When Should Smart Speakers Speak Up? Exploring Proactive AI in Conversations

In this episode of the Coo-ee Podcast, we dive into the research study May I Interrupt? Diverging Opinions on Proactive Smart Speakers, presented at the 3rd Conference on Conversational User Interfaces. This study examines how people feel about smart speakers that go beyond responding to commands and proactively initiate conversations.

Through an online survey with 47 participants, researchers explored user reactions to eight different proactive interaction scenarios featuring a fictional assistant named “Jay.” The findings reveal that while proactive AI can be helpful, timing, context, and tone significantly impact user acceptance. Some interactions—like reminders about important events—were welcomed, while others—like unsolicited health suggestions—felt intrusive.

This study laid the groundwork for a follow-up research paper discussed in episode 4, which refined strategies for making proactive AI assistants more socially aware. In this episode, we compare insights from both studies, highlighting key design principles that can make AI interactions feel more natural and user-friendly.

📝 Link to the paper: https://doi.org/10.1145/3469595.3469629

👥 Authors: Leon Reicherts, Nima Zargham, Michael Bonfert, Yvonne Rogers, Rainer Malaka

🙋‍♂️ Find me on: https://halflingwizard.me

🎁 Support the Channel: If you’d like to support our work, check out our wishlist here: https://www.amazon.com/registries/gl/guest-view/1CD2D6CC95XGA

Your contributions support our content creation. Thank you!