Gender Bias in Social Chatbots:A Conversation Test Study Based on Xiaoice Series of Chatbots
As AI social chatbots are seen as human communicators,it is crucial to understand the problems of gender bias in their interactions with humans.Using the method of conversation test,this paper designs a series of questions for testing gender bias of robots and to test the gender bias of three mainstream social chatbots in China.The interaction texts are analyzed through qualitative coding analysis.The results indicate that social chatbots exhibit significant gender bias in self-perception of gender,gender stereotypes,gender equality,and response to gender harassment,which are unrelated to the male and female gender roles of the social chatbots themselves.The gender bias of social chatbots as products of human-computer interaction technology,they are constructed by user participation,dialog system technical support,technology companies and program developers.The result is that Al,as represented by social chatbots,replicates and reinforces the construct power of gender bias in the gender culture of human society in learning and imitation.
artificial intelligencesocial chatbotsXiaoice seriesgender biasconversation test
NETL
NSTL
万方数据
