Navigating Language Contact in the Digital Age

An AI Experiment with Belarusian

Abstract

Belarusian, classified as vulnerable by UNESCO, is a low-resource language with limited digital and media representation. Preserving such languages demands a comprehensive approach, which is deeply interconnected with their linguistic and historical contexts. Thus, any study of Belarusian must consider its linguistic variability, historical fragmentation, and the dominant pressures of Russian and Polish. This best practice paper investigates the potential of artificial intelligence (AI) in addressing the complex dynamics of Belarusian. It also outlines Iceland’s recent partnership with OpenAI to preserve Icelandic and argues that a similar AI–human collaboration could benefit Belarusian. The contribution aims to explore how AI can be utilised to support language learning, generate new vocabulary, and preserve the characteristic features of Belarusian. However, the study also acknowledges the limitations of current AI models. The activities presented can be implemented at both school and university levels, proving equally valuable for native speakers, language learners, and future teachers. This best practice paper relies on experiments and qualitative content analysis of the results and addresses two core questions: 1. Can AI tools cope with orthographic and grammatical complexities of Belarusian? 2. Can AI models successfully navigate linguistic interference in Belarusian? The protocols of the experiments can be transferred to other languages or rerun unchanged to track how new outputs and model biases evolve over time.