REVIEW 4 major objections 4 minor 4 references
Exploitation All the Way Down: Calling out the Root Cause of Bad Online Experiences for Users of the "Majority World"
T0 review · 4 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash
Pith's one-line read This essay argues that the common root of the online harms faced by users from the Global Majority is the continued exploitation and disregard for their well-being by large tech companies and outside states.
desk verdict A clear, well-cited critical essay whose 'exploitation all the way down' framing is more an interpretive lens than a demonstrated root cause, but it deserves a serious referee as a policy intervention. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The carrying mechanism is the "cycle of harm," a four-stage account that connects user-facing harm, moderator trauma and exploited labor, exclusion from technological benefits, and a false dichotomy between surveillance and neglect. The authors use this cycle, together with a critique of blanket terminology, to argue that exploitation is the root cause rather than a side effect. The cycle is what makes the argument cohere: each stage is treated as a manifestation of the same structural neglect, and the term "Global Majority" is examined as a lens that both mobilizes and obscures.
What would settle it
A representative cross-platform audit of moderation outcomes across many Global Majority languages, controlling for how much each platform invests in moderating that language and how moderators are compensated, would settle the claim: if harm rates track resource levels rather than exploitative labor arrangements, the root-cause thesis fails.
Extended reading notes
Core claim
The paper's central claim is that the common denominator tying the experiences of Global Majority people is "the continued exploitation and disregard for well-being by Big Tech and states outside of the Global Majority." The authors present a cycle of harm with four linked faces: users are exposed to harmful content because moderation is under-enforced and under-resourced in their languages; people from these communities are hired as content moderators and subjected to traumatic, poorly compensated work; their labor is used to train models that do not serve their languages or are unavailable in their countries; and existing moderation alternatives force a false choice between being surveilled or being left unprotected. The essay further argues that this structure resembles colonial exploitation and that the term "Global Majority," while useful for reclaiming power, obscures internal diversity that makes uniform policies ineffective. The remedy proposed is not to extend current Global Minority policies to previously ignored communities, but to support community-driven interventions and participatory design.
Load-bearing premise
The load-bearing premise is that the cited cases and statistics are representative of the broader Global Majority; if those examples are atypical, or if under-moderation stems more from resource constraints than exploitative intent, the causal claim loses its basis.
Editorial extensions
If this is right
- If exploitation is the root cause, then improving automated detection models for more languages will not solve the problem by itself; the labor conditions and policy enforcement structures also need to change.
- Hiring moderators from the Global Majority without fair compensation and support cannot count as inclusion; it extends the same cycle of harm.
- Extending existing Western-derived moderation policies to Global Majority communities would likely reproduce harm, since the term hides important differences in class, ethnicity, and political context.
- Effective interventions should start from community organizing and participatory design rather than from platform-centric policy.
- The surveillance-versus-neglect framing implies that any moderation solution that increases surveillance without addressing exploitation will fail the communities it claims to protect.
Reading between the lines
- A testable consequence the paper does not spell out: if the exploitation thesis is right, then a cross-platform audit controlling for language and region should find that harm rates and moderation failures are predicted by the exploitativeness of labor practices (e.g., third-party subcontracting) rather than by raw resource levels.
- The cycle-of-harm account suggests a concrete design target: platform policies that combine community-owned moderation with transparent appeals could break the false dichotomy between surveillance and neglect, a direction the paper gestures at but does not fully develop.
- The "Global Majority" terminology critique implies that future research should disaggregate outcomes by country, class, and ethnicity; a useful next step would be to build a comparative dataset of moderation harms across several Global Majority communities to test which layers of difference matter most.
- The paper's colonial-continuity analogy could be extended to labor externalization in other industries; if the moderation case follows the same pattern as outsourced manufacturing, we would expect repeated cycles of subcontracting, union-busting, and relocation to lower-cost regions.
Signed reviews
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. This essay argues that the many harms experienced by "Global Majority" users and workers on online platforms—content-moderation failures, traumatic and exploitative labor conditions, surveillance risks, and exclusion from AI benefits—are not separate technical or policy failures but manifestations of one underlying cause: continued exploitation and neglect by Big Tech and by states outside the Global Majority, understood as continuous with colonial extraction. The paper draws on reported cases (Facebook in Myanmar and Ethiopia, WhatsApp misinformation in India, Kenyan and Ethiopian moderator labor, and the YouTube study of Ethiopian users), industry statistics (Meta's 87% US-focused moderation, YouTube's 89.2% English moderators), and prior scholarship on digital colonialism, and it closes by calling for nuance within blanket terminology, participatory design, and engagement with community organizations.
Significance. If the paper's framing were established, it would meaningfully shift policy and research attention from isolated moderation failures to the structural political economy that connects labor exploitation, underenforcement, and user harm. The essay's strength is its synthesis: it brings together labor-rights cases, user-facing harms, and language-technology exclusions in one narrative, and it explicitly acknowledges that the Global Majority is internally diverse—an important corrective to uniform policy proposals. The published venue (a UN IGF coalition report chapter) gives it policy visibility. The paper is also honest about the limits of its own proposals, grounding them in participatory design rather than top-down regulation. That said, it is an essay rather than a systematic study, and its central causal claim about a single root cause is asserted more than demonstrated.
major comments (4)
- [The Cycle of Harm in Moderation and Inclusion] The paper's central claim—that "at the center of this cycle of harm is the exploitation and neglect of the wide swath of communities"—is asserted but not established with a defined causal mechanism. The text shifts between two senses of exploitation: surplus extraction from low-paid moderators (Kenyan lawsuits, ChatGPT data labeling) and the broader neglect of non-English-speaking users (English-only moderators, resource scarcity, enforcement failures). These are different mechanisms with different beneficiaries; for example, the under-provision of moderation for low-resource languages is more parsimoniously explained by cost minimization and missing training data than by an intent to exploit. A root-cause claim must name the mechanism, identify the beneficiary in each case, and specify a contrast class of plausible non-exploitative explanations that are ruled out. Without that, the central claim is underdetermined rather than demonstrated.
- [What Lies Under Blanket Terminologies?] The essay explicitly states that "there is no single 'AI from the Global Majority' because the 'Global Majority' is many" and that class, ethnicity, and political context create different layers of harm. This acknowledgment raises the evidential burden for the preceding claim that a shared mechanism—exploitation—ties all these communities' experiences together. If the realities of, say, an Ethiopian domestic worker in Lebanon, an Ethiopian software engineer in the US, and an Indigenous community on occupied land are as different as the paper says, the reader needs a specification of how each harm traces to the same exploitative root rather than to distinct local conditions. The paper should either reconcile this tension by identifying the common causal pathway at the right level of abstraction or reframe the claim as an interpretive hypothesis rather than a root-cause finding.
- [False Dichotomies of Harm] The section on "False Dichotomies of Harm" asserts that current moderation alternatives present a false choice between surveillance and neglect, calling surveillance an "inevitable consequence" of effective moderation. However, the alleged dichotomy is never precisely defined, and no argument is given for why these two options exhaust the relevant space or why alternatives such as community-based moderation, cooperative governance, or non-profit infrastructure are excluded. The rhetorical question "If the premise of moderation was not capitalistic and exploitative, could we have safer online experiences?" presupposes exactly the exploitative premise that the essay is trying to establish. This is a load-bearing gap, because the "exploitation all the way down" thesis depends on showing that the observed harms are not merely the result of resource constraints or poor implementation, but of an exploitative structure that makes those the only options.
- [Introduction and Discussion] The evidence is curated rather than systematic. The paper relies on a small set of case studies (Ethiopian YouTube users, Kenyan moderators, Facebook in Myanmar/Ethiopia, WhatsApp in India) and selected statistics (90% of users outside the US/Canada, 87% of moderation time on US posts, 89.2% of moderators operating in English, 60% higher regret exposure for non-English users). These are cited without addressing whether they are representative of the Global Majority as a whole. Given that the paper itself insists on the diversity of the Global Majority and criticizes blanket terminology, the inference from selected examples to a single root cause needs explicit justification—or the claim should be softened to a common pattern rather than a root cause. At minimum, the paper should acknowledge the selection basis and state what kind of evidence would count against the root-cause claim.
minor comments (4)
- [Title] The title as printed in the manuscript header reads "Calling out theRoot Cause"—missing a space—and the footnote contains "Global Majoirty." These should be corrected.
- [References] The in-text citation "Dias Oliva, 2020" does not match the reference list entry "Dias Oliva, T., Antonialli, D. M., & Gomes, A. (2021)"; please harmonize the year and author list.
- [Introduction] The sentence "which has lead to a lack of data" should read "led" rather than "lead."
- [What Lies Under Blanket Terminologies?] The phrase "users resort toVPNservicesto access the platforms" is missing spaces around "VPN" and reads as a typographical error rather than an intentional formatting choice.
Circularity Check
No significant circularity: the essay argues a thesis from external evidence and contains no derivation that reduces to its own inputs.
full rationale
This essay makes no quantitative predictions and contains no fitted parameters, equations, or uniqueness arguments. Its central claim—that exploitation is the common tie among Global Majority users' online harms—is a thesis argued from independently cited evidence (e.g., moderator working-condition reports, the YouTube Regrets study, Facebook's America-centric moderation statistics, and the Myanmar/Ethiopia moderation failures). The self-citation to Nigatu & Raji (2024) is used as an empirical illustration of Ethiopian women YouTube users' experiences; it is an independent interview study reporting user harm, and it does not simply restate the essay's conclusion. The introduction's description of the Global Majority as sharing 'a common predicament... exploitation' is a statement of the essay's thesis, not a definitional input from which the conclusion is mechanically derived. Thus no step reduces by construction to its own inputs. Possible weaknesses in causal identification—such as not ruling out resource constraints or ordinary cost minimization as alternative explanations for under-moderation—are correctness or evidentiary risks, not circularity.
Assumptions & free parameters
assumptions (3)
- domain assumption The cited news reports and studies (Perrigo 2022, Tworek 2021, Nigatu & Raji 2024) accurately describe platform practices.
- domain assumption The historical analogy between colonial exploitation and contemporary tech labor is valid and explanatory, not merely rhetorical.
- domain assumption The category 'Global Majority' is analytically meaningful for policy despite internal heterogeneity.
Cite this review
Pith. "Pith review of Exploitation All the Way Down: Calling out the Root Cause of Bad Online Experiences for Users of the "Majority World"." pith.science (2026). https://pith.science/paper/CSKXITWN
@misc{pith2026250114748,
author = {Pith},
title = {Pith review of: Exploitation All the Way Down: Calling out the Root Cause of Bad Online Experiences for Users of the "Majority World"},
year = {2026},
howpublished = {\url{https://pith.science/paper/CSKXITWN}},
note = {Machine review of arXiv:2501.14748}
}
read the original abstract
Global Majority users are exposed to multitudes of harm when interacting with online platforms. This essay illuminates how exploitation in the advances of Artificial Intelligence is tied to historical exploitation and how the use of blanket terminology overshadows the layers of exploitation and harm ``Global Majority'' populations face. It first discusses the multitude of harm content moderators from the Global Majority face, arguing against the current trend of protection through exploitation, then it illustrates the nuances and differences within the Global Majority, and finally, it outlines actionable items to move away from such harm.
Reference graph
Works this paper leans on
-
[1]
Exploitation All the Way Down: Calling out theRoot Cause of Bad Online Experiences forUsersof the“MajorityWorld.” 1 HellinaHailuNigatu, UCBerkeley, USA 2 ZeerakTalat, Edinburgh University, UK AbstractGlobal Majority users are exposedtomultitudesof harmwheninteractingwithonlineplatforms. This essay illuminates how exploitation in the advances of Artificial...
-
[3]
WEIRDFAccTs: HowWestern,Educated, Industrialized, Rich, andDemocraticisFAccT?InProceedingsof the2023ACMConferenceonFairness, Accountability, andTransparency(FAccT'23). Associationfor ComputingMachinery, NewYork, NY, USA, 160–171.https://doi.org/10.1145/3593013.3593985 Shahid, F. &Vashistha, A. (2023) DecolonizingContent Moderation: DoesUniformGlobal Commu...
arXiv 2023
-
[4]
Popular YouTubechannelsproducedavast amount ofcontent, muchof it inlanguagesother thanEnglish. PewResearchCenter. Retrievedfromhttps://www.pewresearch.org/internet/2019/07/25/popular-youtube-channels-produced-a-vast-amount-of-content-much-of-it-in-languages-other-than-english/#:~:text=Meanwhile%2C%2067%25%20posted%20videos%20exclusively,the%20first%20week...
work page 2019
-
[2023]
UDigital ServicesAct (EUDSA) Biannual VLOSE/VLOPTransparencyReport. TechnicalReport. https://storage.googleapis.com/transparencyreport/report-downloads/pdf-report-27_2023-8-28_2023-9-10_en_v1.pdf Horwitz, J. TheFacebookFiles. (2021, October 01). TheWall Street Journal. Retrievedfromhttps://www.wsj.com/articles/the-facebook-files-11631713039 Kaffee, L.-A.,...
arXiv 2023
Reviewed August 11, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.