Can AI help us save endangered languages? - What in the World podcast, BBC World Service

Key Concepts

• Linguicide: The systematic destruction or erasure of a language.
• Language Hierarchy: The social perception that certain languages are more "prestigious" or economically valuable than others.
• Endangered Language: A language at risk of falling out of use, often determined by whether it is being passed down to the next generation.
• AI Hallucination: When AI models generate inaccurate or nonsensical information, often due to a lack of training data.
• Data Sovereignty: The right of communities to control, own, and consent to the use of their linguistic data.
• Nüshu: A unique, historical script created by women for women in ancient China.

---

1. The Crisis of Language Loss

Linguists estimate that approximately half of the world’s languages could disappear by the end of this century. This phenomenon is described as "linguicide," a term coined in the 1980s to characterize the erasure of languages due to external pressures.

• Drivers of Extinction: Research indicates that the construction of infrastructure (roads) and rising socioeconomic status are primary variables in language loss. As communities integrate into the global economy, a "language hierarchy" often emerges, where dominant languages are prioritized for education and employment, relegating indigenous or local languages to a lower status.
• Defining Endangered Status: Contrary to popular belief, the number of speakers is not the sole indicator of endangerment. For example, the Karuk language in California was historically sustainable with only 1,500 speakers. Linguists instead look for functional indicators: Are parents teaching the language to their children? Is the language present in public life or education?

2. The Impact of Language Loss

The loss of a language is not merely the loss of vocabulary; it is the loss of stories, history, and unique worldviews.

• Community Well-being: Research from British Columbia suggests a correlation between the maintenance of indigenous languages and lower youth suicide rates. While correlation does not equal causation, experts suggest that language maintenance fosters "self-continuity" and cultural resilience.
• Intergenerational Disconnect: Language loss creates a "linguistic gap" within families, where grandparents and grandchildren may struggle to communicate, leading to the erosion of oral traditions and ancestral knowledge.

3. AI as a Tool for Preservation

AI is being explored as a potential solution to bridge the gap for languages ignored by major tech platforms.

• Case Study (Ghana): Language activists are developing smartphone translation tools that incorporate speech and audio recognition, which is critical for communities with lower literacy rates.
• Case Study (Nüshu): AI models are being trained on small datasets to translate Nüshu, preserving a unique cultural artifact that was historically restricted to women.

4. Risks and Ethical Challenges

The use of AI in language preservation is fraught with technical and ethical dangers:

• Inaccuracy and Hallucination: AI models often fail to recognize non-standardized languages. For instance, when tested with the Emilian dialect, an AI model failed to identify it, suggesting it was a "constructed language" (conlang).
• Misinformation: Major translation tools have produced "mind-bogglingly bad" errors, such as translating the Manx word for "Manx" (Gaelg) as "English."
• Data Ethics: There are significant concerns regarding who owns the data, whether consent was obtained from the speaker communities, and whether the tools actually address the community's specific needs.
• Inequality: Sophia Smith Galer warns that AI risks inheriting the inequalities of the "analogue world," potentially further disenfranchising communities that are already marginalized.

5. Synthesis and Conclusion

The preservation of endangered languages cannot rely solely on grassroots activism or volunteerism. While community interest is the foundation, institutional support—such as constitutional recognition and funding—is essential.

As noted by the BBC World Service contributors, language diversity is as vital to human society as biodiversity is to the planet. The consensus is that while AI offers a promising technological avenue for preservation, it must be developed ethically, with direct input from speaker communities, and supported by robust institutional frameworks to ensure that technology serves to empower rather than erase cultural heritage.