Strategies for Eliminating Linguistic and Cultural Prejudices in the Integration of General AI Systems
In today's interconnected world, the application of Generative AI (genAI) is expanding rapidly, particularly in busy work environments like manufacturing factory floors. As many as 95% of U.S. companies have adopted genAI, and platforms like ChatGPT and AI-powered Google searches are dominating in 2025.
However, the English-centric nature of many genAI platforms poses challenges, with English making up 67.3% of websites. To address this, it is crucial to consider cultural traditions and linguistic nuances in genAI development. Datasets should be created to capture certain language characteristics, such as onomatopoeia and context-heavy forms of communication, ensuring that the AI reflects the linguistic and cultural context of various regions.
For instance, Japanese, a high-context language, relies heavily on onomatopoeia and subtle shifts in expression to convey intent and inference. Similarly, Edward T. Hall's theory outlines intrinsic differences in how various cultures communicate. Therefore, genAI models should be regionally curated rather than using a universal model that may not be as culturally or linguistically agile.
The United States and China, with their strong technological ecosystems and significant investment, are at the forefront of developing and deploying multimodal generative AI systems. These systems use text, speech, gestures, and images to overcome language and cultural barriers. The US emphasizes practical applications and diverse private investments, while China integrates humanoid robotics widely in services and builds scalable supply chains for key components. Europe, particularly the DACH region, benefits from a mature GreenTech ecosystem and a strong innovation culture supportive of AI startups, fostering sustainable and socially impactful AI development.
To ensure the accuracy and effectiveness of genAI, workshops should be conducted where local teams can test and practice using these tools, providing valuable feedback on their performance. Teams should also be encouraged to always loop in a human to clarify any confusion, mitigating the risk of misinformation or misguidance from genAI tools.
Moreover, AI training workshops should avoid 'echo chambers' by ensuring that new information from AI is not just from an individual, but from a wide range of sources. AI tools need to be continuously updated with real-world data from multilingual sources to learn the nuances of different languages and communication forms.
Serious concerns have been raised about bias in AI-detectors against non-native English writers, and non-native English speakers are often left out of the equation when discussing AI deployment strategies. To address this, employees should be trained on best practices around prompting genAI platforms, with a focus on clarity and principles like fairness and transparency.
Lastly, digital or technology-based communication, specifically that experienced via genAI tools, must include other modes beyond text-based messaging. This is particularly important in environments involving multiple languages or cultures, where communication can be distorted. By considering these factors, we can ensure that genAI serves as a bridge, connecting people across linguistic and cultural boundaries, rather than a barrier.
