is to make visible the specific conditions on which novel forms of globally circulating labour rely when they travel to places like the University of Gulu. To do so, it will be necessary to highlight organisational, infrastructural, social, and knowledge-based decouplings that contribute, to varying degrees, to how these new forms of digital labour overlap, and at times conflict with local work cultures. However, while digital evolutions in the urban centres of African countries, for example Nairobi, Accra, or Lagos have received growing academic attention, literature on the networking of peripheral rural regions is rather sparse; this paper is therefore an attempt to broaden the field of research in this regard.
2
The notion âhumans-as-a-serviceâ2 is used to describe all those activities in digital economies for which there are either no technical or automated solutions (yet) available, or for which it is not deemed economically worthwhile to ever develop a technical solution (Irani 2013, 11; Bergvall-Kareborn and Howcroft 2014). Content managers, for instance, who clean up platforms from violent images and hate speech represent a sore spot, an embarrassment to Silicon Valley solutionism3 that is not able to offer a technical fix for emerging ethical and political problems. Content work in the Philippines and California are often outsourced to other companies and is not granted the same payment nor the luxurious workplace conditions that Facebook or Google engineers enjoy. Humans-as-a-service demarcates the distinction between what Silicon Valley considers real work (building platforms) versus simple content work (Daub 2020). In a similar vein, big car companies like BMW or General Motors keep suspiciously silent when entering into contracts with big data factories in China and India for the production of large chunks of training data for their self-driving cars. But being silent about this heavy reliance on cheap human labour to make automation and AI work, is paralleled by a fear of diminishing and disenchanting public imaginaries of the technologyâs own intelligence (Gray and Suri 2019).
On the one hand, the spectrum of activities and tasks subsumed under the notion of humans-as-a-service is very broad and becomes increasingly difficult to compare. A few minutes of text transcription on Amazonâs âmechanical turkâ4 is very different from the psychological stress to which, for example, content moderators find themselves exposed when cleaning up platforms such as Facebook or YouTube from violent depictions and hate speeches (Roberts, 2019). On the other hand, critical social science studies are increasingly identifying common principles and patterns that underlie as well as emerge from
However, case studies that explore the current and future role and relevance of the internet for this new young field of globally circulating labour address, to varying degrees, the issue of unequal distribution of access to digital services and employment opportunities. One concept that attempts to capture the issue of unequal access is the notion of a growing digital divide (Mariscal 2005, Fuchs and Horak 2008). A central observation of this concept is that an increase in access to the internet and digital services is simultaneously accompanied by an increase in digital disconnectedness of large parts of the worldâs population (World Bank 2016). Despite the World Bankâs concerns regarding growing global inequalities, its framing has been criticized as a quest for yet another technical fix (Donner 2015, Shrum et al. 2007). If such a framing would be effective, a purely technical availability of the internet would have beneficial and accelerating impact on the economic development of a country or region.
An argument that this does not have to be case is provided by Louise Bezuidenhout et al. (Bezuidenhout et al. 2016, Bezuidenhout et al. 2017) showing that the mere availability and access to opensource and open-data science in the context of African universities depends on further requirements and skills in order to unfold positive impact for academic research. The question of when the internet can actually serve as a resource for an individual or a group to cope more efficiently and successfully with the hardships of everyday work is not simply solved by providing access. While I consider this perspective to be valuable and share the associated criticism of the concept of the digital divide, I will propose a perspective that is diametrically opposed to the argumentation of Bezuidenhout. For this, I will argue that we currently find particular forms of digital workâand uses of the internetâwhose functioning builds and relies on the exclusion of, and disconnection from, social, material, personal, and knowledge-based requirements. The planning, logistics, organisation, and core functional/technical principles of digital wage labour in African contexts, for example, are feasible and economically successful, precisely because they disconnect the enterprise from local entanglements and situated knowledge cultures. The effects as well as their side effects will be presented in the following empirical sections.
3
The research project that allowed me to travel to Gulu again in September 2018 was dedicated to the development and expansion of university curricula.5 One initial aim was to introduce social science perspectives on science and technology as part of humanities and engineering courses. Furthermore, the project also aimed to augment teaching capacities through the dissemination of teaching materials from online courses such as Massive Open Online Courses (MOOC s). At this point during the project, my involvement with the universityâs internal information and communication infrastructures (ICT s) intensified, as it quickly became clear that only functioning and reliable ICT s would enable the provision and use of e-learning formats at all. During a meeting with the dean of the Computer Science Faculty, I asked where he would be getting internet access from. While answering my question, he pointed to a container just outside his window and added that he would get good internet connection from Samasource: âTheirs is more reliable and faster than the one we have on campus.â6 Indeed, shortly before, I had been told by the universityâs network manager that the campus Wi-Fi connection was very limited. In order to better understand this parallel and simultaneous functionality and non-functionality of internet connections, it is necessary to look into the more recent history of the northern Ugandan region.
Until 2008, the municipality as well as the surrounding district of Gulu formed the centre of civil war that started in the early 1980s and lasted for decades (Büscher et al. 2018). From my first visit in 2010 until my last stay in May 2019, Gulu has increasingly transformed into a regional commercial centre connecting several important trading arteries. During this time, the city became a hub for international non-governmental organisations whose post-conflict development projects not only brought far-reaching financial resources, but additionally offered new diverse employment opportunities to the region (Redfield 2013). Even until today, most major aid organisations (e.g. UNICEF, Oxfam, USAID, GIZ, MSF) continue to have regional offices in the city. In recent years, Gulu has also served as logistical hub for other crisis
As part of its post-conflict recovery and development measures, Oxfamâa UK-based, non-governmental organisationâinitiated a supra-regional project called Internet Now!. Starting in 2012 the aim was to make internet-based learning and employment options available focusing largely on rural areas. In collaboration with the Gulu University administration, ten containers were placed on campus providing equipment and resources for approximately thirty computer science students and graduates; allowing them to carry out early forms of micro-work. After initial collaborations with Oxfam, Samasource took over the Gulu branch of Internet Now! in 2016.7 Samasource is one of the pioneering organisations in the field of impact sourcing, a subtype of Business Process Outsourcing (BPO), a widespread corporate practice through which sub-processes and tasks are outsourced to other specialised companies. However, unlike BPO, impact sourcing companies claim that the services they provide are managed through employment of marginalised and disadvantaged populations (Heeks 2013). Against this background Samasource promises to address two problems at once: âthe big data problem and the global poverty problemâ (Janah 2017, 118). According to the organisationâs self-portrait, Samasource offers jobs and training for digital activities to people below the poverty line in Kenya, Uganda, Haiti, and India. Since its foundation in 2008, Samasource has launched sixteen branchesâof which some are no longer operationalâin which it employed an estimated eight thousand workers. The organisationâs motto âGive work, not aidâ (Janah 2017, 3) follows a free market logic, with its core assumption that the provision of wage labour opportunities empowers individuals in developing countriesâbetter than, for instance, international aidâto lift themselves and their families out of poverty. The extent to which impact sourcing attempts can actually contribute to the reduction of global poverty is difficult to answer.
For Samasource Gulu, one can assert that it is predominantly students and graduates who are attracted to and recruited by the organisation.8 Even
In order to stay operational in such a heavily competitive field of business outsourcing, Samasource Gulu must guarantee a high level of quality and reliability, starting at the technical and infrastructural level. Running an outpost like Gulu exposes their operations to heavy and long power cuts often lasting several hours a day.9 After the takeover of Internet Now! Samasource added sixteen additional containers and made room for one hundred more working stations. To sustain this extension a new generator had to be purchased, as well as a second internet connection, which now serves as a back-up in case the first connection is interrupted. William, the manager summarises the situation as follows: âWe run our own system here! It is completely independent from the university. If we had to rely on their system we would definitely fail.â10 Even though there are only a few metres between the universityâs own computer lab and Samasource containers, the contrast could not be greater. Extreme heat, high humidity, and fine dust have turned the computer lab into an unused, sad museum offering an unintended glimpse into the history of computer technology. Especially during power outages after sunset, the brightly lit windows of the Samasource containers are often the only light on the entire campus; a light that denotes productivity, but a productivity that is fully decoupled from local realities and spaces. The costs of these material and infrastructural decoupling measures are considerable. The internet connection alone amounts to a cost of at least $20,000 per year. It should be clear by now that the conditions for an organisation like Samasource to have settled in a location like Gulu are fundamentally related to the costly option of disconnecting its operations from the local daily material and infrastructural uncertainties.
4
âSometimes I find myself annotating faces of people I interact with.â11 This statement was made by a participant during a focus group discussion in which interlocutors spoke about the social effects of annotation activities. Other participants also remarked on its psychological effect, which extended into leisure time, and they attributed it to the predominantly repetitive nature of most of the activities. In general, most tasks at Samasource were described as monotonous, boring, and tiring. However, performing these tasks six days a week, for at least ten hours a day, forms the organisational basis and inner logic of data factory work. Samasource founder Leila Janah,12 was inspired by Henry Fordâs assembly line when developing her micro-work principles: âFord figured out a way to break down the making of an incredibly complex machine [the Ford Model T, RU] into small chunks that people with basic training could complete. He moved the Model T from the craftsmanâs studio into the mainstream. The assembly lines of the future apply the same thinking to digital work.â (ITWeb, 2021). Whereas in Fordâs factories it was the assembly line that dictated the rhythm and intensity of work, today the interface between employees and their workload is mediated by a more dynamic software. At Samasource this function is performed by SamaHub, its own in-house digital platform that has been developed over years with the following core functions: âSamaHub is Samasourceâs proprietary training data annotation platform. This web-based task management system helps facilitate large data projects through the customization of task workflows, task distribution, multi-tiered quality control, and project-based training. [â¦] SamaHub has three main functions: task distribution, data structuring work (image annotation, data categorizing, other writing/research) and quality managementâ (Samasource 2019, 1).
Terms such as âtask-decomposition,â âtask-distribution,â or âtask-completionâ indicate sub-steps in which the software automatically splits up projects (e.g., large files of visual data), distributes it, and also returns it to clients and customers as valuable annotated training data. This has become a fairly complex operation in which clients have the option to check and control the data labelling in real-time. They can change data volumes, and also make use of the labelled data almost immediately. The activities indicated in Figures 4.1 and 4.2
Face annotation (Edge AI and Vision Alliance 2019a)
EDGE AI AND VISION ALLIANCE. (2019A). SAMASOURCE DEMONSTRATION OF ITS SAMAHUB TRAINING DATA ANNOTATION PLATFORM. YOUTUBE VIDEO, 2:57. LAST MODIFIED OCTOBER 9, 2019. HTTPS://WWW.YOUTUBE.COM/WATCH?V=FDS09IUAGMO
Geographical area annotation (Edge AI and Vision Alliance 2019b)
EDGE AI AND VISION ALLIANCE. (2019B). SAMASOURCE DEMONSTRATION OF ITS SAMAHUB TRAINING DATA ANNOTATION PLATFORM. YOUTUBE VIDEO, 2:57. LAST MODIFIED OCTOBER 9, 2019. HTTPS://WWW.YOUTUBE.COM/WATCH?V=FDS09IUAGMO
The comparison with Fordâs assembly line can also be found within the execution speed and the timing of individual labelling tasks. For instance, the average time an employee has for annotating a face is usually calculated by the SamaHub platform. With each new face the time is displayed as a countdown above the image. Next to the countdown called default time, is another timer counting up the real time that is called elapsed time that the employee actually needs for fulfilling the task. This temporal regime of (self)control affects the performability of the respective activity in several ways. At the individual level, performance that is measured as too slow will negatively impact the employeeâs internal evaluation, which again affects the future tasks they will be assigned to, as well as employment and contractual negotiations. On a more practical level, if SamaHub detects that the default time is collectively exceeded, adjustments can be made that eventually result in an extension of default time. A helpful conceptualisation of this partial operational logic of labour control through digital platforms is called âlogistical media theoryâ (Rossiter 2016, 7): Logistical labour emerges at the interface between infrastructure, software protocols, and design. Labour time is real-time. The formation of logistical media theory, therefore, requires an analysis of how labour is organised and governed through software interfaces and media technologies that manage what anthropologist Anna Tsing (2009) identifies as âsupply chain capitalism.â (Rossiter 2016, 23)
This logistical extraction of labour is thus based on a twofold decoupling: On the one hand the breaking down of videos into individual frames decontextualizes the content, as it operates with a simplification that allows a decoupling from any local knowledge practices. On the other hand, the software offers a significant decoupling from local, often limited, skillsets and work experiences. Even without any IT skills people can easily be enrolled and made part of this new form of digital value production and supply chain capitalism.
Analogous to Fordâs workers at the assembly line, employees of Samasource also attempted to momentarily escape control, monitoring, and supervision.14 The management used the term âoutliersâ indicating that they are aware of
5 Decoupling from Local Work Cultures
Calculatory practices that relate the scheduled amount of labour to the actual labour needed is part of broader history of organisational learning and industrial labour culture, which in the case of Samasource Gulu, interacts with but also conflicts with, local work practices. This final section will be dedicated to the discussion of the relation between digital factory labour requirements and local labour cultures. Apart from a few companies in the capital city of Kampala, elaborate control and tight time keeping devices are rarely part of the everyday work experience of most Ugandans. In addition to the entrance login, and the login at their workplace, all employees of Samasource have to hand in their mobile phones on entering the building. Mobile phones may only be used during strictly recorded breaks. Most of the interviewees found the mobile phone policy to be the most unpleasant control measure. It was repeatedly indicated that during working hours, employees could not respond, for example, to urgent family needs and other issues.16 It also regularly happened that individual workers completed their daily workload well before the end of their shift. I heard complaints from employees indicating that after having finished early, for example, around three oâclock in the afternoon, they were neither allowed to leave nor could they use the internet for private use until the end of the shift, usually at six oâclock in the evening. One employee described the situation as follows: âWhen you finish early you are just supposed to sit and do nothing.â17 When I asked the management about their policy on that, I was told that there were indeed waiting times every now and then, particularly because quality control managers had to check their team members work
On the one hand the formalisation techniques for labour registration and worktime measures practised at Samasource did not go beyond corporate governance policies usually found in transnationally operating companies. On the other hand, the above-described contradictions regarding the end of shift also suggest that strict formal regulations interact with local forms of social control within corporate contexts. In the words of one informant: âThe problem is not Samasource. The problem is on the ground.â19 The reflection refers to patronage relationships among the personnel through which access to tasks and resources are either granted or restricted. Starting from recruitment and allocation of shifts (day/night), kinship networks, and even ethnicity formed dominant inclusion or exclusion criteria potentially undermining merit principles. Another complaint was that it mattered whether one was favoured by middle management. When I enquired why the management in San Francisco headquarters did not attempt to regulate certain abuses of power, it was mentioned that they were largely kept in the dark about the local conditions. During visits from the San Francisco office, only pre-selected employees who were considered loyal were allowed as contacts.
In light of these social and organisational entanglements, it remains a challenge to judge the extent to which this form of repetitive and uncreative data work is adequately renumerated. During my fieldwork, however, this question repeatedly emerged in a strangely paradoxical manner: When I reported the type of labour practiced at Samasource Gulu to Ugandan colleagues, they rarely shared my critical questioning of the monthly payments. Instead, an income of USh460,000 per month (approximately $120) was deemed fair pay for a university graduate. In the broader national context such income of newly graduated students would sit closer to around USh250,000 (approximately $65) for
Although this justification seems to be rather speculative and less grounded in empirical evidence, at this point I would like to draw attention to a specific feature of Ugandan work culture, which is not taken into consideration in Janahâs statement. In Uganda, as well as in other countries with comparatively low incomes, the line between formal wage labour and informal employment remains a difficult and complex distinction. However, through my research in the Ugandan health sector, I became aware that the distinction between the state/public and the private sector plays a significant role in perception of local work cultures (Umlauf 2017). For example, it is a very likely scenario that a nurse or a doctor employed in the private sector would receive a higher monthly income than in the same job in the public sector. Nevertheless, a large proportion of private sector health workers would prefer to work in the public sector. The main reason, apart from generally higher job security, was the opportunity to participate in further training sessions. As part of internationally funded donor projects, these capacity building missions take place several times a year and are highly attractive, especially because of their daily allowances, which often bring in several additional monthly salaries (Swidler and Watkins 2009). In addition to these benefits of accessing the development industry, almost all government employees were engaged in side businesses constituting an additional source of income (McCoy et al. 2008). In contrast, greater control and regulation of working hours of privately employed health workers make it much more difficult to operate and sustain secondary incomes. The strict labour regulations and control of working hours at Samasource, together with the strongly limited communication options, conflict in various ways with local working cultures and the need to divert sources of income. Against this background a higher pay could thus be justified relatively easily without fear of distorting local wage and labour markets, as indicated in Janahâs statement.
I would like to conclude this section by returning to Samasourceâs ethical claim of lifting people out of poverty by giving them jobs instead of aid. The fact that marginalised tiers of the population are mainly characterised by poor access to formally secured wage labour is reinforced by the aspect that many have only limited knowledge of labour protection and labour law regulations. As already mentioned, we can assume that the majority of employees at Samasource Gulu do not belong to the poorest or most marginalised
6 Conclusion
Entitled âDigital Dividend,â the 2016 World Bank report states that, in a historical comparison, the internet reached many developing countries much faster than was the case for other technologies (World Bank 2016, 5). At the same time and despite the spread of mobile money and the use of social media, the everyday lives of the majority of people living in the so-called developing world still offers very limited interfaces with digital services and their promised benefits. With reference to these observed frictions, I have attempted to outline a more complex narrative of possible and actual forms in which access to digital services and employment opportunities are currently (infra)structured in Uganda. The ethnographic case study of Samasource Gulu was intended to demonstrate the ambivalence with which the spread of the internet takes place. On the one hand, it extends largely to very specific and strongly regulated sub-sectors, such as the digital mass and micro-labour. On the other hand, the study was also meant to illustrate how software-driven tailoring of the internet neither increases digital literacy nor does it produce much value for other parts of peopleâs everyday lives. I have shown that the connection and access to new digital employment opportunities both requires and goes hand in hand with disconnections from other forms of local knowledge systems and work cultures.
It remains difficult to predict what future developments will emerge from this still nascent form of digital wage labour in low-income countries. At
Bibliography
Alexander, Jan. 2018. âHow Samasource Reverses Poverty One Job at a Time.â This is Capitalism. Last modified January 8, 2018. https://www.thisiscapitalism.com/samasource-reverses-poverty-one-job-time/.
Anwar, Mohammed Amir and Mark Graham. 2019. âHidden Transcripts of the Gig Economy: Labour Agency and the New Art of Resistance Among African Gig Workers.â Environment and Planning: Economy and Space 52, no. 7: 1â23.
Bajwa, Uttam, Denise Gastaldo, Erica Di Ruggiero, and Lilian Knorr. 2018. âThe Health of Workers in the Global Gig Economy.â Globalization and Health 14, no. 124.
BergvallâKÃ¥reborn, Birgitta and Debra Howcroft. 2014. âAmazon Mechanical Turk and the Commodification of Labour.â New Technology, Work and Employment 29, no. 3: 213â23.
Bezuidenhout, Louise, Ann Kelly, Sabina Leonelli, and Brian Rappert. 2016. ââ$100 Is Not Much To Youâ: Open Science and Neglected Accessibilities for Scientific Research in Africa.â Critical Public Health 27, no. 1: 39â49.
Bezuidenhout, Louise, Sabina Leonelli, Ann Kelly, and Brian Rappert. 2017. âBeyond the Digital Divide: Towards a Situated Approach to Open Data.â Science and Public Policy 44, no. 4: 464â75.
Büscher, Karen, Sophie Komujuni, and Ivan Ashaba. 2018. âHumanitarian Urbanism in a Post-Conflict Aid Town: Aid Agencies and Urbanization in Gulu, Northern Uganda.â Journal of Eastern African Studies 12, no. 2: 348â66.
Daub, Adrian. 2020. What Tech Calls Thinking: An Inquiry into the Intellectual Bedrock of Silicon Valley. New York: Farrar, Straus and Giroux.
Donner, Jonathan. 2015. After Access: Inclusion, Development, and a More Mobile Internet. Cambridge, MA: MIT Press.
Ettlinger, Nancy. 2016. âThe Governance of Crowdsourcing: Rationalities of the New Exploitation. Environment and Planning A: Economy and Space 48, no. 11: 2162â80.
Ettlinger, Nancy. 2017. âParadoxes, Problems and Potentialities of Online Work Platforms.â Work Organisation, Labour & Globalisation 11, no. 2: 21â38.
Fish, Adam and Ramesh Srinivasan. 2011. âDigital Labour Is the New Killer App.â New Media & Society 14, no. 1: 137â52.
Fuchs, Christian and Eva Hora. 2008. âAfrica and the Digital Divide.â Telematics and Informatics 25, no. 2: 99â116.
Gino, Francesca and Bradley R. Staats. 2011. âSamasource: Give Work, Not Aid.â Harvard Business School Case 912â011, December.
Gray, Mary and Siddharth Suri. 2019. Ghost Work: How to stop Silicon Valley from Building a New Global Underclass. Boston: Houghton Mifflin Harcourt.
Heeks, Richard. 2013. âInformation Technology Impact Sourcing.â Communications of the ACM 56, no. 12: 22â25.
Irani, Lilly. 2015. âThe Cultural Work of Microwork.â New Media & Society 17, no.5: 720â739.
Irani, Lilly and Six Silberman. 2013. âTurkopticon: Interrupting Worker Invisibility in Amazon Mechanical Turk.â CHI 13: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 611â20. Paris.
ITWeb. n.d. âBusiness Technology News.â Accessed January 11, 2021. https://www.itweb.co.za/content/mYZRXM9JyD8qOgA8.
Janah, Leila. 2017. Give Work: Reversing Poverty One Job at a Time. New York: Penguin.
Lee, Dave. 2018. âWhy Big Tech Pays Poor Kenyans to Teach Self-Driving Cars.â BBC News. Last modified, November 3, 2018. https://www.bbc.com/news/technology-46055595.
Lost Research Group. n.d. Lost Research Group. Accessed November 16, 2021. https://lost-research-group.org/.
LXMI. n.d. âUltra Clean Skincare | Organic & Fair Trade | LXMIâ. Accessed January 10, 2021. https://lxmi.com/.
Mariscal, Judith. 2005. âDigital Divide in a Developing Country.â Telecommunications Policy 29, no. 5â6: 409â28.
McCoy, David, Barbara McPake, and Sara Bennett. 2008. âSalaries and Incomes of Health Workers in Sub-Saharan Africa.â The Lancet 371, no. 9613: 675â81.
MITODL Video Services. 2006. Opening Keynote and Keynote Interview with Jeff Bezos. MITODL video, 1:02:38. Last modified December 4, 2018. https://techtv.mit.edu/videos/16180-opening-keynote-and-keynote-interview-with-jeff-bezos.
Redfield, Peter. 2013. Life in Crisis: The Ethical Journey of Doctors Without Borders. Berkeley: University of California Press.
Roberts, Sarah. 2019. Behind the Screen: Content Moderation in the Shadows of Social Media. New Haven: Yale University Press.
Rossiter, Ned. 2016. Software, Infrastructure, Labor: A Media Theory of Logistical Nightmares. New York: Routledge.
Samasource. 2019. Samahub Process and Feature Overview. Samasource. Accessed January 10, 2021. https://www.samasource.com/hubfs/Collateral/SamaHub%20Solution%20Brief%202019.pdf.
Selwyn, Neil. 2004. âReconsidering Political and Popular Understandings of the Digital Divide.â New Media and Society 6, no. 3: 341â62.
Selwyn, Neil and Kirk Facer. 2010. âBeyond Digital Divide: Toward an Agenda for Change.â Handbook of Overcoming Digital Divides. Constructing an Equitable and Competitive Information Society, Volume 1, edited by Enrico Ferro, Yogesh Kumar Dwivedi, J. Ramon Gil-Garcia, and Michael D. Willaims, 1â20. Hershey, PA: IGI Global.
Shrum, Wesley, Keith Benson, Wiebe Bijker, and Klaus Brunnstein. 2007. Past, Present and Future of Research in the Information Society. New York: Springer.
Standage, Tom. 2003. The Turk: Life and Times of the Famous 18th Century Chess-Playing Machine. New York: Walker & Co.
Swidler, Ann and Susan Cotts Watkins. 2009. ââTeach a Man to Fishâ: The Doctrine of Sustainability and Its Effects on Three Strata of Malawian Society.â World Development 37, no. 7: 1182â96.
Techrepublic. 2014. âSilicon Valleyâs âSolutionismâ Issues Appear to be Scaling.â Last modified August 25, 2015. https://www.techrepublic.com/article/silicon-valleys-solutionism-issues-appear-to-be-scaling.
Umlauf, René. 2017. Mobile Labore. Zur Diagnose und Organisation von Malaria in Uganda. Bielefeld: Transcript Verlag.
Vallas, Steven and Juliet B. Schor. 2020. âWhat Platforms do? Understanding the Gig Economy.â Annual Review of Sociology 46, no. 5: 273â94.
World Bank. 2016. World Development Report 2016: Digital Dividends. Washington D.C.: World Bank Publications.
Interview #2, Manager Samasource, Gulu, September 19, 2018.
The term was coined by Amazon founder Jeff Bezos during an opening keynote at MIT Emerging Technologies Conference in 2006 (MIT 2018).
Silicon Valley solutionism refers to prevailing ideology and beliefs that for each and every social challenge there is technical solution (see Techrepublic 2014).
âThe Turkâ or âMechanical Turkâ was a eighteenth century mechanical man dressed in a Turkish costume, that was capable of playing chess. While the set-up was to pretend that the Turk could play automatically it was actually fake, since its chess skills relied on a human hidden inside the machine (Standage 2002).
The project funded by the German Research Foundation (DFG) entitled Curriculum Development (2017â19) was based at the Department of Social Anthropology at Martin Luther University Halle (see Lost n.d). During the field visits of three weeks each (September 2018 and May 2019), I conducted three focus group discussions (University staff, students, Samasource staff) and ten semi-structured interviews with current and former Samasource staff in Gulu. Other quotations and information are taken from sporadic conversations and chance encounters on Gulu campus between Samasource management and myself.
Interview #1, Dean of Faculty Computer Science, Gulu, September 18, 2018.
Unfortunately, I was not able to piece together more precisely which exact circumstances led to both the initial collaboration and the eventual takeover by Samasource.
It is difficult to answer which tiers of society Samasource mainly recruits its workforce from. However, in locations like Nairobi, Samasource does not operate directly on a university campus, which might potentially have an impact on who is applying and eventually recruited.
During these power cuts, a campus generator can be switched on to supply power to the rest of the campus. However, the cost of diesel consumption has to be paid for from university funds.
Interview #4, Manager Samasource, Gulu, May 6, 2019.
Focus Group Discussion #5, Samasource Staff, Gulu, May 5, 2019.
Leila Janah passed away on 24 January 2020. Until then, she ran the cosmetic label LXMI in addition to Samasource. The label primarily sells products that use shea butter as a key ingredient. The nuts form the shea trees were also harvested in Uganda, and Janah also promoted this branch of the company as providing marginalised women with meaningful and sustainable employment (see LXMI 2021).
When I write here âwithout much interpretative effortâ, I am nevertheless aware that it is up to the interpretation of an individual, for example, where a house exactly begins and ends. These activities carry less moral interpretation than tasks assigned to the already mentioned content moderators. The latter must be able to assess throughout several cultural, aesthetic, religious and ethical codes if something is considered violent or hate speech. See chapter by Helen Robertson in this volume.
However, I would not go as far as to call these practices acts of sabotage.
Interview #4, Manager Samasource, Gulu, May 6, 2019.
In this context, it was also mentioned that young mothers were probably not allowed to breastfeed babies on the company premises during working hours.
Focus Group Discussion #3, Samasource Staff, Gulu, September 19, 2018.
The issue of quality data has gained some importance as it serves as a niche for companies like Samasource, which helps them justify and distinguish their costly and elaborate factory-style operations against the remote and individualised operational characteristic of many other sectors of the gig economy. IT-Tech as well as automotive companies require properly labelled data for their learning algorithms that are often only achieved through strict monitoring and quality control measures. Sending out micro-tasks on crowd platforms, like AMT or Upwork, exposes the data to significant differences in the quality that is produced. See in this regard also Anwar and Grahamâs (2019) study on adaptation practices of African gig workers to cope with the (labour) conditions they are facing on Upwork.
Focus Group Discussion #5, Samasource Staff, Gulu, May 5, 2019.
The threat of potential redundance of this field of employment is already in the minds of the Samasource management: âIn spite of such success stories, however, Janah is frank about the challenges ahead. The greatest concern is that as technology continues to advance, in five to ten years machines might perform the tasks that Samasource is now training people to doâ (Alexander n.d).