ISCA Archive Interspeech 2024
ISCA Archive Interspeech 2024

Behavioral evidence for higher speech rate convergence following natural than artificial time altered speech

Jérémy Giroud, Jessica Lei, Kirsty Phillips, Matthew H. Davis

As AI progresses, our exposure to artificially generated spoken content varying in naturalness and speed increases. This trend is amplified by AI-powered personal assistants’ proliferation, multiplying our interactions with intelligent systems. Research is crucial to understand if phenomena observed in human-human interactions can inform these new interactions. For instance, in everyday conversation, people adjust their speaking rate to match their partner's, a phenomenon known as speech rate convergence. It is crucial for effective communication, occurs automatically and is present in more artificial interaction scenarios. We investigated how the nature (natural vs. artificial) and the presentation rate (normal vs. fast) of the speech signal impact speech rate convergence. Data from 116 participants across two experiments reveal higher convergence towards naturally produced speech compared to artificially time altered speech. Implications for human-machine interactions are discussed.