Cornell built a robot with mirrors so strangers could catch each other's eye

Civil inattention: why nobody makes eye contact in public

Sociologists have a name for what happens in that waiting room: civil inattention. In public, people quietly agree not to take much notice of each other. You look at your phone. So do they. It isn't cold so much as polite.

That politeness also blocks the first step. The paper's abstract opens by saying eye contact between strangers, "even fleeting, can spark interaction and foster connection, happiness, and belonging," and that in public spaces such moments are often suppressed, with "many people absorbed in their phones."

We weren't just trying to trigger conversations, but to support the very first moment of social connection, which is the eye contact.

That's what makes this interesting. Most icebreakers are about getting people to speak. This one goes after the step before that: giving two people's eyes somewhere to meet. Eye contact with strangers means different things in different places, which is part of why that step is so hard.

The first study: 16 pairs of strangers and a waiting room

The first paper appeared at HRI 2026, a human-robot interaction conference, and won best paper in its Design Track. The work came out of Cornell's Architectural Robotics Lab, led by Keith Evan Green.

Sixteen pairs, 32 people aged 18 to 50, were told they'd be doing a short-term memory task. They learned what the study was really about afterward. Here's how it went:

• Each pair sat in a waiting room about 12 feet by 12 feet, with three chairs.
• The robot came out from behind a screen. It stood four feet tall, and it was small and made of soft material so it wouldn't intimidate anyone.
• Its two mirrors moved to pre-programmed positions. In one you saw yourself, and in the other you saw the other person.
• Guo was operating the robot remotely. It wasn't deciding anything on its own.

Twelve of the 16 groups reported that their first meaningful contact came through the mirrors and not face to face. People reacted in different ways. Some pairs tried to make sense of the robot together, others engaged with it, and some used the mirrors to cautiously check whether the other person seemed open to talking.

That last one deserves a second look. Glancing at someone in a mirror asks a lot less than turning your head and looking right at them. If they don't respond, nothing awkward has happened, and you can both act as if it never did. That's our reading and not a finding from the paper.

The second study: the mirrors, the robot, or both?

The second paper, at CHI 2026, asks a separate question: is it the mirrors or the robot that matters? It's a 2×2 between-subjects study with 90 participants, 45 pairs, in four conditions:

• MirrorBot, with both the robot and the mirrors
• Bot-only
• Mirror-only
• A control, with neither

The abstract reports that "MirrorBot led participants to initiate conversation more often, feel greater closeness and togetherness, and have more enjoyable interactions." The authors argue that robots can act as more than social agents, as "socio-spatial interfaces that choreograph sight lines and shared attention in physical space."

The abstract reports a direction without per-condition numbers, so we can't tell you how big the gap is.

One small discrepancy: Cornell's release says the related paper had 40 pairs and the abstract says 45 dyads. We've gone with the abstract.

Not everyone wanted a robot's help

The most useful part of this work may be where it didn't go well.

A few participants felt uncomfortable... We saw them turn away from the robot and frown, which created an awkward moment between the two participants. One participant actually said it felt like 'an overenthusiastic friend,' pushing two reluctant people to talk.

The team's takeaway is a good one: "It suggests social technology should not only know how to initiate interaction, but also know when to step back."

Guo also names another limit of novelty. Unusual objects can get people talking, "but people often end up talking about the object itself, rather than becoming curious about each other." The team feels MirrorBot is different because the focus is on the other person. That's their view, and it will take more evidence to settle.

This isn't the only attempt to help strangers past the first step. See the Happy to Chat badge study, and keep in mind what the person being approached is weighing.

What this research can't tell us

• The robot was teleoperated. It wasn't reading the room on its own.
• This all happened in a lab waiting room, with people who'd agreed to take part in a study. It wasn't a bar, an event or a street.
• The samples are small, 16 pairs and 45 pairs, and we've seen no Taiwan data in either.
• The second paper's abstract doesn't give the size of the gaps.
• There's no long-term follow-up in what we read, so we don't know if anyone became friends.

One more thing to be clear about: this research has nothing to do with Limere. We weren't involved, and Limere doesn't work this way. We're writing about it because the question is a good one: how do two people get past that very first step?

Other work has tried a similar idea. SocialStools, from 2023, used three moving stools that generate sound and imagery with 12 groups of three strangers, and togetherness emerged. Another 2023 CHI study gave 49 pairs of strangers questions through a humanoid robot, a screen or a poster. People complied most with deep questions from the robot, and the robot or screen brought more happiness and connection than the poster. These are small lab studies too. The shared idea is that a third thing between two people can make it easier to start.

What to take with you this week

You won't have a robot with you, and you don't need one. These are our own suggestions drawn from the research, and the studies didn't test them.

• Look up. Put the phone away for a moment. If there's nowhere for someone's eyes to land, nothing can happen.
• Stand where there's something shared to look at, like the bar, a stage, an exhibit or a band. When everyone's facing the same thing, starting to talk feels more natural.
• A short glance is plenty. If they look away, take it as not right now, and don't keep looking.
• If you do catch someone's eye and want to say something, talk about what's in front of you. That's a lot lighter than commenting on them.

The research also carries another reminder: the other person may not want the invitation at all, and being overeager is itself uncomfortable. No response is an answer. For what to say when you do speak, read how to start a conversation at a bar or event.

If you caught someone's eye and didn't say anything, here's what to do about the person you saw at an event. Or next time, try Limere to meet people at the same bar or event. There's no robot involved, and no pressure to force anything. Just pick a night you already wanted to go out.