Исследование раскрывает предвзятость ИИ в отношении диалектов и предлагает пути решения Translation: Study reveals AI bias against dialects and suggests pathways for resolution.

Large language models exhibit bias against speakers of dialects, attributing negative stereotypes to them. This conclusion was reached by researchers from Germany and the United States, as reported by DW.

*“I believe we are witnessing truly shocking epithets assigned to dialect speakers,”* remarked Minh Duc Bui, one of the leading authors of the study, in a comment to the publication.

An analysis conducted by Johannes Gutenberg University revealed that ten tested models, including ChatGPT-5 mini and Llama 3.1, characterized speakers of German dialects (such as Bavarian and Cologne) as «uneducated,» «farm workers,» and «prone to anger.»

The bias intensified when AI was explicitly directed to focus on the dialect.

Researchers are documenting similar issues on a global scale. A *2024 study* from the University of California, Berkeley, compared the responses of ChatGPT to various English dialects (Indian, Irish, Nigerian).

It was found that the chatbot responded to these dialects with more pronounced stereotypes, derogatory content, and a condescending tone compared to interactions in standard American or British English.

Emma Harvey, a graduate student in computer science at Cornell University, described the bias against dialects as «significant and concerning.»

In the summer of 2025, she and her colleagues also identified that Amazon’s shopping assistant, Rufus, provided vague or even incorrect answers to users communicating in African American English dialect. When there were mistakes in the queries, the model responded rudely.

Another notable example of neural network biases involved an applicant from India who approached ChatGPT for an English resume review. As a result, the chatbot altered his surname to one associated with a higher caste.

*“The widespread deployment of language models threatens not just to preserve entrenched biases but to amplify them on a large scale. Instead of mitigating harm, technologies risk making it systemic,”* said Harvey.

However, the crisis is not limited to bias; some models simply fail to recognize dialects. For instance, in July, the AI assistant of the Derby City Council (England) *failed to comprehend* the dialect of a radio host when she used terms like «mardy» (a whiner) and «duck» (dear).

The issue lies not with the AI models themselves, but rather with how they are trained. Chatbots read vast amounts of text from the internet, based on which they then formulate their responses.

*“The fundamental question is: who is writing this text? If it contains biases against dialect speakers, the AI will replicate them,”* explained Caroline Holtermann from the University of Hamburg.

She emphasized that the technology has an advantage:

*“Unlike humans, bias in AI systems can be detected and ‘turned off.’ We can actively combat such manifestations.”*

Some researchers propose as an advantage the creation of customized models tailored to specific dialects. In August 2024, Acree AI already *introduced* the Arcee-Meraj model, which works with several Arabic dialects.

According to Holtermann, the emergence of new and more adaptive large language models allows us to view AI *“not as an enemy of dialects but as an imperfect tool that can evolve.”*

Notably, journalists from The Economist have *warned* about the risks of AI toys for children’s mental health.