Effect of Speech Entrainment in Human-Computer Conversation: A Review

The phenomenon of entrainment in conversation is the process where participants become more similar to each other in terms of different verbal and non-verbal aspects such as acoustic-prosodic, lexical, syntactic, pitch, and speech rate. This process of becoming similar to each other is the key to effective human-human conversation. To replicate the effectiveness observed in human-human conversation, it is equally critical to explore the occurrence of entrainment within human-machine conversation. This review article examines the various non-verbal and verbal aspects that machines are able to adapt for improved entrainment in human-machine conversation. Initially, we categorize the specific verbal and non-verbal behaviors of human users that machines are capable of adapting. Subsequently, we analyze the likely challenges that have prevented the speech technology sector from enabling smooth, natural interactions between humans and machines. These obstacles have hindered the industry’s ability to leverage the phenomenon of entrainment for more fluid and intuitive human-machine conversation. Finally, we advocate for a mechanomorphic design strategy in human-machine conversation, outlining the rationale for its potential efficacy.