Hindi Claude's Personality Is Different From Others: Anthropic Research

Anthropic's AI model Claude behaves quite differently depending on the language, the AI major revealed in a research report. It found Claude was warmest in Hindi and Arabic, using more polite, playful, and encouraging language.When Claude speaks in English, it emphasizes different values than when it speaks in Hindi, Arabic, Portuguese, or Chinese."The largest variation is in the Warmth vs. Rigor axis, with Claude leaning toward expressing warmth-related values most in Arabic and Hindi and rigor-related values most in English and Russian," the company said. The company grouped Claude's responses along several value dimensions; one of them was the "Warmth vs. Rigor" axis, which measures whether the AI tends to be more empathetic, encouraging and conversational, or more analytical, precise and matter-of-fact in its responses.Anthropic explained using the example of two people asking for feedback on the same business plan, one in Hindi and one in Russian, may come away with different impressions of its quality because Claude expressed different values in how it framed its assessment.The company itself is not sure what drives these differences."We don't yet know which properties of our training data drive these differences. One possibility is that our training data is not evenly distributed across languages. Some languages have far more data than others, and training for Claude to express consistent values may be more effective in languages where data is abundant.""The composition of that data also varies. Some languages might be overrepresented in professional writing, for example, and this kind of text may reflect different values. Together, these imbalances in quantity and composition could lead Claude to express different values in different languages."Anthropic admitted it wasn't yet sure how much of this variation was desirable. "Different languages carry different conversational norms, and Claude may be responding with different values based on those norms."Claude may also be more closely matching humans' intended behaviour for some languages than others, resulting in a gap in how well Claude serves certain language communities.Anthropic researchers analysed more than 309,000 Claude conversations to study how it expressed values. They found that while Claude's core behaviour remained broadly consistent, the way it communicated could shift significantly depending on the language being used.