The DOI matches its Crossref record; 111 of the 111 references it cites were found in a registry or an index; 27 of the 97 sources its claims cite were checked for support.
'The plan is just survival': Data Work in Kenya and the Regime of Entrapment
Open accessAbstract
The rapid expansion of the AI industry relies heavily on the production, verification, and maintenance of data, otherwise known as "data work". Companies outsource and offshore this work through global AI supply chains that operate under exploitative conditions. Drawing on semi-structured interviews with Kenyan data workers across platforms and BPOs, this paper examines how such conditions take shape and persist. We argue that workers are caught within a regime of entrapment, a system of interconnected mechanisms that make it difficult for workers to leave or improve their positions. These mechanisms include the push to invest in the promise of ‘AI’ jobs, the use of precarious contracts to govern workers, the capture of regulatory institutions, and the exploitation of global labor arbitrage. Using complementary lenses of neoliberal governmentality, precarity, and supply chain capitalism, we analyze why labor mobilization in this sector remains uniquely constrained. We conclude by outlining an orientation for research and scholarly practice that can support workers’ organizing efforts and contest the structural conditions sustaining this regime.
Overview
CoChat Summary
TL;DR
Drawing on semi-structured interviews with 18 Kenyan data workers, this paper argues that workers are held in AI supply chains by a "regime of entrapment": interlocking mechanisms that make leaving or improving their positions extremely difficult. It theorizes six such mechanisms and argues that labor mobilization in this sector is uniquely constrained by precarious contracts, regulatory capture, and global wage arbitrage.
Key findings
- Workers described six mechanisms binding them to data work: (1) required upfront investment in the promise of AI jobs (equipment, internet, unpaid training), (2) normalized under-employment, (3) governance through precarious or absent contracts, (4) capture of regulatory institutions, (5) exploitation of global labor arbitrage, and (6) a sustained illusion of mobility.
- Reported figures include a platform setup costing "over 100,000 Kenyan Shillings" with at least 16 GB RAM and 15 Mbps internet (Matthew); $40 recurring internet cost, roughly two weeks of platform earnings (Emmet); a COVID-era rate cut from $1.45 to $1.25/hour; $25 total pay for a full month of 8-hour days (Dinah); and ~$2,000 unpaid after coordinating ~15,000 videos (Andrew).
- Wage stagnation and decline were pervasive: Joseph earned roughly the same hourly wage after eight years in BPO work, and Emmanuel reported earning over $350/month when starting on Remotasks before per-task pay fell; Emmanuel contrasted US rates of over $60/hour with "barely over a dollar per hour" in Kenya.
- Workers reported that skills did not cohere into portable credentials — no certificates, non-compete and non-disclosure clauses, and only screenshots or emails as proof of employment — foreclosing exit into other fields.
Methods
- Semi-structured interviews with 18 Kenyan data workers conducted remotely on Zoom in English between July–August 2025, lasting approximately one hour each, with participants compensated 50 Euros; recruitment was facilitated through the Data Labelers Association (DLA), and the study was framed as a workers' inquiry.
- Participants were all located in urban Kenya (e.g., Nairobi, Juja, Machakos); 11 had experience in both BPOs and platforms, 5 only platforms, and 2 only BPOs. Demographics are reported only in aggregate and pseudonyms are used, though platform and BPO names are not anonymized.
- Thematic analysis guided by Charmaz's constructivist grounded theory, with line-by-line and focused coding by three team members, weekly team discussions, reflexive memos, member checking, and reported theoretical saturation.
Limitations
- Stated by the authors: the study was not designed to investigate gender as an analytic category, though exploitation is likely felt differentially based on caring responsibilities and social position; there is "central ambiguity" over whether the regime of entrapment is deliberately orchestrated or emerges from disjointed practices, which they suggest future "studying up" work could address; and their use of Foucault's neoliberal governmentality "only begins to capture the complexities of neoliberalism."
- Evident from the text: the sample is small (18), drawn from a single worker collective in urban Kenya, and interviews were conducted remotely rather than in person; the authors also note that their positionality as researchers largely external to Kenyan data work communities shaped what participants shared and how accounts were interpreted.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
21 Valid · 3 Not valid · 37 Not checked · 108 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Behind the illusions of autonomy and machine intelligence are vast global networks of human workers who clean, tag, classify, and verify data and model outputs [96].”
What the sources say
- Supportedcites [96] [96] James Muldoon, Callum Cant, Boxi Wu, and Mark Graham. 2024. A typology of artificial intelligence data work. Big Data & Society 11, 1 (2024), 20539517241232632. arXiv: Go to Citation Crossref Google Scholar source
The Introduction says AI systems require human workers to collect, curate, annotate, and evaluate data, often through outsourced production networks. “Microwork, crowdwork and AI data work” includes evaluation of model outputs, while “Model evaluation and data verification” describes workers reviewing and correcting predictions and labels. (LLM, conf=0.93)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The organization of contemporary AI data work is entangled with longer histories of colonial extraction and racialized dispossession that shape where value is produced and whose labor is rendered interchangeable.”
What the sources say
- Partly supportedcites [108] [108] Julian Posada. 2021. The coloniality of data work in Latin America. In Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society. 277–278. Digital Library Google Scholar a [...] South and Southeast Asia, and Latin America b [...] and political-economic conditions elsewhere c [...] difference. Attending to Posada’s sourceExperimental — accuracy not yet measured
The paper (p. 1) links contemporary AI data work to centuries-long exploitative relations and describes wealthy institutions in advanced economies benefiting from cheap labor in developing countries. It also describes labor being exchanged as a commodity. Our copy does not specifically establish the claim about racialized dispossession or say that workers are rendered interchangeable. (LLM, conf=0.83)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“The difficulty of locating responsibility within AI supply chains is a built-in feature of how they are organized [140].”
§ 6 Discussion > 6.2 Locating responsibility within the AI supply chain
What the sources say
- Supportedcites [140] [140] David Gray Widder and Dawn Nafus. 2023. Dislocated accountabilities in the “AI supply chain”: Modularity and developers’ notions of responsibility. Big Data & Society 10, 1 (Jan. 2023), 20539517231177620. Crossref Google Scholar a [...] built-in feature of how they are organized b [...] companies, BPOs/platforms, and governments source
The Introduction and background (p. 1) links dislocated accountability to modularity and the divisions of labor it enables. “Reproducing the supply chain” (p. 5) says those divisions create cracks through which responsible AI actions fall, supporting the claim that difficulty locating responsibility arises from how the supply chain is organized. (LLM, conf=0.92)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“Across both settings, we observe a convergence in labor arrangements that echoes Zhang et al.’s [148] observation of hybrid management practices.”
§ 6 Discussion > 6.1 Examining how precarity constrains collective resistance in data work
What the sources say
- Supportedcites [148] [148] Ben Zefeng Zhang, Tianling Yang, Milagros Miceli, Oliver L. Haimson, and Michaelanne Thomas. 2025. The Making of Performative Accuracy in AI Training: Precision Labor and Its Consequences. In Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. ACM, Yokohama Japan, 1–19. Go to Citation Digital Library Google Scholar source
Section 4.2.4 describes Zhang et al.’s observation of hybrid management: workers were managed through both factory-like criteria and platform-style gamification. The citing authors present convergence across their own two settings as their observation, rather than attributing that convergence to Zhang et al. (LLM, conf=0.91)
plainstates an association- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
- Supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“Critical STS scholars remind us that precarity must also be understood as situated within global histories of coloniality and racial capitalism [22, 135].”
§ 6 Discussion > 6.1 Examining how precarity constrains collective resistance in data work
What the sources say
- Not checked — no full text of the sourcecites [22] [22] Paula Chakravartty and Denise Ferreira Da Silva. 2012. Accumulation, dispossession, and debt: The racial logic of global capitalism—An introduction. American Quarterly 64, 3 (2012), 361–385. Go to Citation Google Scholar source
- Supportedcites [135] [135] Lucas Van Milders. 2021. Precarity/coloniality. Theory & Event 24, 4 (2021), 1068–1089. Go to Citation Google Scholar source
The Introduction says precarity should be understood within a global matrix of coloniality and links it to the co-constitutive relations of capitalist imperialism and racialized colonialism. This supports the citing sentence’s paraphrase. (LLM, conf=0.94)
plainstates an associationn = 2- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
74 more claims · 4 with their cited sources checked
Create a free account to see all 169 claims and to verify the 36 references still open.
- Probably not a substantial claim. 50
- It may state two separate claims as one. 61
Citations
Cites
Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective
Haraway, D. (1988) · Feminist Studies · doi:10.2307/3178066 · Cited by 18,364
Are We There Yet? Data Saturation in Qualitative Research
Fusch, P., & Ness, L. (2015) · The Qualitative Report · doi:10.46743/2160-3715/2015.2281 · Cited by 4,460
Good Gig, Bad Gig: Autonomy and Algorithmic Control in the Global Gig Economy
Wood, A. J., Graham, M., Lehdonvirta, V., & Hjorth, I. (2019) · Work, Employment and Society · doi:10.1177/0950017018785616 · Cited by 2,209 · cited 1×
“This normalization echoes accounts of how flexibility rhetoric and algorithmic management encourage workers to govern themselves according to market imperatives [145].”
(2020) · doi:10.7551/mitpress/12255.001.0001 · Cited by 1,706
Represent and Destroy: Rationalizing Violence in the New Racial Capitalism
Melamed, J. (2010) · doi:10.5749/9780816678624 · Cited by 1,137
Supply Chains and the Human Condition
Tsing, A. (2009) · Rethinking Marxism · doi:10.1080/08935690902743088 · Cited by 980 · cited 7×
“Anna Tsing’s scholarship on supply chain capitalism offers a productive lens to understand the structural role of difference in sustaining these global production networks [126].”
InfluentialDal Bo, E. (2006) · Oxford Review of Economic Policy · doi:10.1093/oxrep/grj013 · Cited by 960 · cited 1×
“Regulatory capture could be understood as a process through which regulatory targets "manipulate the state agencies that are supposed to control them" (cf. [33]).”
Labour process theory and the gig economy
Gandini, A. (2019) · Human Relations · doi:10.1177/0018726718790002 · Cited by 946 · cited 1×
“Workers are thus reframed as “users” of platforms instead of being recognized as employees subject to capital-labor relationship [53].”
Turkopticon: interrupting worker invisibility in amazon mechanical turk
Irani, L. C., & Silberman, M. S. (2013) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/2470654.2470742 · Cited by 798 · cited 1×
“Turkopticon is a notable example [69], which was originally created as an activist tool for MTurk workers to evaluate their employers, and has later transformed into a worker-run advocacy group that provides vital resources and infrastructure for MTurk worker communities.”
Riders on the Storm: Workplace Solidarity among Gig Economy Couriers in Italy and the UK
Tassinari, A., & Maccarrone, V. (2020) · Work, Employment and Society · doi:10.1177/0950017019862954 · Cited by 746
Research works printed but not linked: 19
- Janine Berg, Marianne Furrer, Ellie Harmon, Uma Rani, and M. Six Silberman. 2018. Digital labour platforms and the future of work: towards decent work in the online world. International Labour Organization, Geneva.
- Callum Cant, James Muldoon, and Mark Graham. 2024. Feeding the machine: The hidden human labor powering AI. Bloomsbury Publishing USA.
- Kathy Charmaz. 2006. Constructing grounded theory: A practical guide through qualitative analysis. sage.
- NeurIPS – Neural Information Processing Systems Conference. 2023. Code of ethics. https://neurips.cc/public/EthicsGuidelines Accessed: 2025-01-14.
- Kate Crawford. 2021. The Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.
- Michel Foucault. 2003. " Society Must Be Defended": Lectures at the Collège de France, 1975-1976. Vol. 1. Macmillan.
- Michel Foucault. 2008. The Birth of Biopolitics: lectures at the Collège de France, 1978 - 79. Palgrave Macmillan, New York.
- Mary L Gray and Siddharth Suri. 2019. Ghost work: How to stop Silicon Valley from building a new global underclass. Harper Business.
- Steven Hill. 2015. Raw Deal: How the" Uber Economy" and Runaway Capitalism Are Screwing American Workers. Macmillan.
- Ursula Huws. 2014. Labor in the global digital economy: The cybertariat comes of age. NYU Press.
- Chidinma Iwu. 2024. Can African Unions Beat Big Tech? | Nonprofit Quarterly | Civic News. Empowering Nonprofits. Advancing Justice.https://nonprofitquarterly.org/can-african-unions-beat-big-tech/ [Online; accessed 2025-12-01].
- Isabell Lorey. 2015. State of Insecurity: Government of the Precarious. Verso Books.
- Laura Mann and Mark Graham. 2017. The Domestic Turn: Business Process Outsourcing and the Growing Automation of Kenyan Organisations. In Globalization, Economic Inclusion and African Workers. Routledge. Num Pages: 19.
- Richard Mathenge. 2024. Data Workers Organizing – The African Content Moderators Union. Technical Report. DAIR Institute. https://data-workers.org/richard [Videography and editing by P. Teka. Research support by A. Dinika, M. Miceli, L. Sachenbacher, C. Salim Wagner, & K. Kauffman].
- 10.5749/minnesota/9780816674244.001.0001 · DOI
- Milagros Miceli, Paola Tubaro, Antonio A. Casilli, Thomas Le Bonniec, Camilla Salim Wagner, and Laurenz Sachenbacher. 2024. Who Trains the Data for European Artificial Intelligence?Technical Report. European Parliament ; The Left. 1–40 pages. https://hal.science/hal-04662589
- Alex Rosenblat and Luke Stark. 2016. Algorithmic labor and information asymmetries: A case study of Uber’s drivers. International journal of communication 10 (2016) 27.
- Anna Lowenhaupt Tsing. 2015. The mushroom at the end of the world: On the possibility of life in capitalist ruins. In The Mushroom at the End of the World. Princeton University Press.
- UN. 2024. High Level Roundtable on Green Jobs & Skills Development in Kenya. https://www.unep.org/events/conference/high-level-roundtable-green-jobs-skills-development-kenya
Cited by
Factories or workhouses? Digital platforms and the re-feudalisation thesis
Arvidsson, A., & Bonini, T. (2026) · Dialogues on Digital Society · doi:10.1177/29768640261463048 · Cited by 3
(2000) · Time to knit · doi:10.1145/3772318.3809022
Metadata
- Authors
- Kapania, S., Yang, T., Abdelkadir, N. A., Scheuerman, M. K., Miceli, M., Taylor, A. S., & Fox, S. E.
- Affiliations
- Carnegie Mellon University (US)
- Weizenbaum Institute (DE)
- University of Minnesota (US)
- Unaffiliated, Barcelona, Spain
- University of Edinburgh (GB)
- ORCID
- Venue
- Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems
- Publisher
- ACM
- Published
- 2026-04-13
- Pages
- 1-17
- OpenAlex
- W7153972188
- Cited by
- 2
- References
- 103
- Citations by year
- 2026: 2
- FWCI
- 47.08
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- Locations
- CHI Conference on Human Factors in Computing Systems (publishedVersion, cc-by-nc-nd)
- Edinburgh Research Explorer (University of Edinburgh) (submittedVersion, other-oa)
- Funders
- German Federal Ministry of Research, Technology and Space (BMFTR) (16DII134)
- UK Arts and Humanities Research Council (AH/X007146/1)
- NSF (National Science Foundation) (2442153)
- Topics
- Social Sciences › Sociology and Political Science › Digital Economy and Work Transformation
- Social Sciences › Business, Management and Accounting › Industrial relations › Legal and Labor Studies
- Social Sciences › Business, Management and Accounting › Strategy and Management › Global trade, sustainability, and social impact
- Keywords
- data work, artificial intelligence industry, outsourcing, business process outsourcing, precarious employment, labor arbitrage, neoliberal governmentality, supply chain capitalism, labor mobilization, worker organizing, Kenya, data workers
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-29, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-29 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: 'The plan is just survival': Data Work in Kenya and the Regime of Entrapment. https://app.cochat.ai/papers/10.1145/3772318.3791097
- The paper
- Kapania, S., Yang, T., Abdelkadir, N. A., Scheuerman, M. K., Miceli, M., Taylor, A. S., & Fox, S. E. (2026). 'The plan is just survival': Data Work in Kenya and the Regime of Entrapment. Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3772318.3791097
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.