Courtesy of Genetic Programming, Inc.

In the complex and rapidly changing business environment of the early 21st Century, designing an effective and optimized organization for a major project is a great challenge. Project managers have to rely on their experience and/or trial and error to come up with organizational designs that fit their particular projects. The Virtual Design Team (VDT) simulation system, based on the information processing theories of Galbraith (1977) and March and Simon (1958), has helped project managers and organizational designers to model and analyze project organizations in the computer before implementing them in practice. VDT was developed with initial seed funding from CIFE, and can predict schedule, cost and process quality performance of a user-specified organization and work process. However, VDT has no ability to improve or optimize current design automatically. The user must experiment in “What if?” mode with different alternatives to find better solutions that can mitigate the identified risks a given project configuration.  Based on her or his expertise, the user must set up the model, run the simulator, analyze the output, make changes to the input, and repeat these steps until an acceptable output is achieved. The problem has many degrees of freedom so the search space for better solutions is vast.  Thus, exploring the search space manually can be an intimidating activity. It relies on the expertise of the human user and offers no guarantee of optimality.

The proposed research will first design and develop a “post-processing” optimizer for VDT using evolutionary computational methods to help project managers find near optimal designs for their project organizations. Next, we will validate our post-processor by comparing its recommended organization designs to predictions of organizational “contingency” theory. Finally, we plan to conduct organization design charrettes to verify whether or not our model can help project managers design better organizations. 

A preliminary version of our postprocessor optimizer beats the best human trial-and-error solutions developed by more than 40 teams over the past eight years.  The postprocessor was awarded a Silver Medal for human-competitive results in genetic and evolutionary computation at GECCO-2004 Conference.

This project is designed to integrate with a parallel research effort being conducted by Michael Murray, another Ph.D. student in the VDT research group.  Michael's project is being carried out as part of the Global Change Project, a seed project that was also initially funded by CIFE in 2002, and subsequently funded by NSF in 2003 as part of its Management of Knowledge-Intensive Dynamic Systems (MKIDS) program. (The Stanford MKIDS project is described in a Stanford Report Article of June 10, 2003.)   Michael's part of the MKIDS grant attempts to build a post-processor for optimizing resource pool sizes and task start times in VDT using a combination of AI and OR techniques.  The resource pool size and task start dates output from Mike's system will provide the starting point to seed the genetic programming and genetic algorithms analysis that is the core of this research.