ITEA Framework — Index of Transformation and Algorithmic Expropriability

🌐 English · Español · Português · 中文

A multidimensional framework for measuring occupational algorithmic expropriability under the Agentic AI regime.

ITEA measures algorithmic expropriability — the surface of human knowledge that AI can codify and capture from work. Its distinctive move is to separate what can be measured (a potential: expropriability = exposure × codifiability) from the normative event it enables — the transfer of the worker's sovereignty over the exploitation of that knowledge — and from its two consequences: the firm's appropriation of value and the dilution of market exclusivity. This layered view — a worker still possesses what she knows (non-rival), but may lose the right to decide how it is exploited — is what distinguishes ITEA from a single exposure score.

Understand the reasoning: The research narrative and the why of each decision · Data guide and glossary

The ITEA Framework (v3.0) implements this with ten complementary indicators covering 1,016 SOC 6-digit occupations from O*NET 30.2, designed to measure occupational algorithmic expropriability in the Agentic AI regime — distinct from but complementary to task-level automation indices in the Frey-Osborne (2017) tradition or ability-based exposure measures such as the AIOE benchmark of Felten, Raj and Seamans (2021).

External convergent validation against the AIOE benchmark over 769 common SOC 6-digit occupations: r(OAEI v3.0+, AIOE) = 0.797.

Coverage note. OAEI is available for 878 of the 1,016 occupations (86.5%). The 138 without OAEI lack the GEE/wage inputs it composes; among them, the 19 military occupations (SOC 55) are a structural O*NET exclusion (no Job Zone/wage). Details in the dataset review.

This repository advances from v1.45 (October 2025) to v3.0 (April 2026), archiving the previous release as the v1.45-legacy tag. The transition is not cosmetic. It reflects a major methodological revision motivated by a documented case study — three research papers (the trilogy 8A/8B/8C) that put the framework into active empirical use and, in doing so, exposed three specific limitations of v2.1 (the methodological line that paralleled v1.45). Documenting this evolution explicitly is part of the framework's preparation for submission to the Journal of Open Source Software (target window: mid-2026), which values evidence of community use and principled software evolution.

Three research papers in active circulation as of April 2026 motivated each of the three central revisions in v3.0:

• Paper 8A — "The Structural Flaw: Industrial Labour Contract Inadequacy under Agentic AI" (García-Lluis Valencia, 2026a). Targeted journals: Journal of Institutional Economics, Industrial and Corporate Change. → motivated the z-score ITEA aggregation.
• Paper 8B — "Beyond the Pigouvian Trap: Tokenised Intellectual Capital..." (García-Lluis Valencia, 2026b). Theoretical paper formalising the Lindbeck–Snower protective condition dissolution. → motivated the dual-variant OAEI (multiplicative for theoretical alignment + additive for criterion validity).
• Paper 8C — "QE and Senior Workforce Restructuring: A Labour Liability Transmission Channel" (García-Lluis Valencia, 2026c). Targeted journal: Labour Economics. Empirical paper on a novel monetary-policy transmission channel. → motivated the triple-residualised IRA.

The full case-study reasoning, including how each paper's specific empirical finding pushed the methodological decision, is documented in §4.4, §5.4 and §6.5 of the Consolidated Methodology document.

The v1.x line (March 2024 – October 2025) was the operational release cycle: data workbooks, R/Python implementation code, Streamlit dashboard. The v2.x line (April 2026) was the methodological documentation cycle: academic papers anchoring the framework's psychometric foundations. Until this release, the two lines progressed asynchronously — a researcher citing "ITEA v1.45" referred to the same conceptual framework as one citing "ITEA v2.1", but the precise indicator formulas, source data versions, and validation tables were not trivially comparable.

v3.0 unifies both numbering lines onto a single version. The operational repository advances to v3.0 with the v1.45 release archived as a permanent legacy tag for replication of any results published using the v1.x specification. The methodological documentation advances to a single Consolidated Edition that supersedes both v2.0 and v2.1.

Migration from v1.45 is documented step-by-step in docs/MIGRATION.md. Researchers who wish to reproduce results using the v1.x specification can check out the v1.45-legacy tag and follow the version-specific README preserved in that snapshot.

source("code/v3/itea_functions_v3.R")

# Load processed data
library(readxl)
df <- read_excel("data/processed/ITEA_v3_0_Workbook.xlsx",
                 sheet = "ITEA_INDICATORS_v3.0", skip = 2)

# Compute ITEA v3.0 (z-score aggregation)
df$ITEA_v30_check <- itea_v3(df$EAC, df$EIG, df$EIA)

# Compute IRA v3.0 (triple residualisation)
df$IRA_v30_check <- ira_v3(df$CA, df$`IRO v1.45 (legacy)`,
                            df$`GEE v2.1`, df$ITEA_v30_check, df$ICT)

# Compute OAEI v3.0 multiplicative (canonical)
df$OAEI_v30_check <- oaei_v3_mult(df$ITEA_v30_check, df$`GEE v2.1`,
                                    df$ICT, df$IPI)

# Compute OAEI v3.0+ additive (alternative)
df$OAEI_v30_plus_check <- oaei_v3_add(df$ITEA_v30_check, df$`GEE v2.1`,
                                       df$ICT, df$IPI)

from code.v3.itea_functions_v3 import itea_v3, ira_v3, oaei_v3_mult, oaei_v3_add
import pandas as pd

df = pd.read_excel("data/processed/ITEA_v3_0_Workbook.xlsx",
                   sheet_name="ITEA_INDICATORS_v3.0", skiprows=2)

df['ITEA_v30_check'] = itea_v3(df['EAC'], df['EIG'], df['EIA'])
df['IRA_v30_check'] = ira_v3(df['CA'], df['IRO v1.45 (legacy)'],
                              df['GEE v2.1'], df['ITEA_v30_check'], df['ICT'])
df['OAEI_v30_check'] = oaei_v3_mult(df['ITEA_v30_check'], df['GEE v2.1'],
                                     df['ICT'], df['IPI'])
df['OAEI_v30_plus_check'] = oaei_v3_add(df['ITEA_v30_check'], df['GEE v2.1'],
                                         df['ICT'], df['IPI'])

Both implementations are designed to reproduce the values stored in OAEI v3.0 and OAEI v3.0+ additive columns of the data workbook (Δ < 1e-6 over the 1,016 occupations).

ITEA-Framework/
├── README.md                          # This file
├── README_ES.md · README_PT.md · README_ZH.md   # Multilingual versions
├── LICENSE                            # MIT
├── CITATION.cff                       # GitHub-native citation metadata
├── CHANGELOG.md                       # Full v1.0 → v3.0 change history
├── MANIFEST.json                      # SHA-256 checksums for Zenodo deposit
├── .gitignore
│
├── data/
│   ├── processed/
│   │   └── ITEA_v3_0_Workbook.xlsx          # Primary v3.0 dataset (12 sheets)
│   └── legacy/
│       └── Research_Data_Workbook_ITEA_v1.45.xlsx   # Preserved for v1.x replication
│
├── code/
│   ├── v3/
│   │   ├── itea_functions_v3.R              # R reference implementation
│   │   ├── itea_functions_v3.py             # Python reference implementation
│   │   └── tests/                           # Unit tests reproducing the workbook values
│   └── legacy/
│       ├── itea_functions_v1.45.R           # Preserved v1.x implementation
│       └── itea_functions_v1_45.py
│
├── docs/
│   ├── ITEA_v3_0_Consolidated_Methodology.pdf   # Single authoritative methodology document
│   ├── MIGRATION.md                              # v1.45 → v3.0 step-by-step migration
│   └── papers/
│       ├── README.md                             # Pointers to trilogy papers (Zenodo DOIs)
│       └── trilogy_case_study.md                 # Standalone case study
│
└── paper/                                        # JOSS submission
    ├── paper.md                                  # 1,000-word JOSS paper
    ├── paper.bib                                 # Bibliography
    └── figures/                                  # JOSS figures

The full change-log is in CHANGELOG.md.

The trilogy that motivated v3.0 (citable by name in any work using ITEA):

1. García-Lluis Valencia, A. (2026a). The Structural Flaw: Industrial Labour Contract Inadequacy under Agentic AI. Universidad Rey Juan Carlos. Zenodo: 10.5281/zenodo.19592542.
2. García-Lluis Valencia, A. (2026b). Beyond the Pigouvian Trap: Tokenised Intellectual Capital as a Pareto-Dominant Mechanism for the Agentic AI Layoff Externality. Universidad Rey Juan Carlos.
3. García-Lluis Valencia, A. (2026c). QE and Senior Workforce Restructuring: A Labour Liability Transmission Channel. Universidad Rey Juan Carlos. Zenodo: 10.5281/zenodo.19592184.

• O*NET 30.2 (February 2026): 1,016 occupations, 47,810 task statements. U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license; O*NET® is a trademark of USDOL/ETA. See full attribution below.
• BLS OEWS (May 2024): Occupational wage data by percentile, used for the criterion validity test r(OAEI, Wage).
• AIOE (Felten, Raj & Seamans, 2021): External convergent benchmark over 769 common SOC 6-digit occupations.
• Eurostat ESJS2 (referenced for future European replication, §11 of methodology).
• U.S. Census Bureau IDB: Demographic context only (not used in indicator construction).

@software{garcia-lluis2026itea,
  title     = {ITEA Framework: A Multidimensional System for Measuring Occupational
               Exposure to Algorithmic Expropriation under the Agentic AI Regime},
  author    = {García-Lluis Valencia, Alberto},
  year      = {2026},
  version   = {3.0},
  doi       = {10.5281/zenodo.20083102},
  url       = {https://github.com/Aval22/ITEA},
  note      = {Consolidated Methodological Edition. Supersedes v2.0 and v2.1.}
}

GitHub-native citation metadata is available in CITATION.cff.

Released under the MIT License.

• Academic use: Free with citation.
• Commercial use: Contact the author.
• Data: O*NET data is used under the CC BY 4.0 license (attribution below, not public domain); BLS OEWS public domain; AIOE used under Felten et al. 2021 terms.

This product includes information from the O*NET 30.2 Database by the U.S. Department of Labor, Employment and Training Administration (USDOL/ETA). Used under the CC BY 4.0 license. O*NET® is a trademark of USDOL/ETA. Alberto García-Lluis Valencia has modified all or some of this information (the ITEA indicators are derived, not original O*NET fields). USDOL/ETA has not approved, endorsed, or tested these modifications.

Alberto García-Lluis Valencia Universidad Rey Juan Carlos, Madrid Doctoral Programme in Economics and Business ORCID: 0009-0003-1438-1633 Email: alb.valencia@gmail.com Institutional: a.garciaval.2025@alumnos.urjc.es

ITEA Framework v3.0 — Making algorithmic expropriability measurable, multidimensional, and actionable.