Social Networks : Evolving Adaptive Organizations

James Park was a Senior Consultant at Ernst & Young’s Center for Business Innovation where he studied organizational and strategic applications of complexity science, intangible sources of value such as human capital, and performance measurement methodologies. Prior to joining Ernst & Young, James was a special assistant at the U.S. Department of Labor, where he worked on corporate outreach initiatives for the Secretary of Labor. James holds an MA in Industrial/Organizational Psychology from The George Washington University.

In the spring of 1999, a group of Marines stormed the beaches of Northern California in an exercise called “Urban Warrior,” where they tested combat tactics, new technologies, and a different way of organizing—one that diverges from their traditional command-and-control, top-down structure. The Marines realize, as do many progressive companies, that in an environment characterized by uncertainty, a rigid, hierarchical organizational structure can inhibit important characteristics such as flexibility and adaptability. For the Marines, this uncertain environment stems from the complexities of new threats to national security—warlords in Mogadishu, separatists in Kosovo, thugs in Haiti.

The Marines are calling their new approach “networkcentric warfare,” and it involves rethinking the fundamental ways in which they have worked for over 200 years. Network-centric warfare refers not only to an electronic network—new communication technologies are connecting Marines through satellites and LAN systems—but also refers to how Marines define the way they organize on the battlefield. Through exercises like “Urban Warrior,” the Marines want to find out if they can transform themselves into a human network—a structure that many believe to be more adaptive and flexible than a traditional hierarchy.1

Network Organizations

Like the Marines, many U.S. corporations are trying to become more adaptive by moving away from centrally coordinated hierarchies toward a variety of more flexible structures. Intel, for example, often utilizes diverse, cross-functional teams that form around specific business issues or projects. Its organization looks more like a web of teams than a clearly defined hierarchy. 3M also does not follow the traditional approach to organizational design. 3M consistently achieves its goal of having 15 percent of its revenue come from new products by providing managers with the latitude to move from one business unit or laboratory to another without bureaucratic obstruction. Project groups, operating with few constraints from the formal organization, come together to accomplish a task and disband when their work is completed.2

One thing all of these new organizational forms have in common is that they resemble webs or networks—clusters of specialized units coordinated by communication and relational norms rather than by a hierarchical chain of command. “Specialization” is the key word because increased specialization enhances organizational adaptability in several ways. First, by focusing on more narrowly defined task domains, specialists accumulate large amounts of in-depth knowledge and expertise. By focusing their attention, they are also better able to monitor—and more likely to recognize and correctly interpret—indicators concerning impending environmental shifts likely to affect their special areas of expertise. Finally, groups of specialists operating in concert are more likely to craft creative and proactive solutions to complex organizational problems—complex problems are easier to solve when they are broken down into their component elements, and each component is tackled by specialists.3

Scientist Stuart Kauffman calls this division of complex problems “patching.” In his modeling of fitness landscapes,4 Kauffman demonstrates that for a particularly complicated optimization problem with many interdependent components and constraints, it is often useful to divide the problem into a set of nonoverlapping pieces, or patches, and then solve each patch independently. Kauffman uses the image of patches on a quilt to illustrate this idea. The patches are interdependent—optimizing one patch may change the problem to be solved in another patch. For this reason, patch size is very important. Patches that are too small will interfere with one another as each tries to optimize for its own purposes. In contrast, patches that are too large will cause the system to become stable and rigid, with each patch unable to co-evolve out of a suboptimal situation. A patch size that falls between these two extremes allows the system to stay on the edge of chaos, settled into a mutually optimal and orderly pattern, from which it can evolve. What does this mean for business? A hierarchical company with too much control at the top is likely to freeze rigidly into poor compromise solutions. A company broken into too many small, selfishly optimizing departments may churn chaotically.5 From these insights, one potential hypothesis is that the structure of an adaptive organization looks more like a network of interconnected patches, rather than a hierarchy.

If network structures are more adaptive than traditional forms, why aren’t more organizations adopting them? One primary reason is that researchers and practitioners interested in issues of organizational design lack the empirical tools to adequately address how best to coordinate the flow of information and knowledge through a network structure. Recent discoveries, however, taken from a diverse group of disciplines—social networks, graph theory, applied mathematics—combined with the power of computer simulation may provide companies with a tool to help them better understand the dynamics of network organization.

In order to begin to move toward efficient and robust forms of network organization, firms need to do three things:

1. Recognize and accept that networks exist in their organizations.
2. Visualize those networks.
3. Optimize those networks.

Recognize and Accept Networks

Creating an organization that looks more like a network than a hierarchy is not an easy task. It requires, first, a shift in mind-set (i.e., executives must realize that the old top-down, command-and-control structure is ineffective, and in many cases counterproductive) and, second, it requires the ability to implement a new framework that is more flexible and adaptive. A good first step is for people to recognize that embedded in almost every organization is an informal network, which describes what people actually do in order to accomplish their work (e.g., communities of practice). People often ignore formal organizational structures, finding them ineffective and unresponsive. To complete a task, they reach out for resources and relationships they need and they follow leaders of their own choosing, people they know they can rely on. If formal organizational charts do not represent, or support, the way that work actually gets accomplished, they should be modified so that they do.

Visualize the Network

Once the existence of informal networks within an organization is accepted, empirical tools that allow one to actually visualize those networks are necessary to better understand their influence. Researchers and practitioners use a construct called social network mapping (also known as organizational network analysis or organizational network mapping) to provide a clear snapshot of the organization’s informal network structure. Social network mapping uses survey methods to collect data about interactions among individuals—who talks to whom, the nature of the communication, the frequency of the interaction. Once collected, the data is analyzed and then presented both visually and statistically. Mapping interactions this way allows companies to better understand the state of their workplace dynamics. Network mapping tools can x-ray a chosen segment of the organization and derive tangible, quantifiable findings about how work actually gets done.6

Visually, social network mapping produces an organizational chart with lines that connect the boxes of people who exchange information, and a second organizational chart with lines drawn between seekers and givers of approval. Through sophisticated statistical analysis, further detail is available, such as the frequency of interaction, how often information is exchanged, and how often approval is sought and given. The results derived from social network mapping confirm what employees and managers have always intuitively known about their organization but have never been able to articulate—much less document.7

What would happen, however, if these tools were also able to tell you the “optimal” network structure for your firm—one that was connected in to allow a flow of knowledge throughout the organization that kept all employees sufficiently informed but did not, simultaneously, overwhelm or inundate them with information. Recent discoveries by Duncan Watts and Steven Strogatz may provide the answer.

Optimize the Network

In 1998, Duncan Watts, then a postdoctoral fellow in the social sciences at Columbia University and now a postdoctoral fellow at the Santa Fe Institute, and Steven Strogatz, a mathematician at Cornell University, discovered a phenomenon that they called small-world networks. Most research on network structure assumes that networks are either completely ordered or completely random. Research demonstrates, however, that many real-world networks—computer networks, social networks, employee networks—embody elements of both order and randomness. These hybrid models are small-world networks. Small-world networks are unique in that they consist of lots of little clusters, in which a few members of each cluster have connections to other, more distant parts of the network. These longer connections make it easy to find short paths through the network from any one point to any other.



What Watts and Strogatz found was that any set of connected, dynamic components—be they people, electric power stations, or brain cells—can be transformed into a small-world by introducing a few links that serve as short cuts between components. Relatively few short cuts can make big changes in a network, linking clusters of people, power stations, or brain cells together in unexpected ways.8

Translated into a new approach for organizational design, these findings suggest that employee networks in an organization could benefit from the rewiring of links among individuals to enhance their efficiency and robustness. An optimal organizational architecture promotes both efficient communication and effective teamwork. Small-world networks could provide the right balance, allowing a high level of clustering, with only a few links between any two parts of the organization. Careful placement of a wellconnected employee, for example, could go a long way toward improving interactions among departments, without the need to completely restructure them. Clustering remains high (existing teams aren’t disbanded), but the separation of elements is made smaller (the employee forms a link between two formerly disparate groups).9

While Watts and Strogatz’s model of small-world networks provides important insights into optimal network structure, the divide between theory and practices is still significant. In their work, all elements were treated as identical, and all ties were symmetric and equally weighted—assumptions that don’t hold up in the real world. What’s needed is a simulation that is able to model people’s actual decision-making strategies about whom to connect with and whom not to connect with, taking into consideration things like preferences, social norms, and personality types. Building these types of constraints into the model would make it much more representative of the real world. An initial extension of Watts and Strogatz’s work using an complexity-based simulation was done by Kai Shih, from Ernst & Young’s Center for Business Innovation and Stanford University. Through his simulation, Shih was able to “evolve” network structures that exhibited “small world” characteristics, using few, relatively simple rules and constraints. The potential of this work is that you could utilize the same algorithm used in Shih’s model and integrate it with an organizational mapping program. This would allow you to take a current state representation of an organizational network and evolve it into a small-world network, using rules and constraints that accurately represent the firm’s current situation. Instead of randomly rewiring an organization, one would be able to actually run through multiple scenarios “in silico,” testing various organizational structures until an optimal form was discovered.

What do these insights mean for the Marines and other companies searching for ways to become more adaptive? Watts and Strogatz’s model has the potential to provide the framework for more efficient and robust forms of organization. The key, however, is to make the leap from the theoretical to the practical. To transform an organization into a “small world” requires tools that better model the constraints of a real-world environment. Simulations, such as Shih’s, could allow users to evolve small-world networks from a current state situation, using simple rules that take the dynamics of the organization into consideration.

1. Garreau, Joel, “Point Men for a Revolution: Can the Marines Survive a Shift from Hierarchies to Networks?” Washington Post,March 6, 1999.
2. Tetenbaum, Toby J., “Shifting Paradigms: From Newton to Chaos,” Organizational Dynamics,Spring 1998.
3.  Walker, Orville C., “The adaptability of network organizations: Some unexplored questions,” Journal of the Academy of Marketing Science,25 (1), 1997.
4.  Landscape analysis is a mathematical process used to search for the best solution to a multi-variable optimization problem. It utilizes a fitness function, an equation that describes how well each possible solution meets the criteria specified in the problem. Within the search space (i.e., a plot of all the possible solutions), the fitness of a particular solution can be shown graphically as its height. The resulting “landscape” will have peaks in the regions of the search space that contain better solutions; further analysis of the peak regions can be used to refine the results and achieve even more precise solutions.
5. Kauffman, Stuart, At Home in the Universe(New York: Oxford University Press, 1995).
6. Doloff, Phyllis Gail, “Beyond the Org Chart,” Across the Board, Februrary 1999.
7. Ibid.
8. Watts, Duncan J., & Strogatz, Steven H., “Collective Dynamics of ‘Small-World’ Networks,” Nature,June 1998.
9. Watts, Duncan, “It’s a Big Company, But a Small World,” Exec Online,January 1999.

© James Park

blog comments powered by Disqus