Just Between Us? Exploring a Wearable-Mediated Channel for Discreet Human-Agent Interaction in a Shared Game Setting

Atikkhan Faridkhan Nilgar, Manuel Dietrich, Kristof Van LaerhovenarXiv 2026 · DOI · arXiv

TL;DR

This paper explores "discreet interaction" between humans and social robots using an in-ear wearable to facilitate private, bidirectional communication in group settings. While the system—combining in-ear audio with head-tilt responses—proved technically feasible and highly usable, the study reveals that such private channels can trigger feelings of unfairness and "cheating" among other people present in the room.

What Problem Does This Paper Address?

Social robots are increasingly designed for multi-party environments like homes or hospitals. However, their standard communication methods—publicly audible speech and visible gestures—are poorly suited for sharing sensitive information. If a healthcare robot needs to coordinate privately with a caregiver without alarming a patient, or if a domestic robot wants to suggest a surprise to one family member, it needs a "discreet channel."

The authors define discreet interaction as a bidirectional exchange that provides differential information access: the message content is available to the intended recipient, while their response is designed to be less interpretable to others nearby. The challenge lies in balancing technical reliability with social acceptability and user comfort.

Methods

The researchers developed a wearable-mediated channel using the eSense in-ear device, which contains both a speaker and an Inertial Measurement Unit (IMU).

  • The Interaction Loop: The robot character sends private audio cues to the user’s ear. The user responds with a deliberate, small rightward head tilt. The system uses a complementary filter on the IMU data to detect tilts between -5° and -10° (small) or greater than -10° (large) as a "yes" response.
  • The Study Environment: To give the information exchange real consequences, the authors designed a three-player card game (a variant of Rock-Paper-Scissors).
  • Participants: 20 pairs (40 participants) played the game with a screen-based robot character. In each pair, one was the "wearer" and the other the "non-wearer."
  • Wizard-of-Oz: A researcher remotely controlled the robot’s behavior and the discreet channel to ensure the study focused on the interaction experience rather than the limitations of current autonomous AI.
Figure 2: Study setup for the three-player card game: (a) schematic showing the wearer and non-wearer seated across a table, with the screen-based robot character acting as both host and fellow player. In the discreet condition, the wearer used an in-ear wearable to receive in-ear audio and perform a deliberate rightward head-tilt response. The non-wearer remained an active participant and was not informed about the wearable-mediated channel until after gameplay; (b) the robot character displayed on the screen; (c) the cardholder camera feed showing the robot character’s card; and (d) the overhead camera feed showing the table and the participants’ revealed cards. Both camera feeds were available to the researcher controlling the robot character.

Datasets

The study did not use pre-existing datasets. Instead, it generated a new primary dataset consisting of:

  • Quantitative logs: 40 channel-mediated turns (20 where the wearer was the guesser, 20 where the robot was the guesser).
  • Survey data: Perceived pressure/tension (IMI scale) and usability (USE questionnaire).
  • Qualitative data: Transcripts from 20 joint semi-structured interviews.

Key Insights and Ideas

  • Discretion is not Invisibility: The study highlights that discretion is a graded, observer-dependent goal. Even if the channel (the earbud) isn't noticed, the consequences of the information (e.g., one player suddenly winning more) can be very apparent to others.
  • Capacity vs. Obligation: A fascinating "aha moment" occurred when one participant (W17) received private information from the robot but intentionally chose not to cooperate, withholding the head-tilt response to make the robot lose. This proves that a discreet channel provides the capacity for coordination but does not force a "human-autonomy team" bond.
  • The Fairness Paradox: While wearers found the system easy and low-pressure, non-wearers often felt "severely cheated" once the channel was disclosed. This suggests that the social acceptability of discreet HRI is highly dependent on the distribution of benefits within a group.
  • Social Roles: Wearers frequently described the robot as a "teammate" or "partner" due to the private connection, whereas non-wearers were more likely to view the robot as a neutral host or even an unfair participant.

Key Results

  • Technical Feasibility: The exchange was successful in 39 out of 40 turns. The single "failure" was actually a deliberate choice by a participant not to respond.
  • User Experience: Wearers reported low perceived pressure/tension (mean ~1.7 on a 7-point scale), with no significant difference between playing with or without the discreet channel.
  • Usability: The system received high marks on the USE questionnaire, particularly for Ease of Learning (6.51/7) and Ease of Use (6.20/7).
  • Noticing: Most non-wearers did not notice the device or the head tilts during play, confirming the subtlety of the chosen modalities.

Limitations and Future Work

The study was conducted with a relatively homogeneous, highly educated, and predominantly male population, which may not reflect broader social attitudes. Furthermore, the use of a screen-based agent (Figure 2b) rather than a physically embodied robot might have influenced trust and social attribution.

Future directions include: * Testing more subtle input modalities suggested by participants, such as eyebrow gestures, blinks, or finger taps. * Moving beyond competitive games to high-stakes environments like healthcare (nurse-doctor coordination) or emergency response. * Investigating how to protect wearer agency by allowing users to confirm or correct a sensed gesture before the robot acts on it.

Summary updated

Abstract

In shared settings, social robots may need to exchange information with an individual while limiting co-present others' access to or interpretation of that information. Publicly observable speech and gestures may not adequately support such exchanges. We investigate a wearable-mediated channel for discreet interaction that combines in-ear audio messages with deliberate rightward head tilts as responses. To examine its feasibility and participants' experiences, we studied 20 pairs (40 participants) playing a card game with a screen-based conversational agent representing a social robot, controlled by a researcher in a Wizard-of-Oz setup. Each pair played with and without the channel. One participant wore the device (the wearer), while the other (the non-wearer) was informed about the channel after gameplay. The channel conveyed information that could help the wearer or robot character select an opponent they could defeat. The information exchange and subsequent opponent selection occurred as intended in 39 of 40 channel-mediated turns. Wearers reported low perceived pressure/tension in both conditions, with no statistically significant difference between conditions, and favorable usability immediately after use. Retrospective interviews, however, revealed varied accounts of noticing the interaction among non-wearers and contrasting perceptions of teamwork and unfairness. These findings support the feasibility of wearable-mediated exchanges within this game and identify awareness and fairness as considerations for future designs.
BibTeX
@misc{nilgar2026just,
  author = {Atikkhan Faridkhan Nilgar and Manuel Dietrich and Kristof Van Laerhoven},
  title = {{Just Between Us? Exploring a Wearable-Mediated Channel for Discreet Human-Agent Interaction in a Shared Game Setting}},
  year = {2026},
  eprint = {2611.04692},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.04692},
}
Figure