During a recent study, we asked participants how long they thought they would have to speak in order for their voice to be uniquely recognized (i.e., voice print). While their estimates varied widely – from 30 seconds to 30 minutes – most people said about three to five minutes.

The reality is, voice prints can take as little as five seconds to do accurately if the correct phrases are spoken. So the question is, as researchers, should we recommend to design for what people think and feel is the correct length so they feel secure? Or should we collect the minimum?

While I understand why we might do what is comforting for users, I think our job is to convince people that voice biometrics are secure with the minimum amount of effort required. What do you think? Join the discussion!

Check out some of our blogs about voice!

1 Comment

  1. Korey

    Partial agreement.

    What I agree with is that it is incumbent upon the UX community to find ways to instill trust in reliable systems through design – such as a voice print feature that only requires 5 seconds.

    However, I think recognition that in cases like this where user expectations are based on consistent past experience across multiple systems (i.e., most voice recognition systems I have used in the past have been pretty weak… I don’t care what this one says, I don’t believe it is that good) – playing to the users’ expectations a bit can help build that trust more effectively. For example, after 5 seconds of voice printing, indicate that the system has sufficient data, but provide the user with the option to continue voice printing to “further enhance” the reliability of the print. Over time, as trust in a particular system (or class of systems) grows, the crutch can be phased out. A little deception in the interest of building trust is ironic… but I think it can be effective!

    Reply

Submit a Comment

Your email address will not be published. Required fields are marked *

Pin It on Pinterest

Share This