METAXIS Intelligence Workbench (Operator)

AI_LABOR -- Report report_intel_dbab7b01963396b5_v3

version=3 readiness=CONDITIONALLY_READY generated_at=2026-10-02T05:18:27.506330+00:00

Intelligence Object: intel_dbab7b01963396b5 (IO version 3, readiness CONDITIONALLY_READY)

Provenance Inspector for this report →

Current State

US labor force participation rate (World Bank SL.TLF.ACTI.ZS) ranged 71.58-73.02% over 2010-2022. Trend classification: STABLE. No evidence yet connects this aggregate movement specifically to AI adoption -- a general participation-rate change is not treated as an AI effect. [O-2 round 4 update] Real evidence has since been added beyond the World Bank participation-rate series: (1) OECD AI Exposure Measure methodology -- exposure/capability-gap only, explicitly NOT displacement; (2) US BLS JOLTS Aug 2026 aggregate labor-flow data, little-changed month-over-month, no AI attribution in the release; (3) NBER w31161 field experiment: generative-AI assistant raised customer-support productivity 14% on average (34% for novices), a firm-level productivity finding, NOT employment/headcount evidence; (4) Stanford SIEPR 'Canaries in the Coal Mine': 13% relative employment decline for early-career (22-25) US workers in high-AI-exposure occupations since generative AI adoption -- direct observed employment-effect evidence, but scoped narrowly to that cohort; (5) KDI Korea macro FORECAST (not observed): +3.5% TFP over 10 years, ~256,000 jobs/year displaced; (6) ETLA/Finland peer-reviewed population-level study found NO statistically significant wage/employment divergence by AI exposure in the 2 years after ChatGPT's launch -- genuine counterevidence to a broad displacement reading; (7) Fed FEDS Note: US firm/worker AI-adoption rates (18% of firms by end-2025; 41% of workforce using generative AI at work; adoption only, not a productivity or employment-effect measurement). Four real Hypotheses (HL1-HL4) were constructed this round from this evidence base via the canonical determine_canonical_status() write path: HL1 (US AI adoption is measurably increasing) = PARTIALLY_SUPPORTED (single Fed synthesis source, source-tier not yet classified -> UNKNOWN-tier guard fires); HL2 (direct observed employment effect concentrated in early-career US workers) = CONTESTED (Stanford SIEPR support vs. ETLA/Finland null-result contradiction -- a genuine geography/population tension, surfaced not resolved, with the OECD exposure!=displacement framing recorded as a generalization guard); HL3 (measurable productivity effect scoped strictly to generative-AI customer support) = PARTIALLY_SUPPORTED (same UNKNOWN-source-tier guard as HL1); HL4 (AI is driving large-scale, economy-wide job displacement beyond the early-career cohort) = WEAKENED (KDI's 10-year forecast is not yet corroborated, and is actively weakened, by observed aggregate US BLS data showing no AI-attributed change and by Finland's observed null result -- a forecast-vs-observed divergence, not a rejection, since the forecast horizon has not elapsed and geography differs). [O-2B update] 7 additional real claims added (Lightcast wage premium, NBER w33509 firm-level AI-utilization/pay correlation, Atlanta Fed skill-demand-by-education diffusion, ECB euro-area adoption/no-employment-reduction survey, Korea Employment Information Service occupational outlook, Axios/Chicago-Fed non-AI alternative explanation for the 2022-2025 US job-growth slowdown, and Vanguard broad-occupation counterevidence to mass displacement). Of these, the Vanguard counterevidence and the Korea Employment Information Service outlook were wired as new CONTRADICTING evidence on HL4 (economy-wide displacement), and the Axios alternative explanation was added to HL4's alternative_explanations, via the real determine_canonical_status()/upsert_hypothesis() pipeline -- HL4 recomputed and remained WEAKENED (no status flip). The ECB claim was deliberately left UNWIRED from HL4 given HL4's explicit US/Korea/Finland geographic scope; the Lightcast and Atlanta Fed skill-demand claims were deliberately left UNWIRED from HL1 as adoption-ADJACENT rather than direct adoption-rate evidence. A first Synthesis document for this IO was also authored this round (o2b_synthesis_intel_dbab7b01963396b5.json), covering OBSERVED/FORECAST/ADOPTION/EXPOSURE/PRODUCTIVITY/EMPLOYMENT/WAGES/SKILLS/GEOGRAPHIC-DIFFERENCES/TEMPORAL-LIMITS/COUNTEREVIDENCE/ALTERNATIVE-EXPLANATIONS/ATTRIBUTION-LIMITS/WHAT-WE-KNOW/WHAT-WE-DO-NOT-KNOW, grounded only in this IO's existing 15 claims and HL1-HL4.

Hypotheses (canonical status)

CodeStatementCanonical StatusID
H?AI adoption is measurably shifting US labor force participationINSUFFICIENT_EVIDENCEhyp_f5f71f3a181d830c
HL1AI adoption among US firms and workers is measurably increasingPARTIALLY_SUPPORTEDhyp_b365b95148a3a372
HL2There is direct observed employment effect from AI exposure, concentrated specifically in early-career (ages 22-25) US workersCONTESTEDhyp_b3a8bfef91a47a45
HL3AI shows a measurable productivity effect in specific, well-studied task contexts (generative-AI-assisted customer support)PARTIALLY_SUPPORTEDhyp_c4b0429622eb6dd7
HL4AI is driving large-scale, economy-wide job displacement at the national level (beyond the narrow early-career cohort)WEAKENEDhyp_8691a1cb2d7c0d45

Claims (Key Claims) AVAILABLE

근거 / Research Evidence

empty / NOT_AVAILABLE

Statistics

Observation

Forecast

none

반증 자료 / Counterevidence AVAILABLE

Alternative Explanations AVAILABLE

Geographic Applicability AVAILABLE

Temporal Applicability AVAILABLE

Known Gaps AVAILABLE

Sources (출처) -- 15 actually used

Claim IDInstitutionTitleYearTierObservation/ForecastAccess StatusReal URLLink Verification
claim_4765a68987d86c67World BankWorld Development Indicators -- Labor force participation rate, total (SL.TLF.ACTI.ZS), United States2010-2022 (data series)PRIMARYOBSERVATIONOFFICIAL_DATA_PAGEhttps://api.worldbank.org/v2/country/USA/indicator/SL.TLF.ACTI.ZS?date=2010:2022&format=json&per_page=100UNVERIFIED/ACCESS_BLOCKED -- re-attempted O-2B 2026-10-02: WebFetch again refused with PROVENANCE_REQUIRED, and a direct curl retry through the sandbox agent proxy failed with CONNECT tunnel 403 (organization policy denies egress to api.worldbank.org). Two independent access paths tried, both blocked; left UNVERIFIED rather than forced to VERIFIED.
claim_b9370aa7d219d791OECDThe OECD AI exposure measure: Mapping the OECD AI Capability Indicators to occupations2024PRIMARYOBSERVATION of occupational EXPOSURE only -- NOT a displacement or job-loss measurement; the OECD explicitly states actual displacement depends on adoption, regulation, organisational change and social choice.ORIGINAL_SOURCEhttps://www.oecd.org/en/publications/the-oecd-ai-exposure-measure_f3da0f0a-en.htmlVERIFIED by WebFetch 2026-10-02: title and exposure-not-displacement framing confirmed.
claim_4731da53eae6b843U.S. Bureau of Labor Statistics (BLS)Job Openings and Labor Turnover Survey (JOLTS) News ReleaseAugust 2026 data, released September 29, 2026PRIMARYOBSERVATIONORIGINAL_SOURCEhttps://www.bls.gov/news.release/jolts.nr0.htmVERIFIED by WebFetch 2026-10-02: title, date, and job-openings/hires figures confirmed.
claim_fe878b52922b46a4National Bureau of Economic Research (NBER); Brynjolfsson, Li & RaymondGenerative AI at Work (NBER Working Paper No. 31161)2023 (issued April 2023, revised November 2023)ACADEMICOBSERVATION -- a firm-level PRODUCTIVITY (output-per-hour) finding only, not evidence of employment/headcount change.WORKING_PAPERhttps://www.nber.org/papers/w31161VERIFIED by WebFetch 2026-10-02: title, authors, and 14% productivity figure confirmed. Full paper text is gated behind NBER's working-paper paywall; the abstract/landing page itself is openly accessible, hence WORKING_PAPER rather than ORIGINAL_SOURCE.
claim_8fd5da8f27c9fc2bStanford Institute for Economic Policy Research (SIEPR); Brynjolfsson, Chandar & ChenCanaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial IntelligenceAugust 2025ACADEMICOBSERVATION -- a direct employment-effect finding SCOPED ONLY to early-career (ages 22-25) US workers in high-AI-exposure occupations; must not be generalized to the whole labor market.ORIGINAL_SOURCEhttps://siepr.stanford.edu/publications/working-paper/canaries-coal-mine-six-facts-about-recent-employment-effects-artificialVERIFIED by WebFetch 2026-10-02: title, authors, and scope confirmed.
claim_30fdf98bc1a9c818Korea Development Institute (KDI), reported by Newsis인공지능(AI)의 거시경제적 영향 분석 (Analysis of AI's Macroeconomic Impact) -- as covered by NewsisJuly 22, 2026SECONDARYFORECAST / PREDICTIVE -- a 10-year macroeconomic PROJECTION, not an observed/realized outcome; must never be cited as grounds for an already-happened job-loss claim.SECONDARY_SOURCEhttps://mobile.newsis.com/view/NISX20260722_0003718118VERIFIED by WebFetch 2026-10-02: article content and KDI attribution confirmed. This is the news article, not the KDI report itself -- labeled SECONDARY_SOURCE per 33B/33E, not promoted to PRIMARY.
claim_8e305d0c30d23514ETLA (Research Institute of the Finnish Economy); Kauhanen & RouvinenAssessing Early Labour Market Effects of Generative AI: Evidence from Population Data (Applied Economics Letters)June 2025ACADEMICOBSERVATION -- COUNTEREVIDENCE to displacement: no statistically significant wage/employment difference found by occupational exposure.ORIGINAL_SOURCEhttps://www.etla.fi/julkaisut/akateemiset-julkaisut/assessing-early-labour-market-effects-of-generative-ai-evidence-from-population-data.mdVERIFIED by WebFetch 2026-10-02: title and authors confirmed.
claim_0da20db21fe1463aFederal Reserve Board; Jeffrey S. AllenMonitoring AI Adoption in the U.S. Economy (FEDS Notes)April 3, 2026PRIMARYOBSERVATION -- ADOPTION rate data only, not a productivity or employment-effect finding.ORIGINAL_SOURCEhttps://www.federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-s-economy-20260403.htmlVERIFIED by WebFetch 2026-10-02: title, author, date, and survey figures confirmed.
claim_46bf083230ad547eUNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://lightcast.io/resources/blog/beyond-the-buzz-press-release-2025-07-23NOT_TESTABLE -- no curated verification record
claim_67e1764a3b61ca91UNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://nber.org/papers/w33509NOT_TESTABLE -- no curated verification record
claim_c195b340a92a8d90UNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://www.frbsf.org/research-and-insights/publications/system-research-atlanta-fed/2025/08/by-degrees-measuring-employer-demand-ai-skills-educational-requirements/NOT_TESTABLE -- no curated verification record
claim_ea807921100c2653UNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://www.ecb.europa.eu/press/blog/date/2026/html/ecb.blog20260826~e1c1a89999.en.htmlNOT_TESTABLE -- no curated verification record
claim_cfb369e965025922UNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://www.venturesquare.net/1085101/NOT_TESTABLE -- no curated verification record
claim_25cb5dc45e396783UNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://www.axios.com/2025/08/19/trump-immigration-jobs-growthNOT_TESTABLE -- no curated verification record
claim_1a69cf2e8f4f0f92UNKNOWN -- not in curated source registryUNKNOWN -- not in curated source registryUNKNOWNUNKNOWNUNKNOWNACCESS_BLOCKEDhttps://www.axios.com/2025/12/17/ai-jobs-market-wagesNOT_TESTABLE -- no curated verification record

Reports (version history)

VersionGeneratedReadinessSourcesClaimsEvidence Snapshot.json.pdf_operator.html_print.html_product_public.html
v12026-10-01T05:37:21.781785+00:00CONDITIONALLY_READYUNKNOWNUNKNOWNfefe68ccc8afa64aOKOKMISSINGOKOK
v22026-10-02T02:56:50.510890+00:00CONDITIONALLY_READYUNKNOWNUNKNOWNfefe68ccc8afa64aOKOKOKOKOK
v3 (this page)2026-10-02T05:18:27.506330+00:00CONDITIONALLY_READYUNKNOWNUNKNOWNfefe68ccc8afa64aOKOKOKOKOK

Version Diffs (via report_engine.diff_reports())

TransitionDiff Types
-- → v1BASELINE
1 → v2CLAIM_STATUS_CHANGED, GAP_ADDED
2 → v3CLAIM_STATUS_CHANGED

Provenance

Full Report → IO → Hypothesis → Claim → Evidence → Source chain →