The Muldoon Study
The Muldoon study is the most rigorous academic examination of what happened to the organization Janah founded after her death. Based on fieldwork at three of Sama's East African delivery centers, the study documented conditions that included wages of approximately two dollars per hour to workers while the outsourcing firm billed up to twelve dollars per hour to clients including OpenAI and Meta, insecure employment, tightly disciplined labor management, gender-based exploitation, and inadequate support for content moderators exposed to traumatic material. The study's significance for the Orange Pill Cycle is structural: it provides the empirical test case for the claim that institutional commitments require continuous human stewardship to survive market pressure, and it identifies the specific countervailing institutional forces — organized workers, civil society oversight, regulation — required to sustain dignified labor when individual leadership is insufficient.
The Infrastructure of Extraction — Contrarian ^ Opus
There is a parallel reading that begins not with institutional erosion but with the material substrate that makes digital labor possible: the cables, servers, and energy systems that connect Nairobi to San Francisco. From this vantage, what the Muldoon study documents is not the failure of a social enterprise but the predictable outcome of a global infrastructure designed for extraction. The two-to-twelve wage ratio isn't market pressure eroding noble intentions — it's the architecture working exactly as designed, routing value from periphery to core through the same channels that have carried raw materials northward for centuries.
The study's call for countervailing measures — organized workers, civil society, regulation — assumes these forces can emerge within the existing infrastructure. But the physical and digital systems that enable remote content moderation were built to maximize value extraction while minimizing local agency. The Kenyan workers traumatized by violent content are connected to Meta's servers through networks they don't control, working on platforms they can't modify, producing value that flows through payment rails they can't influence. The infrastructure itself is the institutional structure, and it was never designed to sustain dignity. It was designed to move value efficiently from places where labor is cheap to places where capital accumulates. What looks like institutional erosion from Janah's perspective looks like system optimization from the perspective of the cables. The question isn't how to build countervailing power within this infrastructure but whether dignified digital labor is possible at all when the means of connection are themselves instruments of extraction.
In the AI Story
The study is not merely a critique of Sama but a theoretical contribution to the broader debate about whether market-based models of social impact can sustain dignity at scale. The authors argue that 'competitive market-based dynamics generate a powerful force that pushes such companies towards limiting the actual social impact of their business model in favour of ensuring higher profit margins. This force can be resisted, but only through countervailing measures such as pressure from organised workers, civil society, or regulation.'
The finding does not discredit Janah's founding vision. It specifies the conditions under which the vision can be sustained. Janah herself provided one form of countervailing pressure — leadership committed to maintaining the institutional line against market logic. What the study demonstrates is that leadership alone is insufficient over time, because leadership is temporary by nature. Durable institutional structures — the kinds of countervailing power the labor movement developed in earlier industrial transitions — are required for the commitments to outlast any individual champion.
The study's documentation of the content-moderation work is particularly significant. Workers hired to moderate violent and disturbing content for Meta described psychological trauma for which the organization provided inadequate support. A group of 184 moderators filed suit alleging unfair termination and poor working conditions. The specifics illustrate how the general pattern of market erosion concentrates its damage on the workers least able to bear additional cost.
The study's relevance to the AI transition extends beyond Sama. As foundation models increasingly depend on large-scale human annotation — for RLHF, for safety training, for specialized domain adaptation — the conditions under which that labor is performed become load-bearing for the technology itself. The Muldoon study demonstrates that the industry's current approach to labor supply chains produces exploitation that undermines both the workers and the quality of the models their labor trains.
Origin
The study emerged from the Fairwork project at Oxford's Internet Institute, a sustained research effort on platform labor that Mark Graham has led for over a decade. The specific focus on Sama reflected the organization's prominence as both a founding example of impact sourcing and a contemporary supplier to leading AI companies.
Fieldwork was conducted in 2022 across three Sama delivery centers in Kenya and Uganda, with interviews, document review, and direct observation producing the evidentiary base for the published findings.
Key Ideas
Empirical test case. The study provides the clearest available empirical test of the thesis that institutional commitments erode under market pressure absent countervailing forces.
Two-to-twelve wage ratio. The specific finding that workers received $2/hour while clients were billed $12/hour illustrates the concrete distributional consequences of institutional erosion.
Content moderation trauma. The documentation of inadequate support for workers exposed to violent content illustrates how market erosion concentrates damage on the most vulnerable workers.
Countervailing measures. The study's theoretical contribution is specifying what kinds of institutional architecture — organized workers, civil society oversight, regulation — are required to sustain dignity where leadership alone is insufficient.
Debates & Critiques
The study has been contested primarily by Sama's corporate response, which argued that its practices represented meaningful improvement over local labor market alternatives. The argument is empirically compatible with the study's findings but misses the normative target: the question is not whether Sama's practices exceed the worst available alternatives but whether they honor the founding commitments the organization was built around. The study's finding is that they do not, and that the gap between founding commitment and current practice illustrates the institutional fragility the Orange Pill Cycle must reckon with.
Appears in the Orange Pill Cycle
The Scale-Dignity Paradox — Arbitrator ^ Opus
The tension between Edo's institutional erosion thesis and the infrastructure critique depends entirely on what question we're asking. If we're asking whether market pressure erodes social commitments over time, Edo's framing is 90% correct — the Muldoon study provides compelling evidence that without countervailing forces, organizations drift toward extraction. The two-to-twelve wage ratio documents this drift precisely. But if we're asking whether dignified digital labor is achievable within current global systems, the infrastructure critique dominates (75%) — the physical and economic architecture of global digital work may be fundamentally incompatible with sustained dignity at scale.
The study's documentation of content moderation trauma illustrates where both views converge: these workers suffer because institutional commitments weakened (Edo's point) AND because the infrastructure treats their wellbeing as an externality (the contrarian point). Here the weighting is 50/50 — we need both lenses to understand why workers bear such disproportionate harm. The countervailing measures the study proposes — worker organization, civil society, regulation — are necessary but may be insufficient if the underlying infrastructure remains extractive.
The synthesis suggests reframing the problem as a scale-dignity paradox: digital labor achieves global scale through infrastructure that inherently resists local dignity. Janah's innovation was attempting to route dignity through extractive channels. The Muldoon study shows this works temporarily with extraordinary leadership but fails systematically over time. The solution likely requires both the countervailing institutions Edo emphasizes AND alternative infrastructure models the contrarian view implies — perhaps regional data processing centers, worker-owned annotation cooperatives, or fundamentally different architectures for connecting global demand with distributed labor. The question isn't whether to reform institutions or rebuild infrastructure, but how to do both simultaneously while AI systems increasingly depend on exactly the kinds of labor relations the current system makes exploitative.
Further reading
- James Muldoon, Callum Cant, Mark Graham, and Funda Ustek Spilda, "The poverty of ethical AI," AI & Society, 2023.
- Mary Gray and Siddharth Suri, Ghost Work, Houghton Mifflin, 2019.
- Fairwork Project reports at Oxford Internet Institute, 2018–present.
- Billy Perrigo, Time reporting on OpenAI and Kenyan moderators, January 2023.