Generative AI reshapes job requirements in global labor markets
A recent study investigates more than 150,000 job advertisements in English from 2018 to 2025 to evaluate the impact of generative AI on labor market requirements. This research integrates data from twelve open-access sources and a public API, utilizing techniques such as lexical skill extraction, semantic framing, topic modeling (including BERTopic and LDA), clustering (KMeans), and time-series forecasting (ARIMA). Skills are classified into five categories: AI_Data, Routine, Soft_Meta, Domain_Specific, and Leadership. The study examines whether generative AI serves as an augmentative or substitutive influence across various industries, analyzing the relationships between skills through correlation matrices and cross-sector analyses. Semantic changes in job postings over time are assessed using sentence-transformer embeddings and cosine similarity.
Key facts
- Over 150,000 English-language job postings analyzed
- Data spans 2018 to 2025
- Sources include twelve open-access datasets and one public API
- Methods: lexical skill extraction, semantic framing, BERTopic, LDA, KMeans, ARIMA
- Five skill dimensions: AI_Data, Routine, Soft_Meta, Domain_Specific, Leadership
- Explores augmentative vs. substitutive role of generative AI
- Uses sentence-transformer embeddings and cosine similarity
- Published on arXiv with ID 2605.00843
Entities
Institutions
- arXiv