Reproduce “How Early Adopters Used Generative AI Worldwide: Variation by Country Income and Language” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Madeleine I. G. Daepp, Isaac Slaughter

Researchers from Microsoft Research and the University of Washington investigated global usage patterns of generative AI among early adopters, focusing on how these patterns vary by country income and language. They found that users in lower-income countries tended to apply AI more for schooling and market work, while those in higher-income countries used it more for leisure; additionally, English was disproportionately relied upon, especially in regions with poorly supported local languages, linking model performance to real-world language use.

Key insights

  • Generative AI is being utilized for human capital accumulation (schooling, market work) at higher rates in lower-income countries, challenging traditional "second-level" digital divide narratives observed with internet adoption.
  • Linguistic capabilities of LLMs directly influence user behavior, with English serving as a de facto lingua franca in regions where local languages are poorly supported, leading to potential language-based digital divides.
  • The study established a robust, AI-based classification methodology for real-world AI use cases, extending beyond typical knowledge-work taxonomies.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.