A Study Of The Anthropomorphic And Emotional Expression Of Intelligent Voice Virtual Assistant

Virtual assistants(VA)are AI products commonly used in the lives of users and are an important commercial application of conversational assistants.There is still relatively little research on voice-based virtual assistants on the voice channel,and in particular,there is a strong need for experimental research on the degree of anthropomorphism and emotional expression of VA speech.This study discusses the effects of VA voice anthropomorphism,gender factors,intonation,and contextual emotional expression on perception,trust,willingness to use,and recommendation of human-computer interaction through a pre-experiment and two experiments.Experiment 1 was a three-factor within-subject design that explored HCI outcomes at different levels of anthropomorphism by adjusting the level of VA voice anthropomorphism(low vs.high).Experiment 2,also a three-factor within-subject design,introduced the variable VA intonation affect(neutral vs.positive vs.negative)to further investigate the effect of consistency in intonation x context affect on HCI outcomes.Both experiments were pre-experimented to verify the validity of the experimental material,and both considered the role of VA gender as a factor.The specific findings of the study are as follows:(1)The degree of VA speech anthropomorphism significantly influenced the user’s perception of the VA.Highly anthropomorphic VAs were more social,lifelike,and less intrusive than less anthropomorphic VAs,and also gained more human-computer trust and a higher willingness to use and recommend.(2)In highly anthropomorphic VA,the emotional expression of VA intonation also has an important impact on the human-computer interaction outcome,with VA expressing positive emotions and gaining better user perception.(3)The emotional expression of VA tone needs to be adapted to the text context,and when the emotional expression of both is consistent,the higher the sociality and vitality,the lower the interference,and the higher the human-machine trust and willingness to use/recommend.(4)There are some differences between male and female VA performance in different dimensions,but the differences in VA gender are unstable,especially in the gender effects on sociality and trust.In summary,the main conclusion of this study is that the degree of anthropomorphism,the tendency to express emotions,and consistency all significantly affect the outcome of human-computer interaction in the VA design of the voice channel,including various aspects of perception,trust,and willingness to use/recommend.