Chatbots Are Unpopular in Asia: Cultural Adaptation Missing
A researcher from the University of Canterbury found that most voice assistants are trained on Western examples, which makes them ineffective in Asia. He proposed a four-step scheme to embed cultural features into chatbots from the start, rather than adjusting them later. For instance, in Jakarta a bot should ask about traffic jams instead of the weather.
Professor Vic Naidoo from the University of Canterbury, together with colleague Karman Kaur Chadha, developed a simple four-step scheme to integrate cultural characteristics into chatbots immediately. The first step is to embed local norms, language, and context into the data. The second is to teach the bot to adjust its tone, pace, and style of communication during the conversation. The third is to explain how the bot works in a language understandable to that audience. The fourth is to organize oversight, cultural risk checks, and involvement of local residents in evaluation. Naidoo cited an example: an Australian company launched chatbots in Indonesia, Thailand, and Vietnam, but people simply did not want to talk to them. Engineers translated responses into local languages, but that was not enough. In a Western environment, the bot starts a conversation with a remark about the weather, but in Jakarta it is much more natural to ask about traffic jams. Without such configuration, the bot loses the user. Naidoo believes that the main question is whether the system benefits the person interacting with it. If a person does not trust the bot and does not want to deal with it, all investments are wasted.
Source: Hightech.fm —
original