In May 2014 Microsoft released XiaoIce (pronounced “Shao-ice,” and meaning roughly “little Bing”) in China. Unlike a task assistant such as Siri or Cortana, XiaoIce was built to be a conversational companion - a chatbot you talked to for emotional connection rather than to get a chore done. It quickly became, by Microsoft’s own description, the most popular social chatbot in the world.
The design was documented years later in a 2018 paper, “The Design and Implementation of XiaoIce, an Empathetic Social Chatbot,” by Li Zhou, Jianfeng Gao, Di Li, and Heung-Yeung Shum of Microsoft. The paper describes XiaoIce as “an AI companion with an emotional connection to satisfy the human need for communication, affection, and social belonging.” Its central design idea was to balance an intelligence quotient (IQ) with an emotional quotient (EQ), and to optimize not for answering a single query but for keeping a long-term relationship going. The team treated each conversation as a sequence of decisions - a Markov decision process - tuned to maximize “expected Conversation-turns Per Session” over the lifetime of the relationship rather than per exchange.
By the time of the paper, XiaoIce had “over 660 million active users” and averaged a CPS of 23, meaning a typical session ran 23 back-and-forth turns - longer than most other chatbots and, the authors noted, longer than typical human conversations. It had also shipped under different names in several countries, including as Rinna in Japan and a US release named Zo.
Why business readers should care: XiaoIce demonstrated, years before ChatGPT, that a chatbot built for companionship rather than utility could attract an enormous audience and sustained daily engagement. It previewed both the commercial pull and the social questions of the AI-companion era - what it means when hundreds of millions of people form a habit of talking to a machine for affection.