Brain Trust (28 page)

Generally, there are two ways to study networks. One is by creating mathematical models—like Simon Levin at Princeton, who we met in the entry about trendsetting and fish schools—and then poking and prodding these models in various ways. The other is to study existing networks, which allows you to see how behaviors or information flow in the real world (see, for instance, the book
Connected
, which details the ways behaviors like smoking and obesity flow through a social network in a
Massachusetts town). The first allows you to adjust the network design to see how tweaks affect its function. The second allows you to see real effects in real people.

But what if you could do both, at the same time?

Michael Kearns found a way to have his cake and eat it too. “About six years ago I started running these behavioral lab experiments in which I bring moderately large groups of people into the lab, impose various network structures, give them a game to play for real money, and study how they do,” says Kearns, CS and information science guru at the University of Pennsylvania and the Wharton School. Thus he could design and tweak the network, while also seeing how real people act within it.

For example, in one of his games, subjects tried to coordinate colors. Kearns gathered thirty-six undergrads and stuck them in a room so that they could only see certain of their neighbors, then he started a one-minute timer. Everyone in the room could display red or blue, and if at any time in that one minute everyone coordinated on either color, the whole room got paid. If not, no payday for the undergrads. After the minute, Kearns switched the network structure, and they played again.

So there was high incentive for agreement. “We designed this just after the 2008 democratic primaries,” says Kearns, in which you might remember Democrats had a high incentive for agreement—the sooner Dems could coordinate on Clinton or Obama, the sooner they could start hacking at the throats of Republicans instead of at one another. This proved challenging because, for one reason, while unanimity was good for all, people within the network tended to feel strongly that either Clinton or Obama was more good.

Similarly, Kearns threw into the mix of his lab games unequal payoffs—for some people, coordinating on red paid $1.50 while agreeing on blue paid only $0.50. For others, the reverse was true. So you’d really rather agree on your high-paying color, but
barring that, the other’s better than nothing. Did these networks still get paid, or did they devolve into infighting and general uncoordinated badness in which everyone suffered?

It depended on the network.

For example, in one experiment Kearns gave thirty players a blue preference and only six players an equal and opposite red preference. But he stuck the shorthanded red players in special spots in the network, with the largest number of neighbors. “We ran twenty-seven experiments like this,” says Kearns, “and in twenty-four of these, the network was able to reach an agreement.” Can you guess which color they agreed on? In every case, it was the preference of the highly connected minority—little, social red carried the day. “The minority opinion will dominate the outcome if the minority is sufficiently well connected,” says Kearns. This might remind you of the effect of special-interest lobbyists.

You also might remember the earlier Simon Levin story, in which he showed mathematically that the behavior of a small, committed minority needed connectivity to flow through a population and change social norms. Well, you can see the mathematical model in human action in Kearns’s lab.

But one interesting finding is that even within these scripted networks in which Kearns says he “tries to shoehorn human subjects into settings in which they perform like ants,” personality continues to influence outcomes. One aspect of personality that’s especially clear in coordination games is stubbornness—are you willing to switch your color in the face of a developing majority that disagrees with you? But the effect of stubbornness isn’t necessarily all bad, as you might expect. Sure, if a network includes too much stubbornness, players end up entrenched in their little, opposing fiefdoms. But its opposite is equally detrimental—if everyone’s too willing to flip-flop, the network does just that, oscillating wildly between colors without ever coming to full consensus on either.

This is like the group of indecisive friends trying to pick a restaurant. Maybe we should eat Thai! Sure. Or what about Mexican? Um, OK. Or Chinese? Sounds good. And you end up getting all muddled in overagreement. At some point the network needs some stubbornness.

In another game, Kearns gave humans a game that’s classically hard for computers. In the map-coloring puzzle, you have four colors and must shade the countries on a map so that no two neighboring countries are the same color. If tiny Switzerland goes orange, it affects the color of China, and the rippling of this change quickly requires massive computational chutzpah.

But humans role-playing countries color themselves rather quickly. “One lesson I’ve learned that transcends all types of experiments,” says Kearns, “is that I’m surprised how good people are at this stuff.” He points to this as hope for more and more ambitious crowdsourcing.

Still, there are things that computers do better than humans, and things that one brain does better than many. “If a problem can be broken into a gazillion pieces, you can crowdsource,” says Kearns. But if the pieces themselves require coordination, a problem may still best be solved by good old-fashioned expertise. I apologize in advance for the following sports simile (and not even a sport I play), but it’s like golf: Sure you could crowdsource a hole, with hundreds of people teeing off and then playing only the best ball. But wouldn’t this problem be more efficiently solved by pre-2009 Tiger Woods?

Imagine your problem. Any problem.

Is it “chunkable,” like needing thirty recipes for lightning-fast dinners or the best Monty Python quotes or suggestions from your geeked-out friends for scientists to interview for a book you’re writing? If so, you might throw it out to FB or Twitter or whatever social networking site seems most applicable (be sure to provide incentive, likely framing it as entertainment or offering some sort
of credit to the solvers). If the problem requires backstory and foresight, consider looking up a leading expert or making yourself into one. Or is it simply a question of firepower? Likely, there’s software and/or a bigger, badder box to help with that.

And then join Kearns in hoping that someday soon there will be a middle path that uses all three (see following coolness).

“I have a research fantasy that we’re far from
but that I like to think about sometimes,” says Kearns. Today there exist “compilers” that take a computational problem and recruit components from a network of computers to solve it. This allows you to design a problem without worrying about memory management, or CPUs, or virtual versus physical memory, or any of the other computational limits of your solving system (within reason). “I like to imagine a crowdsourcing compiler,” says Kearns. This compiler would break down a problem into its components and then recruit the optimal tool for solving each. Maybe one chunk requires expertise—the compiler would scroll through the Proceedings of the National Academy of Sciences publications until finding, recruiting, and motivating the top expert. One chunk could simply be computed, and the compiler would pull together the resources for it. And another component might best be crowdsourced, and the compiler would put out feelers into the human online world, creating an incentive like a game or a salary that gets a human network to solve the needed piece.

“We’re moving into a new era,” say Kearns, “in which human computing interfaces with computer computing.” This isn’t the old sci-fi scenario of übertech dominating humans, nor is it today’s model of humans using tech as tool, but a completely new scenario in which humans and the machines we’ve created collaborate to solve problems in ways neither could possibly do on their own.

Puzzle #17:
Map Problem

Use only four colors to shade the following map so that no touching states share the same color.

If you’ve ever watched
Survivor
, you know that not all tribes are created equal. Some are rancorous and repressive, me-centered and backstabbing, while others are cooperative and inclusive, honest, and even idealistic. David Logan, expert in organizational communication at USC’s Marshall School of Business, knows how to make your tribe the latter.

As you might imagine, Logan’s studied these tribes mostly in the context of businesses, which he divides into five tribal stages.

The first he defines with the phrase “life sucks.” “It’s not that people in these organizations don’t have individual core values but that the organizational culture says you have to undermine these values to survive,” he says. You may be forced to cheat to get ahead in the company or encouraged to lie to customers. Thus the battle between core and company values and the overall sucking of life.

In the next tribal level, it’s not that life sucks as a whole, only that each individual thinks “my life sucks.” “Employees say ‘I made suggestions but nobody listened,’ or otherwise deflect accountability,” says Logan.

Stage three includes 48 percent of the organizations Logan’s documented in his eight-and-ongoing years of study. This stage is defined by the idea that “I’m great and you’re not,” he says. People might have positive individual relationships with many others in the tribe, but there’s little coming together. You might solicit other group members to gain agreement for your ideas, but it creates little pods of stage two around the core group.

Leveraging the spirit born of shared values, 22 percent of tribes are able to make the leap from “I’m great” to “We’re great.” This is stage four—“the first stage at which the group becomes aware of
its tribalness,” says Logan. You can tell you’re there when a two-person conversation that’s interrupted absorbs and integrates the interrupter—if you’re all truly in the same tribe, there’s every reason to be inclusive and none to be exclusive.

So that’s it—the four stages of tribal development. You can read more about it in Logan’s book
Tribal Leadership
(with John King and Halee Fischer-Wright).

Only, that’s not it. There’s a fifth stage, “and these groups create amazing things,” says Logan, “like reconciliation in South Africa or Apple famously asking the question, How can I create a computer so simple that even my mom could use it?” The theme of a stage-five tribe is “life is great,” but the problem is that stage-five tribes can be idealistic to the point of being dreamy and not tied to the market, “like an Internet start-up that says ‘We don’t need cash, we’ve got clicks!’ ” says Logan. In his view, it’s ideal to stay at stage four, while infrequently dipping into stage five to ask, How do we make history? or How do we shake up the industry? “Stage five is pure leadership,” says Logan, pointing out that stage four is a nice mix of leadership and management, while stage three is pure management drowning out leadership, and below that not even management functions.