Working together doesn’t always work

The benefits of working together are so widely recognized that they’re cliché. Two heads are better than one. Many hands make light work. Great minds think alike. None of us is as smart as all of us. Be a team player. If you’ve never said one of these middle management tropes, you’ve probably heard one (or seen one on an inspirational poster near the water cooler).

These phrases capture a universally held ethos that’s often instilled by the time you reach kindergarten: Sharing and helping out are qualities worthy of admiration. But in some situations, working together has negative consequences. It’s not that too many cooks spoil the broth; rather it’s that collaboration can prevent people from thinking outside the box. Are we on the same page? Let me explain.

There’s a lot of scholarship on whether sharing information improves problem solving. Research by both social science management types and computer science network scientists suggests that networks with a high degree of clustering, or connectedness, are a force for good. Connections encourage information sharing, which increases coordination, minimizes redundancy and generally improves the odds of success. Other research (also from both camps) suggests that clustering can undermine performance, stifling creativity and innovation. It turns out that both are true, and success depends on the task at hand.

An interdisciplinary team (that’s right) from Boston University, Harvard and Northeastern reasoned that the conflicting results might emerge because people were comparing apples and oranges. So the researchers decided do an experiment that clearly differentiated two kinds of tasks: seeking information (finding all the pieces of a puzzle) and seeking solutions (putting the pieces of the puzzle together). The researchers modified a web browser-based platform called ELICIT (for Experimental Laboratory for Investigating Collaboration, Information-sharing, and Trust), which was initially developed by a U.S. Department of Defense Command and Control Research Program in the wake of 9/11. Their experiments conducted via the platform entailed a whodunit task, in which participants had to predict the who, what and where of an impending terrorist attack.

Participants worked in teams of 16 that had one of four network structures: a caveman network that has four different cliques (four caves of four, each cave connected to two others); a hierarchy network (also four cliques, but arranged with a central clique connected to the other three); a rewired caveman network (a “small world” network, constructed by removing some links from the caveman network and adding others that create shortcuts); and the ring network (a ring where each person is connected to two others, one on each side, creating a circle with no clustering or central command).

The players could take four different actions: search for clues from an online fact set; share clues with neighbors they were linked to in their network; register theories in designated who, what, where, and when boxes; and check their neighbors’ solutions whenever and as often as wanted. After 51 rounds, a pattern emerged. The clustered networks fared better at information-gathering, or finding more pieces of the puzzle. But clustering inhibited coming up with unique solutions — different ways of putting the pieces together.

Teams with clusters gathered more information, not because they searched more but because they searched better, the researchers note in a new paper. Communication among team members, enabled by the caveman and hierarchy networks, made puzzle-piece finding more efficient and those teams found more unique facts. But unclustered networks (ring and rewired caveman) were better at creatively putting the pieces together. These less-clustered networks generated more unique theories than those in which people could communicate with neighbors and neighbors of neighbors.

“When you think about global organizations that communicate virtually, we tend to say, the more knowledge we can share, the better,” says study coauthor Ethan Bernstein an organizational scientist at Harvard Business School. “Our message is that that doesn’t necessarily come without trade-offs.”

For those of you in management, Bernstein and colleagues suggest that collaboration should be thought of as a switch that can be toggled, rather than permanently turned on. Working together is a good way to gather all relevant information, but for generating solutions, people should brainstorm by themselves. “In order to be creative in problem-solving, you might need to go your own way,” says coauthor Jesse Shore, an information systems scientist at Boston University.

Collaboration, like a truism captured in a cliché, works very well — so well that it is over-used and can become a missed opportunity for a more powerful approach. Working together should be a thoughtful choice, dependent on the desired outcome, not a silver bullet, one-size-fits-all solution. Because, to paraphrase Michael Jordan, while there is no ‘I’ in team, there is in win.”