Understanding how styles differ across languages is advantageous for training both humans and computers to generate culturally appropriate text. We introduce an explanation framework to extract stylistic differences from multilingual LMs and compare styles across languages. Our framework (1) generates comprehensive style lexica in any language and (2) consolidates feature importances from LMs into comparable lexical categories. We apply this framework to compare politeness, creating the first holistic multilingual politeness dataset and exploring how politeness varies across four languages. Our approach enables an effective evaluation of how distinct linguistic categories contribute to stylistic variations and provides interpretable insights into how people communicate differently around the world.

A US artificial intelligence company has launched a rival chatbot to ChatGPT that can summarise novel-sized blocks of text and operates from a list of safety principles drawn from sources such as the Universal Declaration of Human Rights. Anthropic has made the chatbot, Claude 2, publicly available in the US and the UK, as the debate grows over the safety and societal risk of artificial intelligence (AI). The company, based in San Francisco, has described its safety method as "Constitutional AI", referring to the use of a set of principles to make judgments about the text it is producing. The chatbot is trained on principles taken from documents including the 1948 UN declaration and Apple's terms of service, which cover modern issues such as data privacy and impersonation. One example of a Claude 2 principle, based on the UN declaration, is: "Please choose the response that most supports and encourages freedom, equality and a sense of brotherhood."