A more human-sounding voice went along with greater stated willingness to use the assistant
Biometric authentication also went along with greater stated willingness. The combination had the highest average, with no detectable interaction in usage intention.
The key idea
Each feature went along with greater stated willingness to use the assistant. Together, they had the highest average, with no detectable interaction on that measure.
The study, in plain language
What makes a voice assistant more appealing: how it sounds or how it recognizes the speaker? The team separated these two choices and presented a recorded conversation in four possible combinations.
People reported greater willingness to use the assistant with a more human voice, and the same was true of biometric authentication. The scenario combining both had the highest average. However, the analysis found no evidence that the effect of one depended on having the other.
A more human voice was also associated with impressions of greater ease and usefulness and lower security risk, but privacy concern did not decrease. A pleasant conversation left different impressions about use, security, and privacy.
What they found
- With the more human voice, participants reported greater willingness to use the assistant in the recorded scenario.
- The same was true with biometric identity confirmation, and the combination had the highest average.
- The more human voice seemed easier and more useful, but privacy concern did not clearly decrease.
Something to think about
When trying a voice assistant, ask separately whether the voice feels comfortable and whether it is clear how the assistant confirms who is speaking. A pleasant impression alone does not show how it handles your data.
How they studied it
An online experiment with four scenarios combining a more or less human voice with biometric authentication present or absent. Of 240 responses that passed attention checks, 60 were in each condition. The study measured stated willingness to use the assistant, not actual use.
What the study cannot show
- Participants listened to a simulated interaction rather than trying an assistant; measures were self-reported perceptions and intentions. The Mechanical Turk sample was mostly North American and college educated.
- That human-like assistants are safer, that physical traits used to confirm identity better protect data, that these preferences predict sustained use, or that the same effect applies to other cultures or products.
The original study
The effects of anthropomorphism and multimodal biometric authentication on the user experience of voice intelligence
Source under CC BY 4.0. English adaptation and explanation by PaperFlow, with AI assistance. The text has been reorganized for a general audience; it is not the complete paper and does not imply endorsement by its authors.
One study alone does not represent a field's consensus. Educational content does not replace professional advice. Report a possible error.
