Can scientists prove whether individual behaviors, such as eating or exercise habits, are contagious? Research from Alex 'Sandy' Pentland, Professor of Media Arts and Sciences and the Director of the Human Dynamics Lab at MIT, shows that may be possible.
"If individual behavior is contagious, then we can change this behavior either by changing the behavior of several influential elements in the social network, or by changing the social network itself," concludes Pentland.
Pentland and his research team studied 70 college students living in a residence dormitory at a North American university, with the student subjects spread evenly across the freshmen, sophomore, junior, and senior grade levels. Participants were given a Windows mobile phone, through which scientists collected data from self-reported surveys designed by experts in political sciences and medicine. Cell phone sensors recorded proximity and location every six minutes and documented communication.
The data from the surveys included:
- The relationships of close friends
- Political opinions and involvement
- Sharing Facebook photos
- Sharing Twitter and blog posts
- Participation in on-campus organizations
- Health conditions and physical exercise
- Demographic information
The data from these sensors also included dyadic
relationships such as who called whom, who sent short messages to whom, the physical
proximity of the individuals, and attributes such as Wi-Fi hotspot scanning,
which tracked where the individuals were at any given time.
Pentland's research team discovered that relationships and behaviors co-evolve over time in a student dormitory. For example, the team found:
- The shared living space—the dorm room, dorm halls, and common areas—were the most important factor for building friendships. Students were five times more likely to report another student in his sector as a friend than another student living in a different sector.
- The second most important factor for building friendships was academics and extracurricular activities. Students were five times more likely to count a classmate or teammate as a friend.
- Friends influence behavior by "infecting" one another with how frequently they visit places. The research showed 15 pairs of friends shared all on-campus activities that were surveyed, and 30% of friend pairs shared over 50% of their on-campus activities. In comparison, non-friends shared less than 10% of on-campus activities. In particular, pairs who participated in aerobic exercise about three times weekly were more likely to be friends.
Pentland and his research team were able to use the data
gathered from the mobile phones to model the dynamics of students' friendships. "By using this model," says Pentland, "We demonstrate how relationships and
individual behaviors interact with each other and are able to further predict
friendship and further project behavioral interactions."
As the saying goes, knowledge is power. And thanks to data analysis, the opportunity to study how relationships influence behavior has provided untapped potential in creating tools to help individuals change their behaviors. "The cell phones," says Pentland, "record relationships and behavior in much deeper detail, and provide almost limitless data-mining potential. The behavior and interactions recorded by the cell phones show clear daily, weekly, and yearly patterns, and increased behavioral complexity."
The next logical step is to use this data goldmine and analysis to create effective tools that help individuals do what has proved to be so difficult: change their behaviors, not only by giving them real time insight into the patterns that influence their everyday life, but by giving them specific approaches and suggestions to help them create a new pattern.
Alex 'Sandy' Pentland is a Professor of Media Arts and Sciences at MIT Sloan and frequently teaches in MIT Sloan Executive Education programs