Human interaction with voice-AI

Voice artificial intelligence (voice-AI), like Amazon’s Alexa and Apple’s Siri, is becoming more prevalent in our daily lives. Recent research highlights that social factors, such as gender and likability, influence how we communicate with virtual assistants, much like human-to-human interactions.

To explore this phenomenon, we conducted a study where people engaged in a question-and-answer session with virtual assistant voices representing British, American, and Indian English dialects. We tested how the perceived identity of the voice — whether it sounded like a device or a human — and the voice’s dialect affected participants’ responses. We found that people had the most positive experiences with voices in the Indian English dialect, and shifted their speech to be more like those voices. On the other hand, people had the least positive experiences with voices in the American English dialect, and shifted their speech to be less like those voices.

Our results underscore the need for further research into human-voice-AI interactions and how social attitudes shape these exchanges. Understanding these dynamics can guide the design of more effective voice-AI products, ultimately enhancing user experiences with virtual assistants. This is crucial for the development of intuitive and relatable voice-AI technologies to serve varying global cultures.