Introducing voice-to-voice moderation

For a while now, we've been building an AI moderator that feels as natural, engaging, and authentic as conversing with a real human. At the same time, we’re focused on making sure researchers retain full control of the participant experience without sacrificing the depth, quality, or reliability of the responses they gather.

So far, our AI moderator has proven effective at engaging participants in text-to-speech conversations, where the moderator asks questions in a text format and participants speak their responses. Today, we're introducing a new moderation mode: voice-to-voice moderation.

Voice-to-voice mode lets participants hear the research questions spoken aloud and reply in a more natural way, using their voice. In voice-to-voice mode, the AI moderator listens to their response and understands when to move the conversation forward with dynamic follow-up questions.

This has been made possible thanks to significant advancements in foundational large language models, many of which now support voice generation. We've reached a point where producing low-latency, realistic conversations between humans and AI is achievable without participants feeling stuck in the uncanny valley (that uneasy sense you get when interactions feel almost, but not quite, human). Our voice-to-voice moderation feels genuinely conversational and maintains the depth, nuance, and accuracy you've come to expect from our AI moderator.

When we first launched voice responses as an input method for our AI-moderated studies, we saw that voice-based responses resulted in far more detailed and nuanced feedback from participants. We think voice-to-voice mode represents an equally big improvement to the participant experience, as allowing participants to talk openly makes sharing their thoughts easier. They don't need to worry about typing up their answers and instead can simply say what's on their minds. As a result, the responses collected in voice-to-voice mode feel more authentic, and participants say that the voice-to-voice moderator feels less rigid and much more conversational and engaging.

Voice-to-voice mode works across every block and study type available on Wondering and can be applied to blocks individually. It is also supported in any of the distribution methods you use, whether your study is link-based, embedded directly into your product with our in-product targeting, or sent through our participant panel.

Our goal has always been to help you achieve the quality and richness of human-moderated research, combined with the efficiency, scale, and convenience that AI moderation enables. Voice-to-voice mode significantly narrows the gap between participating in a human-moderated study and an AI-moderated one.

How do I turn on voice-to-voice moderation in my study?

To change the moderation mode for your study, click the "Study settings" button in the study builder, and choose the mode you want for your study:

‍

At the start of studies in Voice-to-voice mode, a custom message will be shown asking participants to grant access to their microphone. You can customize this message by going to Study settings, choosing Voice-to-voice moderation and then customising the Permissions screen message:

Voice-to-voice moderation mode is available starting now for all Wondering users. You can use this new mode in your existing AI-led studies and experience firsthand the difference it makes to the participant experience.

Author

Axel Thomson

Founder

Author

Continue reading

Understand your customers with AI-driven research