ORMS Today
December 2000

Designing Organizations

CMOT launches success on a solid, scientific foundation

By Douglas A. Samuelson

For a few minutes, forget our profession's usual focus on mostly technical, highly specific problems, and consider a new answer to an old question: how can we design organizations that work well?

No, this isn't about tools for planning and budgeting, or models of how an organization's objectives can be optimally attained. The design of the organization itself is the subject of a new spin-off field from mainstream OR/MS: computational and mathematical organization theory (CMOT).

As the name implies, this approach does involve mathematical and computational methods. Artificial intelligence, particularly agent-based systems models, and simulation are the methods most often used. The novelty is that these and other methods are applied not to how the organization does specific tasks, but how the organization, its components and its environment interact. This integrated view enables analysts to solve complex performance problems.

One of the acknowledged leaders of the field is Kathleen Carley, professor jointly in the Heinz School of Public Policy and Management, the Department of Social and Decision Sciences, and the Department of Engineering and Public Policy at Carnegie Mellon University. She focuses largely on social networks, especially knowledge transfer. Powered by a 1999 National Science Foundation grant, CMU has the largest graduate program in CMOT (renamed Computational Analysis of Social and Organizational Systems, or CASOS) in the country. Using the grant funding, CMU has also taken over hosting the annual workshop which was held for many years in conjunction with the INFORMS spring national meeting. This year's workshop, held in July, brought together about 70 participants and highlighted the variety and accomplishments of the field and its leaders.

Designing Successful Project Teams

Ray Levitt, professor of civil engineering at Stanford, focuses on developing systematic ways to put effective teams together — balancing costs, risks and expected functions. "My first exposure to practical problems of organizations," he says, "was in construction projects for my father's company when I was 14, growing up in South Africa. We had a bunch of wonderful engineers who were lousy managers. I studied management, and eventually concentrated in construction management at Stanford, including three courses from (Stanford organizational scientist) James March. I branched out into psychology. I came away convinced that we ought to be able to design organizations as we design bridges — balancing risks, costs and benefits in a rigorous, systematic way. We just had to figure out how."

Levitt studied construction safety for his Ph.D., using stochastic learning theory, decision analysis and other applied probability tools. "I began to see the power of that approach," he says. "With the right kind of computer tools, which I haven't found yet, you ought to be able to analyze organizations with finite-element tools."

What Levitt has found — created, actually — is a powerful modeling method, Virtual Design Teams (VDT), and simulation software (Vité) to run "what-if" analyses of how alternative designs would perform. The focus is on pattern matching, especially fitting skill sets into task requirements, by adjusting parameters in a discrete-event simulation. A process model diagram, as shown in Figure 1, represents the team.

Figure 1: Designing successful project teams: Process model of John Deere illustrates high concurrency of design and prototyping activities, causing many communication links (green) and rework cycles (red).

Vité is now a commercial spin-off, including software development, sales and service, consulting and support activities. The company has successfully tackled a number of cases, including:
  • a redesign for John Deere that cut the company's new product development time for heavy machinery in half and significantly improved quality, then captured the streamlined process completely enough to apply it to other product lines;

  • cutting a product development cycle from 19 months to 12 months for a major computer manufacturer — with only two days to do the analysis;

  • reducing development time from three years to two years for a sub-sea oil production module for Norway's Stratfjord Sub-Sea Satellite Project, while ensuring that it would operate reliably for several years, unattended, undersea; and

  • reducing development time from five years to one year for a complex launch vehicle for the commercial market for Lockheed Missiles and Space Company, while maintaining quality and reducing costs.
"Of course, studying at Stanford, I got excited about decision analysis," Professor Levitt says, "but I've run into a lot of problems applying it to complex construction projects. It's well documented that decision analysts often run into trouble getting the decision-makers' utilities, which turn out to be critical to the analysis. What we encountered in contractor risk pricing was even worse; we couldn't decide whose utilities we needed! Is it the contractor, the city manager, the city engineer, the city council, the transit authority, the riders, the voters or who else? With the tools we've developed, we can let everyone play 'what-if' and be confident of reaching a workable solution."

Matching Organizational Culture to Function

Professor Richard Burton of the Fuqua School of Business at Duke is another long-standing member of the CMOT community. He, too, traces his work back to ideas introduced by James March and economist Herbert Simon, but he has worked in many ideas from organizational psychology and sociology as well. He developed OrgCon, a rule-based expert systems program that assesses misfits between an organization's aims and environment and the strategy, technology, management style, degree of formality, degree of hierarchy and degree of complexity, among other factors, its leaders claim to prefer.

The underlying theory is summarized in a book he and Borge Obel, of the Department of Management at Odense University, Denmark, wrote five years ago. The key idea is to define several critical aspects of an organization and then lead the user, via an easy-to-follow interface, through questions to elicit his or her perception of where the organization fits into this framework. OrgCon's rule-based engine then points out possible misfits, such as a company trying to succeed in the fast-paced high-tech world with a bureaucratic, risk-averse management style and a high degree of centralization of decision-making.

Professor Obel recently used OrgCon in five Danish companies. While the results are still preliminary, he reports that one company decided to implement all the recommendations and expects a significant improvement in managing relationships between divisions, an area they had previously neglected. Another identified a critical need for resources to upgrade production technology and is now working to do that. A third had a less pleasant result; after the managers had extensively discussed their different views of the company's strategy and goals, an area the OrgCon analysis had identified as a major misfit for them, one manager decided to resign.

At the INFORMS national meeting in San Antonio in November, professors Burton and Obel presented a statistical study focusing on misfits they identified in 334 Danish corporations. They found that the presence of one or more misfits was associated with lower performance, as they had expected. Interestingly, the number of misfits, as long as there was at least one, did not matter much. "If everything is fine and you change one thing, you may create several misfits, because variables interact," Obel explains. His advice: "To improve an organization's performance, you have to take a holistic approach, not try to do it piecemeal."

Annual Workshop Displays Variety of Subjects

Organizers of the CMOT/CASOS workshop took Obel's advice to heart as evidenced by the 50 presentations at this year's event. Topics included:
  • diffusion of knowledge within organizations;

  • development of trust and collaboration;

  • depth vs. breadth in knowledge management strategies;

  • effects of different promotion systems on organizational performance;

  • information processing in medical care;

  • the electric power spot market;

  • farmers' adoption of agricultural innovation;

  • modeling consumer preferences and the acceptance of product innovation; and

  • how to work on the Moon and Mars.
Referring to the last topic, Maarten Sierhuis of the NASA Ames Research Center observes that when people on Earth try to work with sophisticated robots on Mars, "the 45 communication delay (between asking a simple question and receiving the reply) will change the nature of the collaboration significantly."

1978 Economics Nobel Laureate Herbert Simon, a long-standing member of Carnegie Mellon's faculty, served as the conference's keynote speaker. Professor Simon described the conference's topics as a natural outgrowth of his own work in economic decision-making. He broke away from the concept of companies devoted solely to maximizing profit. Instead, he postulates that decision-makers in a company attempt to find satisfactory solutions to the problems they face, taking into account what other decision-makers are doing. Decision-makers, he says, "satisfice" rather than optimize. This insight led to many connections between economics and other social sciences.

Now, Simon says he thinks about why market processes work so well with so little formal attention to fitting their mechanisms into other organizational processes. Markets work with little structure, he explains, because organizations and social systems consist of clumps that interact intensely at low levels. As you go higher in an organization, the processes are bigger and slower, and information takes longer to get recognized. "Your importance can be measured," he quips, "by how long you could be dead in your office before anyone noticed."

These bigger, higher-level processes also have more interdependence. Dividing processes among smaller, less interdependent sub-organizations increases adaptability and effectiveness. "Nearly decomposable organizations," Simon explains, "can improve more and faster, so they evolve more effectively."

Simon notes that people can know many more things by learning from others than they could by themselves, so everyone gains from belonging to social networks. This observation leads to bounded rationality, the idea that decision-makers settle for "good enough" solutions rather than always seeking to optimize. Such behavior is appropriate, because anyone's knowledge is a vast simplification of reality. "I believe in 'satisficing,' he explains, "because I can't find optima."

Therefore, Simon asserts, we see the growth of large organizations because they use identification with others to harness self-interest to the organization's interests, and they grow via combining social groups into a nearly decomposable structure. "The challenge," Simon concludes, "is to maintain a morality of centers of power," so that people feel comfortable remaining loyal to the organization, and the organization remains open to new information and perceptions. Asked about his most important piece of advice for graduate students, he replies, "Keep your curiosity."

The Role of Information Technology

Not surprisingly, the analysis of organizations now includes the subject of information technology. Late last year, a team led by Kathleen Carley and Noshir Contractor, professor of speech communications and psychology at the University of Illinois, won an NSF Knowledge and Distributed Intelligence Initiative grant to study how information technology is reshaping organizations, focusing especially on the evolution of knowledge networks. Ray Levitt, John Kunz and Francoise Barr from Stanford, Stanley Wasserman and Andrea Hollingshead from the University of Illinois and Peter Monge and Janet Fulk from the University of Southern California are also on the team. They claim the "project team is itself a knowledge network," with expertise in communication, computer science, engineering, psychology, sociology, organization science, social networks, statistics and urban planning.

Carley states, "This study will be the first large-scale application in which multiple computational models of organizations are simultaneously used to generate hypotheses and explore theoretical implications for organizations." The project will draw field data from at least a dozen organizations, including several multinational ones. "Today, information technology is radically altering organizations, and we can use that same technology in this inter-university project to study the evolution of new organizational forms in the 21st century," Carley adds.

Most likely we'll hear more about it, along with many other topics, at the next CASOS conference, set to take place at Carnegie Mellon in July 2001.

In The Beginning

Kathleen Carley traces the origins of CMOT back 13 years, when Michael Masich put together a conference in the Netherlands. Two years later, Richard Burton, professor of business at Duke, ran a workshop at a TIMS/ORSA (now INFORMS) conference, sponsored by the TIMS College on Organization. Carley recalls there were about 10 participants — modelers and some more general " organizations people."

The group grew and began to meet regularly, often in conjunction with the spring TIMS/ORSA meetings. "It was exciting," she relates, "because we had artificial intelligence people and organizational psychologists and sociologists all talking to each other." That wasn't and isn't happening often elsewhere.

About five years ago, Kluwer started the CMOT Journal, with Carley and colleague Al Wallace as editors. The Multi-Agent Systems Journal started up a year later.

The field has continued to grow and expand. Late last year, Carnegie Mellon won a five-year Integrated Graduate Education and Research Training (IGERT) grant from the National Science Foundation to support what Carley calls "a new kind of graduate education." The program, with 60 graduate students, pulls together computer modeling, organization theory and social networks, all focused on real-world applications.

— Douglas Samuelson


  1. Burton, Richard M., and Obel, Borge, "Strategic Organizational Diagnosis and Design: Developing Theory for Application," Boston, Massachusetts: Kluwer Academic Publishers, 1995.

  2. Computational and Mathematical Organization Theory Journal, Kluwer.

Doug Samuelson, a frequent contributor to OR/MS Today, is president of InfoLogix, Inc., a consulting company in Annandale, Va.

E-mail to the Editorial Department of OR/MS Today: orms@lionhrtpub.com

OR/MS Today copyright © 2000 by the Institute for Operations Research and the Management Sciences. All rights reserved.

Lionheart Publishing, Inc.
2555 Cumberland Parkway, Suite 299, Atlanta, GA 30339 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com/

Web Site © Copyright 2000 by Lionheart Publishing, Inc. All rights reserved.