Who would you rather have as a fitness partner: a paragon of athleticism and dedication who could motivate you to exceed your current level of fitness or an equal, with whom you could exchange tips and encouragement on the road to better health?

Or neither? According to a new study led by University of Pennsylvania’s Damon Centola, participants in an online fitness program ignored the fitness aptitude of their potential partners.

“Instead they chose contacts based on characteristics that would largely be observable in regular, offline face-to-face networks:  age, gender and body mass index,” Centola said.

As more people turn to the Internet to help them improve their health and fitness, Centola, an associate professor in the Annenberg School for Communication, wanted to examine how people sought out health partners in an online forum. He coauthored the study with Arnout van de Rijt, an associate professor in Stony Brook University’s Department of Sociology. The paper was published online in the journal Social Science and Medicine.

Centola and van de Rijt partnered with an existing online fitness website to recruit 432 participants to be part of their new “Health Improvement Network.” All of the participants shared ten pieces of information: their age, gender, ethnicity, body mass index (BMI), fitness level, diet preferences, goals for the program and favorite exercise, as well as their average exercise minutes and intensity level. 

The researchers divided the participants into six groups. Each participant was then randomly partnered with six “health contacts” in their group with whom they could exchange information. Over a five-week period, the participants were given the opportunity to select new health contacts and drop existing ones. The only information on which they had to base their choice was the set of ten characteristics that the other members of the group had shared. Participants had no knowledge of who each other’s health contacts were, or whether there were any “highly connected” individuals. This allowed the study to reveal which characteristics participants would use to make their connections.

The researchers anticipated that group members would select health contacts who shared similar exercise routines or interests, or even fitness “leaders” who were very fit and could serve as motivational role models. Yet in five of the six communities, participants did neither. Rather, the community members showed a strong inclination to choose contacts whose age, BMI and gender were like their own.

The participants’ tendency to “make ties to ‘the devil they know,’” the authors wrote, “may unintentionally limit their opportunities for finding health information from sources that they are not normally exposed to.”

The findings suggest that although people in online health programs are beckoned with the possibilities of meeting healthier people who can provide them with information about new kinds of exercises and better strategies for getting healthy, they self-select into networks that look very similar to the kinds of networks that people typically have offline: people with similar age, gender and BMI profiles as themselves.

Health programs can work around this human tendency, Centola noted, by actively recommending “health buddies” based on characteristics that are hard to connect to offline, but easy to find online, such as people who are good motivational partners, or partners who prefer similar exercises, or are working to increase their endurance to similar levels. 

“Our findings suggest that the trick to an information-rich online community,” Centola said, “is to encourage new kinds of ties by reminding participants just how valuable these online relationships can be.”

The research was supported by the James S. McDonnell Foundation, the MIT Research Support Committee and the National Science Foundation.