Project Updates
The project started ahead of original schedule in July of 2003 since one of the Prof. John Koza’s colleagues, who was helping us with the initial scoping of our genetic programming design, was leaving the area. The early start of the project gave us the extra boost needed to achieve the promising initial results described below. The first and second year activities, some complete and some in progress, include the following:
In order to improve the design and development of the postprocessor, Bijan KHosraviani have attended a workshop sponsored by the National Science Foundation (NSF), conducted a literature review, and enrolled in relevant coursework. The NSF-sponsored workshop has provided an opportunity to become familiar with theories and computational modeling approaches used both in commercial and academic settings. The literature review has been focused to gain information about genetic algorithms, genetic programming, evolutionary methods, user-interface design, experimental designs, and optimization and validation of computational models.
We designed and developed Pro-Opt V1.0, the first version of the postprocessor that uses genetic programming (GP) techniques to find a near optimal solution to project organization problems. In this version, the GP postprocessor only modifies some of the individual/sub-team attributes such as skill level, attention allocation, and structural policies such as centralization, formalization, and matrix strength. By modifying just this limited set of attributes, Pro-Opt V1.0 was able to evolve an organization design whose performance against the specified set of objectives came within approximately 3% of the best design that student teams have been able to achieve in a realistic, standardized project organization design problem over the past five years.
In September of 2003, we submitted an NSF proposal to the Division of Design, Manufacturing and Industrial Innovation (DMII) of NSF, based on the first three months of work under this grant.  The proposal was evaluated as “good” by three reviewers but ended up just below the threshold of funding in this panel, with a recommendation to “revise and resubmit”.  With the breakthrough results that we have achieved in the last three months, we feel confident that we can resubmit a competitive proposal to NSF later this year.

We have our first CIFE working paper, "Organization Design Optimization Using Genetic Programming" (#WP085), in progress. We are planning to present this paper at the Genetic and Evolutionary Computation Conference (GECCO-2004) in Seattle Washington in June of this year. Subsequently, we plan to publish the final draft in one of the related scientific journals.
We are in the process of developing the second version of the Pro-Opt postprocessor. Version two will be able to evolve near-optimal configurations of three topological relationships that define a project organization:  the assignment of activities to team members, the supervisory structure of the project, and the structure of participation in various project meetings. We have already implemented the utility program that makes changes to activity assignments and percentage allocation for each activity. With this new addition, our GP postprocessor has been able to find the best solution ever found for the standardized project organization design problem mentioned above.

We submitted and presented a paper at North American Association for Computational Social and Organizational Science Conference in at Carnegie Mellon University in Pittsburgh in June 2004. Our presentation received good comments from scholars in the organizational science field such professor Richard Burton from Duke University. View the paper at Proposals/Paper section.

We submitted and presented a late-breaking paper at GECCO-2004 in Seattle Washington in June of 2004. Our paper won the silver medal for the 2004 Human-competitive award in Genetic and Evolutionary Competition. We were able to win this award because our automatically generated results satisfied the following three criterias: Our result was equal to or better than the most recent human-created solution to a long-standing problem for which there has been a succession of increasingly better human-created solutions. Our result solved a problem of indisputable difficulty in its field. Our result won a regulated competition involving human contestants. View the paper at  Proposals/Paper section. For more details see: http://www.genetic-programming.org/gecco2004hc.html