Think about your social network. How many of your friendships started out as a “friend of a friend”? How often have your professional connections come through a common contact?

This familiar pattern points to a fundamental principle in social networks. For decades, social scientists have observed that if two individuals share a mutual connection, they are significantly more likely to form a direct tie themselves — a phenomenon called “triadic closure.” Yet, a core question remains: Does the mutual tie cause this connection, or do other factors — such as shared interests or simply more chances to meet — account for its prevalence?

In a new paper, “Tendencies toward triadic closure: Field-experimental evidence,” now published in Proceedings of the National Academy of Sciences, MIT Sloan School of Management professor Dean Eckles, Cornell University professor and MIT visiting professor David G. Rand and Oxford Internet Institute, University of Oxford professor, and MIT research affiliate Mohsen Mosleh offer the first clear experimental evidence that the simple existence of a shared social connection plays a direct role in forming social ties, and offer new insights into the forces behind human connection.

Unravelling the ‘Why’ Behind Our Connections

“If you map out an individual’s social network, you’ll usually see a lot of triangles, where two people with a shared acquaintance are also friends themselves,” explained Eckles. “But if I’m friends with Bob, and he’s friends with Alice, do I become friends with Alice because I want to know Bob’s other friends, or just because I see her frequently at Bob’s barbecues or because we all share a hobby?”

It’s typically almost impossible to know. But in the digital world, these conditions are far easier to isolate. In their study, which builds on previous work on how birds of a partisan feather flock together online, Eckles, Rand, and Mosleh used X (formerly Twitter) to randomly modulate the presence of a mutual connection and measure its impact on triadic closure.

The Experiment: Isolating the Significance of a Minimal Connection

The researchers’ study centered on a field experiment where they created convincing “human-looking” bot accounts (“A users”). These A users followed active X users (“B users”) – not influencers, but regular individuals. Some B users quickly followed the A accounts back. Next, researchers identified thousands of “C users” – individuals who followed one of the B users, but had no other connection to the bots. This created an “open triad”: the C user followed the B user, and the B user now followed one of the A users. Then came the core of the experiment where the researchers used the bot accounts to follow C users and tracked whether the C users followed them back. To determine the effect of a mutual connection, C users were divided into two groups:

• For the first group (the “open triad” condition), the researchers maintained the connection between the A and B users. When an A account followed a C user in this group, the C user’s profile would display the B user as a mutual follower with the A account.
• For the second group (the control condition) researchers strategically broke the A-B connection by briefly blocking and unblocking accounts before the A account followed the C user. This ensured no mutual connection was visible on the A user’s profile. (Alternative control conditions followed the C user with a different A user that did not share a connection.)

“The key was to create a situation where the only difference between our experimental groups was the presence or absence of a single, visible mutual connection,” said Rand. “On platforms like X, if someone follows you and you go to their profile, you instantly see if you have mutual followers. The effect of this minimal cue was what we were testing.”

The presence of a mutual connection increased the likelihood that a C user would follow back the A user by 35%. While significant, this increase is much smaller than the rate of triadic closure typically seen in observational studies of existing social networks.

“Our findings clearly support the idea that people are disposed to close an open triangle per se — to make the connection of a connection, ” Eckles explained. “But the effect isn’t nearly as powerful as earlier observational data. This suggests that other factors, like shared interests, also come into play in real life when friends of friends befriend each other.”

Professional Implications

Social scientists have suggested that overly strong tendencies toward triadic closure could be a barrier to accessing new information and opportunities, since these new contacts were already only two hops (or degrees of separation) away.

“Usually, most of our closest friends and family are already connected and tend to have the same information to offer us,” said Mosleh. “So often triadic closure leads us to form largely redundant connections, perhaps at the expense of forming ties that would bridge to new communities and new opportunities.”

The value of these new connections is especially clear in a professional context, where access to new information can lead to innovation, advancement, and opportunities. The study’s finding that a mutual acquaintance alone can drive new relationships, said Eckles, should encourage professionals to both actively leverage existing contacts to expand networks.

The team believes that managers, too, can take advantage of the tendency toward triadic closure. For example, strategically encouraging new relationships that span team functions or locations, potentially through tools like Slack or company intranets, can help employees gain critical knowledge and work more effectively.

“In remote working environments, where mutual acquaintances are less likely to meet casually, it’s especially important to seek out these interactions,” said Eckles. “Food for thought the next time you send a connection request.”

Download the full paper, Tendencies toward triadic closure: Field-experimental evidence, authors Mohsen Mosleh, Dean Eckles and David G. Rand, now published in Proceedings of the National Academy of Sciences journal.

Image credit: MIT Sloan School of Management.

-ends-