Research suggests that while bad performers are weeded out in large groups, they may thrive in smaller ones
When NYU mechanical engineering student Elizabeth Krasner (Tandon ’20) is in class or people-watching on the street, she can’t help but notice how certain individuals seem to draw others toward them, naturally taking on a leadership role in a study group or among friends.
And after conducting more than two dozen experiments and publishing a paper on the subject, she offers a simple explanation for the phenomenon: superior performance.
“Usually whoever comes up with better ideas is associated with, if not called out as, a leader,” Krasner says. “What we think is human nature is human nature, but it’s also mathematically true. It was cool to see that in the data.”
In large groups, the idea of majority rule, or the “wisdom of the crowd,” is well documented. But less is known about how individuals make decisions in smaller settings. Krasner was part of a team of researchers led by NYU Tandon School of Engineering professor Maurizio Porfiri and postdoctoral researcher Shinnosuke Nakayama, who designed a study to test how inter-individual relationships emerge in small groups. The resulting paper, “Social information and spontaneous emergence of leaders in human groups,” was published last semester by the Journal of the Royal Society Interface. According to the study, “good performers emerged as leaders over time, exerting stronger influence on others.”
And the implications are far reaching: understanding how networks of people function is crucial to the study of epidemics, for instance. (Porfiri and the study’s co-author, Lorenzo Zino, collaborated in December 2018 on an article about how social activity affects the spread of disease.)
In the leadership study, participants were randomly chosen and assembled in groups of five. Each person was given a clicker connected to a display screen that flashed a certain number of dots for half a second at a time. After seeing the image, participants used their clicker to guess how many dots they had seen; all their answers were anonymously displayed on the screen along with the past performance of each participant. Then, the subjects had 10 seconds to change their answer, if they wanted, to another.
Over time, instead of choosing to rely only on their own intuition or to follow the wisdom of the crowd—whichever answer was selected by the most people—participants “processed the performance of others in deciding how to change their answers,” according to the study. Good performers were more likely to be copied even when their answers were in the minority, the study found.
The authors also conducted an experiment in which the past performance of people’s initial guess—not the answer they chose after they had an opportunity to change their response—was displayed on the screen. In that version, people changed their answers more on average than in the original experiment. In both versions, participants increased their individual performance by changing their answers.
Nakayama says the coevolution of the individual performance and the network was one of the most interesting findings for him.
“If a leader has a wrong choice, then that information will propagate,” he says.
The experience was both a scientific and personal boon for Krasner, who helped design the experiment, build the clickers, create the user interface for the screen, and code the underlying software. She also ran “each and every one” of the experiments, and now enters her senior year as a published author.
Porfiri says the results now inspire him to ask: “What are the drivers of network change? Why are people choosing to build contacts with other people?” He and Nakayama aim to continue to study network formation in small groups, including, potentially, the effect of a “malicious attack” on the network, such as an actor posing as a participant intentionally getting a wrong answer.
“The wisdom of the crowd is very good for filtering out bad performers and malicious attacks,” Porfiri says. “But in a small group, will we suffer?”