Behavioral Diversity in HRI 

The Challenge

Telepresence robots are increasingly deployed in shared institutional spaces: offices, hospitals, campuses. But most deployments are designed around the operator's needs, not the needs of the people who unexpectedly encounter the robot in a hallway. Brands and institutions lack clear data on what actually drives a bystander to trust, avoid, or engage with a robot they never asked to meet.

Our Approach

To measure real-world bystander response, we ran an in-the-wild deployment of a telepresence robot roaming the engineering building at McGill University. Rather than a lab study with scripted interactions, the robot moved through genuine hallway traffic, and we conducted on-the-spot interviews with people who encountered it.

We tested reactions across three conditions:
- Condition 1 - Animated Blinking Eyes
- Condition 2 - Male Face
- Condition 3 - Female Face

There were numerous people who saw and interacted or avoided the robot, from whom we interviewed 37 individuals. To analyze the interview data, two researchers independently used the MoFASA framework - which is built to trace why a specific person decided to show a specific response to a robot - which they perceive as appropriate. The researchers picked 15 interactions (5 from each condition) and analyzed them independently. The two researchers reconstructed the same underlying explanation for the participant's behavior in [12 of 15 / 80%] of cases — strong evidence that the patterns we found reflect real, repeatable drivers of bystander behavior, not one analyst's read of the room.

Key Insights & Discoveries

Our research showed that bystander trust in an unfamiliar robot is shaped less by the robot's capability and more by how legible its presence and purpose are:

• Ambiguity, Not the Robot Itself, Drives Avoidance: The most common driver of distance keeping wasn't fear of the robot, it was the uncertainty about its purpose. Participants who couldn't tell what the robot was doing there, or who it belonged to, consistently kept their distance regardless of how they felt about robots in general.

• A Human Face Changes the Interaction, Not Just the Impression: When the robot displayed a visible human presence (a face, a person on-screen), bystanders were more likely to treat the encounter as a social one - waving back, approaching, or engaging directly - rather than treating the robot as an object to route around.

• Physical Presentation Reads as Function: Participants repeatedly connected the robot's appearance directly to how they judged it. A robot perceived as visually "messy" read as more unpredictable and fragile, while a cleaner, more legible design read as more trustworthy - independent of what the robot actually did.

• Context Is Requested, Not Assumed: Across conditions, one of the most consistent asks from participants was simply more context: visible signage, a clearer sense of ownership, or an explanation of purpose. In its absence, people filled the gap with their own assumptions.

The Takeaway

Successful robot deployment in shared human spaces isn't primarily an engineering problem - it's a legibility problem. Our findings show that bystanders don't need a robot to be more capable to trust it; they need it to be more explainable. By pairing robots with visible human presence, clear framing, and clean physical design, institutions and brands can deploy robots into public space without triggering the discomfort and avoidance that ambiguous automation typically produces.