Measuring Task-Level Technological Exposure: A Language Model Approach

Andre Mouton
Additional contact information
Andre Mouton: Department of Economics, Wake Forest University

No 132, Working Papers from Wake Forest University, Economics Department

Abstract: This paper develops methods and tools for measuring the exposure of occupational tasks to technological substitution. Patent abstracts and task statements are matched and classified by a small, open-source language model. The model is fine-tuned and validated against a foundation AI, achieving accuracy improvements of roughly 5X over conventional ‘word embedding’ approaches. Model fine-tuning and a rules-based match threshold are critical for realizing these gains. The approach replicates stylized facts about IT exposure, but diverges sharply from survey-based measures of AI automation risk, which systematically understate exposure among high-wage occupations. A living dataset and Python package allow researchers to measure exposure across user-defined technology and task categories, with minimal time lag and at fine temporal resolution.

Keywords: Technological Exposure; Task Automation; Language Models; Patent Analysis (search for similar items in EconPapers)
JEL-codes: C45 C82 J20 O33 (search for similar items in EconPapers)
Pages: 33
Date: 2026-02-02
References: Add references at CitEc
Citations:

Downloads: (external link)
https://drive.google.com/file/d/1sDmdqQZ4BVJgitUWT ... MYh/view?usp=sharing

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:ris:wfuewp:022172

Access Statistics for this paper

More papers in Working Papers from Wake Forest University, Economics Department Contact information at EDIRC.
Bibliographic data for series maintained by Don Shegog ().