Social robots that interact with groups of people face an interesting design problem: how do you let a robot share something with one person in the room without everyone else noticing? A study by Atikkhan Faridkhan Nilgar, Manuel Dietrich, and Kristof Van Laerhoven, accepted to the 25th International Conference on Mobile and Ubiquitous Multimedia (MUM) 2026, tested one answer — a wearable channel combining in-ear audio messages with deliberate rightward head tilts as the response.
The setup
Twenty pairs (40 participants total) played a card game against a screen-based conversational agent standing in for a social robot, controlled behind the scenes in a Wizard-of-Oz setup. One participant in each pair — the wearer — had the device; the other, the non-wearer, was only told about the channel after the session ended. Each pair played both with and without the discreet channel active.
The channel’s job was narrow but specific: feed the wearer (or the robot character) information that could help pick which opponent to target next — the kind of tactical advantage that would look suspicious if said out loud at the table.
It worked — mostly
The information exchange and subsequent opponent selection happened as intended in 39 of 40 channel-mediated turns, a strong feasibility result. Wearers reported low perceived pressure or tension whether or not the channel was active, with no statistically significant difference between conditions, and rated the device’s usability favorably right after use.
The more interesting findings came out in retrospective interviews. Non-wearers gave varied accounts of whether they’d noticed the discreet interaction happening at all, and described contrasting perceptions of teamwork and fairness — some read the arrangement as a clever bit of game mechanics, others as something closer to cheating.
Why it matters beyond the card table
The study’s authors frame awareness and fairness as open design considerations for any future system built around discreet human-agent channels — a wearable that quietly shares information with one person in a shared space raises the same questions whether it’s a tabletop game, a classroom, or a workplace meeting with an AI assistant whispering in someone’s ear. The paper is on arXiv (2611.04692) under Human-Computer Interaction and Robotics.