what can a single individual do to increase rate of idea flow in their part of their social network? Fortunately, there are many ways. In 1985, Bob Kelly of Carnegie Mellon University launched the now famous Bell Stars study. Bell Laboratories, a premier research laboratory, wanted to know more about what separates a star performer from the average performer. Is it something innate or can star performance be learned? Bell Labs already hired the best and the brightest from the world’s most prestigious universities, but only a few lived up to their apparent potential for brilliance. Instead, most hires developed into solid performers but did not contribute substantially to AT&T’s competitive advantage in the marketplace.
What Kelly found was that star producers engage in “preparatory exploration”; that is, they develop dependable two-way streets to experts ahead of time, setting up a relationship that will later help the star producer complete critical tasks. Moreover, the stars’ networks differed from typical workers’ networks in two important respects. First, they maintained stronger engagement with the people in their networks, so that these people responded more quickly and helpfully. As a result, the stars rarely spent time spinning their wheels or going down blind alleys.
Second, star performers’ networks were also more diverse. Average performers saw the world only from the viewpoint of their job, and kept pushing the same points. Stars, on the other hand, had people in their networks with a more diverse set of work roles, so they could adopt the perspectives of customers, competitors, and managers. Because they could see the situation from a variety of viewpoints, they could develop better solutions to problems.
For the entire community, we measured activity levels by using the accelerometer sensors embedded in their mobile phones. Unlike typical social science experiments, FunFit was conducted out in the real world, with all the complications of daily life. In addition, we collected hundreds of thousands of hours and hundreds of gigabytes of contextual data, so that we could later go back and see which factors had the greatest effect.
On average, it turned out that the social network incentive scheme worked almost four times more efficiently than a traditional individual-incentive market approach. For the buddies who had the most interactions with their assigned target, the social network incentive worked almost eight times better than the standard market approach.
And better yet, it stuck. People who received social network incentives maintained their higher levels of activity even after the incentives disappeared. These small but focused social network incentives generated engagement around new, healthier habits of behavior by creating social pressure for behavior change in the community.
And:
Unexpectedly, we found that the factors most people usually think of as driving group performance—i.e., cohesion, motivation, and satisfaction—were not statistically significant. The largest factor in predicting group intelligence was the equality of conversational turn taking; groups where a few people dominated the conversation were less collectively intelligent than those with a more equal distribution of conversational turn taking. The second most important factor was the social intelligence of a group’s members, as measured by their ability to read each other’s social signals. Women tend to do better at reading social signals, so groups with more women tended to do better...
When people are behaving independently of their social learning, it is likely that they have independent information and that they believe in that information enough to fight the effects of social influence. Find as many of these “wise guys” as possible and learn from them. Such contrarians sometimes have the best ideas, but sometimes they are just oddballs. How can you know which is which? If you can find many such independent thinkers and discover that there is a consensus among a large subset of them, then a really, really good trading strategy is to follow the contrarian consensus. For instance, in the eToro network the consensus of these independent strategies is reliably more than twice as good as the best human trader.
And:
to answer the question of how habits form, my research group studied the spread of health behaviors in a tightly knit undergraduate dorm for one year. In the Social Evolution Study, led by PhD student Anmol Madan and myself, with Professor David Lazer helping with design of the experiment and data analysis, we gave all the participating students smartphones with special software so that we could track their social interactions with both close friends and acquaintances. In total, this study produced more than five hundred thousand hours of data and included face-to-face interactions, phone calls, and texting, as well as extensive surveys and weight measurements. These hundreds of gigabytes of data allowed us to examine what goes into the creation of habits.
One particular health behavior that we focused on was weight change and on whether this was more influenced by the behavior of friends or by peers in the surrounding community...
exposure to the behavior examples that surrounded each individual dominated everything else we examined in this study. It was more important than personal factors, such as weight gain by friends, gender, age, or stress/happiness, and even more than all these other factors combined. Put another way, the effect of exposure to the surrounding set of behavior examples was about as powerful as the effect of IQ on standardized test scores.
It might be asked how we can know that exposure to the surrounding behaviors actually caused the idea flow; perhaps it is merely a correlation. The answer is in this experiment we could make quantitative, time-synchronized predictions, which make other noncausal explanations fairly implausible. Perhaps even more persuasively, we have also been able to use the connection between exposure and behavior to predict outcomes in several different situations, and even to manipulate exposure in order to cause behavior changes. Finally, there also have been careful quantitative laboratory experiments that show similar effects and in which the causality is certain.
Therefore, people seem to pick up at least some habits from exposure to those of peers (and not just friends). When everyone else takes that second slice of pizza, we probably will also. The fact that exposure turned out to be more important for driving idea flow than all the other factors combined highlights the overarching importance of automatic social learning in shaping our lives.
And:
How do we choose who to vote for? Do our preferences also come from exposure to those around us? We tackled this question in the Social Evolution experiment by analyzing students’ political views during the 2008 presidential election.9 The question we asked was: Do political views reflect the behaviors that people are exposed to or are they formed more by individual reasoning? By giving these students specially equipped smartphones, we monitored their patterns of social interaction by tracking who spent time with whom, who called whom, who spent time at the same places, and so forth.
We also asked the students a wide range of questions about their interest in politics, involvement in politics, political leanings, and finally (after the election), we inquired which candidate had received their vote. In total, this produced more than five hundred thousand hours of automatically generated data about their interaction patterns, which we then combined with survey data about their beliefs, attitudes, personality, and more.
When sifting through these hundreds of gigabytes of data, we found that the amount of exposure to people possessing similar opinions accurately predicted both the students’ level of interest in the presidential race and their liberal-conservative balance. This collective opinion effect was very clear: More exposure to similar views made the students more extreme in their own views.
Most important, though, this meant that the amount of exposure to people with similar views also predicted the students’ eventual voting behavior. For first-year students, the size of this social exposure effect was similar to the weight gain ones I described in the previous section, while for older students, who presumably had more fixed attitudes, the size of the effect was less but still quite significant.
But what did not predict their voting behavior? The views of the people they talked politics with, and the views of their friends. Just as with weight gain, it was the behavior of the surrounding peer group—the set of behavior examples that they were immersed in—that was the most powerful force in driving idea flow and shaping opinion.
From Pentland’s Social Physics:
More (#2) from Social Physics:
And:
More (#1) from Social Physics:
And:
And: