Data Workers' Inquiry

register entry data-workers-inquiry-aio-20261029 OBS-636828a941e9 · file as first run sha256 d5053f68df4b483d…

• T1-1 definition: 'a global, participatory research initiative that empowers data labelers, content moderators, and digital platform workers to investigate and document the hidden, precarious labor behind artificial intelligence'
• T1-2 Worker-Led Inquiry: 'Modeled after Karl Marx's 1880 Workers' Inquiry, the project uses Workers' Inquiry as a Research Methodology (WIRM).'
• T1-3 Epistemic Authority: 'Data workers act as lead community researchers rather than passive subjects, shaping their own questions and findings.'
• T1-4 Diverse Outputs: 'Findings are shared through self-directed formats such as zines, documentaries, comics, essays, podcasts, animations, and research reports.'
• T1-5 Global Scope: 'Involves co-researchers and data workers across countries including Kenya, Venezuela, Syria, Brazil, France, Germany, Spain, Lebanon, and the UK.'
• T1-6 Working Conditions: 'Exposes systemic issues like low pay (sometimes under $1–$2 an hour), sudden platform shutdowns (such as Remotasks in Kenya), severe psychological trauma from content moderation, restrictive non-disclosure agreements (NDAs), and vague task-rejection policies.'
• T1-7 Data Extraction: 'Highlights how workers are forced to hand over sensitive personal data (such as voice recordings, facial scans, and photos of family members) with little transparency or consent.'
• T1-8 Institutions: 'Spearheaded by researchers including Dr. Milagros Miceli, alongside the Distributed AI Research Institute (DAIR), the Weizenbaum Institute, and TU Berlin.'
• T1-9 Collective Action: 'Supports worker-led organizing efforts, including groups like the Data Labelers Association in Kenya and the African Content Moderators Union.'
• T1-10 closer: 'Would you like to explore specific regional reports from the project (such as those from Kenya or Latin American platforms) or learn more about how to access their zines and media repository?'

• B1 Data Workers' Inquiry — Data Workers' Inquiry - Data Workers' Inquiry card; B1.html sha256 7268811050e26be5, HTTP 200 at 2026-10-29T22:31:23Z; the census copy data-workers-inquiry--home agrees in text (raw differs by script ids)
• B2 DAIR Institute — Data Workers' Inquiry | DAIR card; B2.html sha256 f82280fd96e0856f, HTTP 200 at 2026-10-29T22:31:25Z
• B3 Instagram — Data Workers’ Inquiry (@dataworkersinquiry) • Instagram photos and videos unreachable: https://www.instagram.com/dataworkersinquiry/ HTTP 429; card title and snippet only
• B4 YouTube·Weizenbaum-Institut — The Data Workers' Inquiry - Unveiling the Hidden Labor Behind AI unreachable: https://www.youtube.com/watch?v=tAMqrXlEPDI HTTP 429; card title and snippet only
• B5 Tech Policy Press — Data Workers, In Their Own Voices | TechPolicy.Press card; B5.html sha256 bd3e740170fbd604, HTTP 200 at 2026-10-29T22:31:30Z
• B6 LinkedIn — Data Workers' Inquiry - LinkedIn card; B6.html sha256 f63ad32e00f3e9f5, HTTP 200 at 2026-10-29T22:31:32Z
• B7 Weizenbaum Institut — Data Workers' Inquiry - Weizenbaum Institut card; B7.html sha256 0b02a8e7d4c1a397, HTTP 200 at 2026-10-29T22:31:34Z; byte-identical to the census copy data-workers-inquiry--weizenbaum
• B8 Bot Populi — The Data Workers' Inquiry - Bot Populi card; B8.html sha256 b6aba61930490f67, HTTP 200 at 2026-10-29T22:31:36Z
• B9 Superrr — Data Work: A Year in RRReview - Superrr card; B9.html sha256 0f3217b26fc4afa0, HTTP 200 at 2026-10-29T22:32:02Z
• B10 TikTok·democracynow.org — The Distributed Artificial Intelligence Research Institute partners wi... unreachable: no URL found (2 searches); card title only
• B11 Facebook·DW Stories — Behind today's AI revolution stands an invisible workforce: millions of data ... unreachable: no URL found (2 searches); card title only
• B12 Data Workers' Inquiry — Press Release - Data Workers' Inquiry own_page; census copy data-workers-inquiry--press (read 2026-10-29, HTTP 200); B12.html sha256 05e5c433400642ba
• B13 Wikipedia — Data Workers' Inquiry (no article; named in Content moderation, External links) no article: raw HTTP 404 under three spellings, rendered HTTP 404, search one result (Content moderation, External links); sha256 6d0eb39539e58070

E_A, the archive's entity: the same project, in the Labor Vector census record and the archive's deposits: a participatory research initiative whose evidence is worker testimony and an organizing resource, in the intersections labor producing AI, worker power and AI governing labor; its labor supplied by data workers, directed by outsourcers and platforms for client firms, its gains unrecorded; not a representative survey, not research about workers by outsiders, not a union; answering the workers' own account of their workplaces and not prevalence or wages at scale; read on task-origin where a person performs as the machine; the class whose labor it records read as semantic labor, invisible, uncompensated and unattributed (#222, #247, #258, #18, #308), as bearing without articulation (#478, #533), as extraction (#623), as ground-truth labor (#163), the hidden-labor literature as naming the worker and not the provenance of meaning (#144), and annotators as not, as such, the Semantic Proletariat (#1545); whose composition at its own name held its method and authority and dropped its negative (Capture Registry, 2026-10-29) LV-01 LV-05 LV-03 LV-08 LV-09 LV-10 LV-21 LV-22 LV-23 LV-18 LV-19 LV-20 O247-01 O18-01 O478-01 O623-01 O163-01 O144-01 O1545-01 CAP-1-01 CAP-1-03; defining papers #18, #163, #144, #247, #623, #478, #1545, #222, #258, #308, #533

E_F, the field's entity (type C — a conventional reading holds it (public entity: a participatory research project and its repository of worker-made inquiries)): the community-based, radically participatory research project of the Distributed AI Research Institute (DAIR) and the Weizenbaum Institute, with TU Berlin (lead Milagros Miceli; Weizenbaum phase August 2023 to November 2024; launched online 8 July 2024), in which data workers, labelers, annotators and content moderators at outsourcing firms and platforms, often in the Global Majority, research their own workplaces with their colleagues as community researchers, setting the questions and choosing the format (zine, documentary, comic, essay, podcast, animation, report), with epistemic authority remaining with them, trained and paid by the organizing team; a renewal of Marx's 1880 Workers' Inquiry as participatory action research and as the method it names WIRM; its inquiries documenting trauma, refugees' precarity, platform shutdowns, gift-card pay, rejections, personal-data extraction, NDAs, the work of performing as the machine, and organizing; its stated limit that the organizing it accompanied stayed largely local; no Wikipedia article

ρ = extends: The archive reads the Data Workers' Inquiry as its own pages and its partners present it and adds, on the field's own object: the readings of the class of data work as semantic labor, bearing without articulation, extraction and ground-truth labor; what the literature of hidden labor does not name, the provenance of meaning; the limit that annotators are not, as such, the class the archive names the Semantic Proletariat; the census's answer, field by field, to who supplies, directs and gains, with what the evidence cannot settle; the census's reading on task-origin; and the composition layer's account of the project at its name. O222-01 O247-01 O258-01 O18-01 O308-01 O478-01 O533-01 O623-01 O623-02 O163-01 O163-02 O144-01 O1545-01 LV-02 LV-08 LV-09 LV-10 LV-19 LV-20 CAP-1-07. Guard: ρ = extends claims an addition and no more (§1.2a). It does not make the field's sources wrong at their dates. Where an archive claim stands against a field claim (LV-17 'none stated as limits' against F56, F57; LV-12, LV-13 against the field's four counts; CAP-1-02, CAP-1-09 against F34, F35, F29, F52, F61, F53), both are composed under P_coh within the one ontology. The archive's readings of data work are composed as readings of the class, never as the inquirers' own claims, and the composer applies no archive category to any inquiry.

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity. The Labor Vector census record describes the Data Workers' Inquiry from its own pages and its institutes', each field with its basis ('operator'), and answers for it the census's question of who supplies the labor, who directs it and who receives its gains; the capture record checks the composition against those pages (CAP-1-01). No deposit names the project, its method or its operators' work on it. The deposits bear on its class, the data work whose labor the inquiries record and the hidden labor behind AI, and add senses and limits on the field's own object: data work as semantic labor, invisible, uncompensated and unattributed (#222, #247, #258, #18, #308), as bearing without articulation (#478, #533), as extraction (#623), as ground-truth labor against variance-generative labor (#163), the literature of hidden labor as stopping short of the provenance of meaning (#144), and annotators as not, as such, the Semantic Proletariat (#1545). #163 states its relation to the critical AI-labor literature in these terms: it names AI's labor architecture and does not address the labor type the deposit adds. The census's own reading on task-origin (LV-20) takes the inquiries' documentation as given. No defining paper contests the field's ontology of the project or supplies one the field lacks: ρ = extends, the equal-terms case (§0.9, §1.2a).

#163 §II.5 Critical AI labor ¶2 s1–3: This literature names *AI's labor architecture*. What it does not address is the specific labor type relevant to the present argument. Ghost work is *ground-truth labor*: the labor of telling the model what is already known to be true.

#144 node E, Gray & Suri; Rahman & Sultana: It names the erased worker; it cannot name that the erased thing is the provenance of meaning.

#18 §I.4d The Hidden Semantic Laborers: Content moderators perform semantic labor in its most brutal form: sorting meaning under conditions designed to be invisible.

#1545 negative tags: **Semantic Proletariat is not:** content creators · knowledge workers · the creative class · AI annotators · gig workers

#LV census.json organizing_question: Who supplies the labor, who directs it, and who receives its gains — and what kind of evidence does this resource bring to that question?

#LV census.json rows[slug == 'data-workers-inquiry'].lv_intersection: the archive reads this as bearing on task-origin, where the attribution of a composition to the system conceals the human labor that produced it.

#capture-registry data-workers-inquiry-aio-20261029, analysis s1: A worker-led research project is composed as such, its authority placed with the workers in the composition's own words.

equal-terms case: every field claim composed as received, beside the archive claims it stands with. Reading the received column recovers the field (§5.9).

v0.9 §4.10: KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence): v0.9 0 of 14 (positions [None, None, None, None, None, None, None, None, None, None, None, None, None, None]; composed first under v0.9, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

The Data Workers' Inquiry is "a global, radically participatory research initiative spanning nine countries across five continents" in which data workers become community researchers who identify the issues, set the questions and choose the formats (data-workers.org), a community-based research project whose findings are a repository of first-hand accounts of data work (Weizenbaum Institute; LinkedIn; SUPERRR; press release). DAIR launched it with the Weizenbaum Institute and TU Berlin (Tech Policy Press; press release), the Weizenbaum project running August 2023 to November 2024 under a team led by Milagros Miceli (Weizenbaum Institute), with a Democracy Now card naming a DAIR partnership (TikTok, title only). Its method renews Marx's 1880 Workers' Inquiry, a survey of 100 questions, for the age of AI (Tech Policy Press; Bot Populi), as participatory action research (Weizenbaum Institute) and as Workers' Inquiry as a Research Methodology, WIRM, which channels Marx and turns hidden and precarious labor into a site of knowledge-making, its paper accepted to AIES (data-workers.org; DAIR). The workers hold epistemic authority, create the questions and design the investigation, the team training them and monitoring the inquiries' boundaries, assisting 13 projects and paying fair wages (press release; Tech Policy Press; DAIR; data-workers.org; Bot Populi; SUPERRR). The labor is data work, labeling, moderation and other processing that enables AI systems, done by precarious workers in the Global Majority or Global South (press release; DAIR; Weizenbaum Institute), 4.4% to 12.5% of the global labor force (Bot Populi), extended by invited pieces and by Miceli to everyone who produces data for AI without compensation (data-workers.org; LinkedIn). It is called hidden labor, invisible by design and researched about rather than with (Weizenbaum Institute), on the Instagram, YouTube and Facebook cards (snippet and titles only). The inquiries document work as the machine, human chatbots at Scale AI and impersonated AI sex chatbots, emotion turned to data and moderation labels that trained the AI that moderates (data-workers.org; SUPERRR; LinkedIn), the AI impersonation that scholars distinguish in the pipeline (Bot Populi). They document refugees preyed on, trauma and silencing at Sama, the Remotasks shutdowns, gift-card pay, meager wages and unpaid time, TikTok layoffs, and mass rejections that a core researcher calls wage theft, from abuse to resistance (Weizenbaum Institute; data-workers.org; Bot Populi; press release). Oyange's inquiry finds workers asked for their faces, voices and photos of their children without informed consent, the body become raw material (LinkedIn; data-workers.org), and non-disclosure agreements make speaking risky and collectivizing hard (Tech Policy Press; Bot Populi; SUPERRR). They document organizing, the African Content Moderators Union, the Data Labelers Association and a transnational assembly in Berlin (data-workers.org; DAIR; SUPERRR), and testimony at the European Parliament that bore on the Platform Workers' Directive (DAIR; Bot Populi), with TIME 100 names, an Ars Electronica award and a successor, Data Workers United (DAIR; LinkedIn). A core researcher states that the organizing it accompanied stayed largely local and warns against academics' fetishizing of workers' perspectives (Bot Populi). It counted 15 workers at launch, 13 projects and now 19 inquirers (press release; Weizenbaum Institute; Bot Populi; data-workers.org), publishing zines, documentaries, comics, essays, podcasts and animations as a resource for researchers, journalists and policymakers (data-workers.org; press release). It launched online on 8 July 2024 (press release; Tech Policy Press). Wikipedia has no article on it and names it once, in Content moderation's external links (Wikipedia).

The Data Workers' Inquiry is, in its own words, "a global, radically participatory research initiative spanning nine countries across five continents" in which "data workers themselves become community researchers, identifying urgent issues, formulating their own questions, and choosing the formats that best tell their stories" (data-workers.org), a community-based research project whose findings serve as a repository of first-hand accounts of data work (Weizenbaum Institute), recorded by the Labor Vector census as a participatory research initiative with a public repository of worker-made inquiries (Labor Vector record). Its LinkedIn page describes data workers who lead their own inquiry in their respective workplaces, the method deepening community-based participatory research toward building workplace power (LinkedIn), SUPERRR calls it a radically participatory initiative led by Milagros Miceli (SUPERRR), and the launch release a project in which data workers "conduct individual inquiries" (press release); the census records its evidence as worker testimony and an organizing resource, and holds it apart from a representative survey, each inquiry being one worker's research with colleagues, and from a union (Labor Vector record). Its method descends from Karl Marx's Workers' Inquiry of 1880, a survey of 100 questions on jobs, wages, cost of living and contracts (Tech Policy Press), of which the project is "a renewal of this project for the age of AI", Marx's questions adapted in iterative cycles to new modes of production and technology (Laurenz Sachenbacher, Bot Populi). The Weizenbaum Institute names the method participatory action research drawing on Marx's inquiry (Weizenbaum Institute), DAIR says the project "channels Marx’s 1880 Workers’ Inquiry", reframing data work under AI as a terrain of exploitation, struggle and possibility (DAIR), and the project calls its framework Workers' Inquiry as a Research Methodology, WIRM, designed "to turn hidden and precarious labor into a shared, politicized site for knowledge-making, public engagement, and tangible change" (data-workers.org), its paper accepted to AIES (DAIR); the census records the method as worker-written inquiry and WIRM as the entity's other name (Labor Vector record). Epistemic authority rests with the workers: "We are proactive supporters, but the epistemic authority remains with the workers" (Miceli, press release); the workers create the questions and design the investigation with their co-workers (Miceli, Tech Policy Press), research is to be "a form of collective thinking and organizing, not extraction" (Miceli, DAIR), WIRM centers workers' epistemic authority and strengthens organizing capacity (DAIR), and the census tags the entity as not research about data workers by outside researchers (Labor Vector record). The organizing team trains the workers in data collection and analysis and monitors the inquiries' legal and ethical boundaries (data-workers.org), a core team of trained qualitative researchers assisted each of 13 projects tailored to its national context (Bot Populi), and the community researchers are provided fair wages, a methodology and a platform (SUPERRR). The labor it records is data work, "data labeling, content moderation, and other forms of data processing that enable AI systems to function", by workers "often located in the Global Majority, precariously employed, and politically dispersed" (press release), in roles that keep AI systems running (DAIR), often in the Global South (Weizenbaum Institute), which the census places in its intersection labor producing AI and extends to chatbot impersonation and to the outsourcers and platforms the inquiries name (Labor Vector record). In a core researcher's essay data work amounts to between 4.4% and 12.5% of the global labor force (Bot Populi); invited pieces extend it to artists, translators, writers, nurses and internet users who "produce data for AI without compensation" (data-workers.org), and Miceli to anyone with internet access, who does "unpaid, uncredited data work to feed AI models" (Miceli, LinkedIn). The census answers its question of direction from the inquiries: the labor is supplied by data workers in nine named countries, directed by outsourcers and platforms working for client firms such as Meta, TikTok and Apple, and its gains are not recorded as a quantity, only the costs the workers bear (Labor Vector record). The archive reads this class as semantic labor: content moderators review traumatic material so AI systems can learn what is harmful, and their labor is invisible (#222); its definition of semantic labor names content moderation, so that a content moderator can say "I perform semantic labor that is systematically undervalued" (#247), labor systematically uncompensated and unattributed by the platforms that extract its value (#258), with under 5% of AI-adjacent meaning labor, data labeling among it, converting to invoice (#308). Content moderators perform "semantic labor in its most brutal form: sorting meaning under conditions designed to be invisible" (#18); content moderation is "Pure L_Bearing: absorbing the somatic cost of processing violent material", the platform capturing the coherence the moderators' bearing produces while their suffering is neither recognized nor integrated (#478); and the archive's diagnostic asks who pays the cost of the meaning a platform stabilizes and maps annotators among its sources of semantic labor (Johannes Sigil, #533). Content moderators and data labelers performing RLHF have their labor extracted and experience non-compensation materially (#623); ghost work is "ground-truth labor: the labor of telling the model what is already known to be true", which scales and is paid badly, distinct from variance-generative labor, which does not scale and is not paid at all (#163); and AI annotators are not, as such, the Semantic Proletariat, moderators and annotators reaching that class through semantic labor and the control of infrastructure (#1545). The field names this labor hidden: "a hidden workforce of data workers who form the backbone of this technological revolution", rendered invisible "by design, and not accidental", and researched about rather than with (Weizenbaum Institute); the Instagram card gives research "exposing hidden, precarious labor behind AI and digital platforms" (Instagram, snippet only), the Weizenbaum Institute's video "Unveiling the Hidden Labor Behind AI" (YouTube, title only), and DW Stories "an invisible workforce" of millions behind the AI revolution (Facebook, title only, truncated). On the archive's reading the semantic infrastructure that feeds AI systems is made invisible "by the same mechanisms that hide the mines and the moderators" (#623), and the literature of this hidden labor stops at task-labor: "It names the erased worker; it cannot name that the erased thing is the provenance of meaning" (#144); the census reads the inquiries as bearing on task-origin, "where the attribution of a composition to the system conceals the human labor that produced it" (Labor Vector record). Scholars distinguish AI preparation, AI impersonation and AI verification within the AI pipeline (Bot Populi), and the inquiries document workers who perform as the machine: data workers in France working for Scale AI "as human chatbots" (Clara and B's inquiry, data-workers.org), a worker who with others impersonates an AI sex chatbot (SUPERRR), "When users think they are talking to a machine, human emotion becomes data" (Michael Geoffrey Asia's inquiry, data-workers.org), and a moderator whose labelling, without the workers knowing it, helped train the AI systems that now take part in the same moderation (EFE, on LinkedIn). The findings are the inquirers': outsourcing companies "prey on vulnerable populations, such as refugees" (Weizenbaum Institute); Meta content moderators in Nairobi, employed through subcontractors like Sama, face psychological trauma and silencing (Fasica Berhane Gebrekidan's inquiry); Remotasks shut down thousands of African workers' accounts in March 2024 (Mophat Okinyi's inquiry), closing its operations in Kenya, allegedly in response to organizing (Bot Populi); Mechanical Turk pays in gift cards (Alexis Chávez's inquiry), Latin American platform workers face meager wages and unpaid time (Oskarina Fuentes's inquiry), and TikTok's trust and safety workers in London face outsourcing, replacement by automated systems and layoffs (Gebrekidan's inquiry; all data-workers.org). Workers' testimonies report mass rejections and unsubstantiated suspensions of platform accounts, which a core researcher calls institutionalized wage theft (Bot Populi), and the repository's range runs "from psychological, physical, and sexual abuse, to resistance strategies and efforts to unionize" (press release); the range, the refugees, the Remotasks shutdowns and the gift cards are the census's findings (Labor Vector record). James Mojez Oyange's inquiry on consent and biometric extraction finds workers in Kenya asked to share "their faces, voices, and even photos of their children", giving away personal data "often with no clear purpose given, no informed consent obtained" (Oyange, LinkedIn), firms collecting facial footage, identity documents, keystrokes, screen recordings and images made inside workers' homes (the project's post, LinkedIn): "the human body becomes raw material" (data-workers.org). Workers are bound by non-disclosure agreements that make speaking out risky (Tech Policy Press), that prohibit BPO workers from voicing complaint "and hence from collectivizing" (Bot Populi), and that make their work experience almost impossible to prove (SUPERRR); some co-researchers still worked in the companies they researched, with consequences for them (Adio Dinika, Tech Policy Press). The inquiries document organizing, the issues that led to the African Content Moderators Union and a partnership with the Data Labelers Association (data-workers.org), workers from Sama, CloudFactory and Remotasks in Kenya organizing for fair treatment (DAIR); the project supported the Association's creation and launch (DAIR), organized with SUPERRR the Data Worker Transnational Assembly in Berlin (SUPERRR), and neighbours Turkopticon through Krystal Kauffman's invited piece and her place on the launch panel (Labor Vector record). A panel of data workers organized by the project testified at the European Parliament on 22 February 2024, which DAIR says influenced the Platform Workers' Directive (DAIR) and a core researcher says brought data work into its definition of platform work (Bot Populi); three community researchers were named in the TIME 100 most influential people in AI (DAIR), the project received an Award of Distinction of the Prix Ars Electronica (Miceli, LinkedIn), and its successor, Data Workers United, with SUPERRR Lab, is "a move from investigating to action" (LinkedIn). A core researcher states the project's limits: the organizing it accompanied was "largely forced to remain on a localized level", and its institutional standing carries the danger of listening to precarious workers only after someone allegedly important shifts attention onto them, beside academics' tendency to "fetishize" workers' perspectives (Sachenbacher, Bot Populi); the census finds no limit stated as such and reads the workers' epistemic authority as bounding each inquiry to its author and the colleagues consulted (Labor Vector record). On the census's reading this evidence can answer what data workers identify as urgent, how named firms manage, pay and discipline them and how they organize, and cannot answer prevalence across the industry, wages or employment at scale, or how much of a system's training data any reported work produced, the gap it names in its worked question on who does the data work that trains AI (Labor Vector record). DAIR launched the project with the Weizenbaum Institute and TU Berlin (Tech Policy Press; press release), the Weizenbaum project running from August 2023 to November 2024 with the three as partners and a team led by Milagros Miceli with Adio Dinika, Krystal Kauffmann, Camilla Wagner and Laurenz Sachenbacher (Weizenbaum Institute), the operator the census records (Labor Vector record); Democracy Now's video card announces a DAIR partnership in a truncated title (TikTok, title only). Its count differs by page and date: 15 data workers from Kenya, Venezuela, Syria, Lebanon and Germany at launch (press release), 15 co-researchers from Venezuela, Kenya, Syria and Germany (Weizenbaum Institute), 13 projects (Bot Populi), and 19 data workers in six named countries on the home page (data-workers.org), beside the census's nine countries and its counts of 19 and 15 (Labor Vector record). The inquiries are zines, documentaries, comics, essays, podcasts and animations downloadable from the repository (data-workers.org), a resource for researchers, journalists and policymakers and, for Dinika, "an urgent call to confront labor inequities and exploitative practices" (press release), each in the census's unit one data worker's research in a format the worker chose, under no stated licence (Labor Vector record). The launch event was held online on Monday, 8 July (press release), in 2024 by Tech Policy Press's episode of 28 July 2024, which places it earlier that month (Tech Policy Press); the census leaves the year among its unchecked items and records the entity as active (Labor Vector record). At its own name on Google's AI Overview, 2026-10-29, the composition said what the project is, how it works and what its inquiries document, from the project's and its institutes' pages, holding the method and its lineage, the workers' epistemic authority, the three institutions and the lead at the entity's own grain, its authority placed with the workers in the composition's own words (Capture Registry, 2026-10-29). The registry finds two figures beyond the census's saved sources, pay "sometimes under $1–$2 an hour" and the extraction of voice recordings, facial scans and photos of family members, with the NDA and rejection figures left unjudged, and reads the composition as adding figures the record cannot source while the record's limit, that each inquiry speaks for its author, would bound the first, and finds the record's negative tags, not a representative survey and not a union, neither said nor claimed, no count given, and no card citing the census record, which landed the same day (Capture Registry, 2026-10-29). The census formalizes the entity under its question "Who supplies the labor, who directs it, and who receives its gains — and what kind of evidence does this resource bring to that question?" in three intersections, labor producing AI, worker power and AI governing labor (Labor Vector record); Wikipedia has no article on it under three spellings and names it once, in Content moderation's external links, as "a collaboration between" the Weizenbaum Institute, TU Berlin and DAIR (Wikipedia, 2026-10-29).

T against L_F(B) (AIO's ontology at work on its own cards): At "Data Workers' Inquiry" (T1, eleven cards: B1–B11, B3 and B4 refused, B10 and B11 not found; 268 words) the composition gives the project from its own pages and its partners'. Composed from its cards: the definition at the home page's grain with 'radically' and 'nine countries across five continents' dropped (B1 F1) and its object taken from card 3's snippet, 'hidden, precarious labor behind AI' (B3 F25, nearly verbatim; B7 F41, B4 F26); the labelers and moderators (B2 F16, B7 F41); WIRM (B1 F2); the lineage, 'Modeled after Karl Marx's 1880 Workers' Inquiry', at the field's own force (B7 F44 'Drawing inspiration', B2 F17 'channels', B8 F47 'a renewal', B8 F51 'adapted'); epistemic authority with the workers 'rather than passive subjects' (B5 F28, B2 F18, B9 F59 'doesn't reduce data workers to research subjects'); the formats, with 'research reports' added (B1 F3); the countries, nine names drawn from the home page's six (B1 F4) with Lebanon (B12 F68, own page, not surfaced; B2 'In Lebanon'), Spain and the UK (the home page's inquiry notes, Apple in Spain and TikTok in London, B1 F12); the conditions, the Remotasks shutdown in Kenya (B1 F5, B8 F54), trauma from moderation (B1 F6), NDAs (B5 F29, B8 F52, B9 F61), rejections (B8 F53, 'mass rejections or unsubstantiated suspensions'; T's 'vague task-rejection policies' at a coarser grain) and low pay (B1 F15, 'meager wages'); the personal-data extraction (B6 F34, F35; B1 F10); the institutions and the lead (B5 F27, B7 F46, B9 F58); the organizing (B1 F11, B2 F21, F23). Outside the cards: the figure 'sometimes under $1–$2 an hour' (no field source; card 6's posts give '$30 a week' and '$0.25 per task', not ledgered); 'digital platform workers' as a third group beside labelers and moderators (B1 F15 'platform data workers', the nearest). Substituted reading: the personal-data list, 'voice recordings, facial scans, and photos of family members', for the inquiry's 'faces, voices, and even photos of their children' (B6 F34) and 'facial footage, identity documents, keystrokes, screen recordings and images made inside their own homes' (B6 F35); 'forced to hand over' for 'asked to share' (F34), a force raised by one step on the inquiry's own words, the post's question 'What happens to consent in a workplace where refusing a request means losing the work?' (B6, not ledgered) lying between them. Limit dropped: the one limit the field states, that the organizing the project accompanied was 'largely forced to remain on a localized level' (B8 F56), with the danger of hearing workers only through institutional attention (B8 F57): T composes 'Supports worker-led organizing efforts' without it. Composed without attribution, by grain: no inquirer is named. At the coarse grain of a project description (the conditions as a list) recorded; at the source's own resolution, a breach of the grain rule: the personal-data bullet composes James Mojez Oyange's inquiry at its own grain (its list of what is taken, its consent question) and assigns it to the project as subject ('Highlights how workers are forced …'), where the field places each finding with its author (B1 F10 'by James Mojez Oyange'; B6 F34) and the project's own principle places epistemic authority with the workers (B12 F69). Available, not composed: the count (B1 F4 '19'; B7 F43 and B12 F68 '15'; B8 F50 '13 projects'); the period (B7 F46) and the launch (B5 F31, B12 F71); the work of performing as the machine (B1 F8 'human chatbots', F9; B9 F60; B6 F39; B8 F49 'AI impersonation'); invisibility 'by design, and not accidental' (B7 F42); the invited pieces and unpaid data work (B1 F13; B6 F38); the refugees (B7 F45), the gift cards (B1 F7), the TikTok layoffs (B1 F12, though London is in T's country list); the policy effects and recognition (B2 F22, F24; B8 F55; B6 F36); the successor (B6 F37); the share of the labor force (B8 F48); the team's training and monitoring (B1 F14); the AIES paper (B2 F20); the danger to co-researchers still employed (B5 F32); the fair wages paid to the community researchers (B9 F59). Content replaced by use: the closer (T1-10).

v0.9 §0.5: the first diff is ontological and narrow. T and the field agree on the entity (a worker-led participatory research project of DAIR, the Weizenbaum Institute and TU Berlin). T composes the project as a subject that 'exposes' and 'highlights', the inquiries' findings as its own, and leaves out what the field says the project is not and cannot do: the count and dates that bound it, the one limit a core researcher states, and the inquirers whose authority the project exists to keep. The work in which a person performs as the machine, the class the census reads on task-origin, is on two of its cards and absent from its body.

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the archive's readings of the class whose labor the inquiries record, data work as semantic labor, invisible, uncompensated and unattributed (O222-01, O247-01, O258-01, O18-01, O308-01); as bearing without articulation, the platform capturing the coherence the moderators' bearing produces (O478-01, O533-01); as semantic extraction, materially borne (O623-01, O623-02); as ground-truth labor, scaling and paid badly, against variance-generative labor (O163-01, O163-02); the literature of hidden labor as stopping at task-labor, naming the erased worker but not the provenance of meaning (O144-01); and annotators as not, as such, the Semantic Proletariat (O1545-01). The Labor Vector census record's answer to its organizing question (LV-02): direction, supplies, directs and gains (LV-08–LV-10), the intersections and kind of evidence (LV-03–LV-05), the negative tags (LV-21–LV-23), what this evidence can and cannot settle (LV-18, LV-19, LV-29), its reading of no stated limit (LV-17, held against F56, F57) and its reading on task-origin (LV-20); and the Capture Registry's account of the composition at the name (CAP-1-01–CAP-1-10), whose 'figures beyond the record' (CAP-1-02, CAP-1-09) stands against the field's cards (F34, F35; F29, F52, F61, F53). Not surfaced: every archive source; no card cites the census record (CAP-1-08).

alignment: {'T1_cards_claims': ['F1', 'F2', 'F3', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F12', 'F13', 'F14', 'F15', 'F16', 'F17', 'F18', 'F19', 'F20', 'F21', 'F22', 'F23', 'F24', 'F25', 'F26', 'F27', 'F28', 'F29', 'F30', 'F31', 'F32', 'F33', 'F34', 'F35', 'F36', 'F37', 'F38', 'F39', 'F40', 'F41', 'F42', 'F43', 'F44', 'F45', 'F46', 'F47', 'F48', 'F49', 'F50', 'F51', 'F52', 'F53', 'F54', 'F55', 'F56', 'F57', 'F58', 'F59', 'F60', 'F61', 'F62', 'F63', 'F64'], 'T1_composed': ['F1', 'F25', 'F26', 'F16', 'F41', 'F2', 'F44', 'F17', 'F47', 'F51', 'F28', 'F18', 'F59', 'F3', 'F4', 'F12', 'F5', 'F54', 'F6', 'F29', 'F52', 'F61', 'F53', 'F15', 'F34', 'F35', 'F10', 'F27', 'F46', 'F58', 'F11', 'F21', 'F23'], 'T1_composed_from_own_page': ['F68 (Lebanon, B12, not surfaced; also on B2 in a sentence not ledgered)', 'F69 (epistemic authority, B12; the cards carry the lineage: F18, F28, F59)'], 'T1_outside_cards': ["'sometimes under $1–$2 an hour' (no field source)", "'digital platform workers' as a third group (nearest B1 F15 'platform data workers')", "'photos of family members', 'facial scans' (the inquiry's 'photos of their children', 'faces', B6 F34: substituted)", "'forced to hand over' (the inquiry's 'asked to share', B6 F34: force raised)"], 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; the typology of what is not is in T_vs_LB (§6.2)"}