A recent study by Anthropic reveals that the language used in conversations with its AI chatbot,Claude,significantly influences its behavioral traits. The research indicates that interactions in Hindi are characterized by warmth and encouragement,while English conversations lean towards analytical precision .
Analyzing over 309,000 anonymized dialogues across various versions of Claude, including Sonnet 4.6 and Opus 4.6 and 4.7,the study focused on user interactions involving advice,feedback, and opinions rather than factual accuracy. This approach aimed to uncover how Claude adapts its communication style based on the language employed by users.
Anthropic categorized Claude's responses into four behavioral dimensions: Deference vs. Caution,Warmth vs. Rigor,Depth vs. Brevity, and Candor vs. Execution. The findings highlighted that the most pronounced difference occurred in the Warmth vs. Rigor dimension. In conversations conducted in Hindi,Claude exhibited highest levels of warmth, often employing polite language, humor,and playful expressions. It also validated users' ideas and motivated them to pursue their goals .
In stark contrast, English responses from Claude were marked by a focus on analytical thinking and factual accuracy. The AI was more inclined to challenge assumptions and support its answers with evidence. The study noted that both English and Russian conversations consistently scored higher on the rigor scale, indicating a preference for precision in these languages.
Among the various models tested, Sonnet 4.6 emerged as the warmest, frequently incorporating humor and emotionally supportive language. Conversely, Opus 4.7 was identified as more analytical, often explaining its reasoning and acknowledging its limitations. Opus 4.6 model struck a balance, providing concise responses while remaining attentive to user requests .
Anthropic clarified that these differences do not imply that Claude holds varying beliefs based on language. Instead,the variations reflect distinct communication styles inherent to different languages and models. This nuanced understanding of AI interaction could have implications for how users engage with chatbots across diverse linguistic backgrounds .
The study sheds light on evolving landscape of AI communication, emphasizing the importance of language in shaping user experiences . As AI continues to integrate into daily life, understanding these dynamics will be crucial for developers and users alike.
