Can a clear invitation make strangers more likely to talk?

The hardest part may be knowing who is open

Most strangers do not know whether the person beside them wants a conversation or wants to be left alone. A smile, headphones or an empty chair can all be read in more than one way. The social risk is small but real: one person may avoid an interaction both people would have welcomed, while another may start one the other person did not want.

That makes a visible preference potentially useful. It does not promise chemistry. It reduces one piece of uncertainty before either person acts.

The experiment made social preference visible

Researchers created an outdoor seating area on a university campus. People who entered were randomly assigned to see a female research confederate with one of three colored cup-and-plate signals: open to conversation, prefer solitude, or a control color with no social meaning. Participants in the two signaling conditions displayed their chosen social preference; control participants stated a preference but received white plates. The confederate did not know what the colors meant, made eye contact and smiled, but did not initiate conversation.

The study was preregistered before data collection. Of 812 people who entered, the final analysis included 563 participants in 498 sessions after applying the planned exclusions. Most were university students, and some arrived with one or two companions. The researchers accounted for people sharing a session in their models.

Those exclusions matter. Ninety-five people did not notice the confederate, while 154 in the two meaningful-signal conditions did not correctly identify the signal. The result therefore describes what happened when the cue was noticed and understood—not what any colored object would do by itself.

A clear invitation changed reported behavior

Participants reported a new interaction counted by the study as a new connection at different rates across the three assigned conditions:

The social-versus-control difference was statistically significant in the preregistered logistic model.

Reported interactions were also longer and deeper on the study’s rating scales. Mean interaction length was 1.84 with the social signal versus 1.30 in control; mean depth was 2.03 versus 1.38. Those are scale scores, not minutes or a count of friendships. The measured behavior was a conversation or connection reported during one campus session.

The three conditions reveal more than a simple yes. An explicit invitation increased reported interaction, while an explicit preference for solitude reduced it. Legibility worked in both directions.

More conversation did not improve every outcome

The social signal did not produce a conventionally significant direct increase in reported state social connection over control. Positive mood was only marginally higher than control, with p = .051, and was not significantly higher than in the solitude condition.

The authors’ preregistered mediation models were consistent with reported interaction carrying some benefit for connection and mood. But an indirect model does not erase the null direct comparisons. The most secure finding concerns reported behavior: when openness was explicit, participants were more likely to report interacting in this setting.

Openness is context, never consent

One campus, one confederate and a mostly young student sample cannot establish the same effect in a train, bar, workplace or culture with different norms. Participants reported their interactions; conversations were not independently recorded and assessed. The experiment also says nothing about durability, compatibility or whether a new contact became a friend.

Most importantly, openness to a conversation is not blanket consent. People can change their mind, decline an approach or end an interaction. A solitude signal should work as a boundary, not as a challenge. Our editorial inference is to design for reciprocal clarity: make a preference easier to read while preserving an easy way to say no.

One encounter cannot reveal what follows

Assignment to the visible cue changed the probability that participants reported talking with the confederate. That is already meaningful: access can start with reducing uncertainty, not with changing a person. But the experiment stops at the encounter. It cannot tell us whether a new, relatively loose connection ever carries information or opportunity beyond that moment.

Participants reporting a new connection

Open-to-talk signal

56%

Uninformative control

41%

Prefer-solitude signal

20%

Li et al. (2025). The analysis followed preregistered exclusions and describes signals that participants noticed and correctly understood.