What is to be Done

The outline above of major areas of complex systems research and applications provides a broad view in which many specific projects should be pursued. We can, however, single out three tasks that, because of their importance or scope, are worth identifying as priorities for the upcoming years: (1) transform education; (2) develop sets of key system descriptions; and (3) design highly complex engineering projects as evolutionary systems.

Transform Education

The importance of education in complex systems concepts for all areas of science, technology, and society at large has been mentioned above but should be reemphasized. There is need for educational materials and programs that convey complex systems concepts and methods and are accessible to a wide range of individuals, as well as more specific materials and courses that explain their application in particular contexts. A major existing project on fractals can be used as an example (Buldyrev et al. n.d.). There are two compelling reasons for the importance of such projects. The first is the wide applicability of complex systems concepts in science, engineering, medicine, and management. The second is the great opportunity for engaging the public in exciting science with a natural relevance to daily life, and enhancing their support for ongoing and future research. Ultimately, the objective is to integrate complex systems concepts throughout the educational system.

Develop Sets of Key System Descriptions

There are various projects for describing specific complex systems (NOAA 2002; Kalra et al. 1988; Goto, Kshirsagar, and Magnenat-Thalmann 2001; Heudin 1999; Schaff et al. 1997; Tomita et al. 1999), ranging from the earth to a single cell, which have been making substantial progress. Some of these focus more on generative simulation, others on representation of observational data. The greatest challenge is to merge these approaches and develop system descriptions that identify both the limits of observational and modeling strategies, and the opportunities they provide jointly for the description of complex systems. From this perspective, some of the most exciting advances are in representation of human forms in computer-based animation (Kalra et al. 1988; Goto, Kshirsagar, and Magnenat-Thalmann 2001; Heudin 1999), and particularly, in projecting human beings electronically. Pattern recognition is performed on realtime video to obtain key information about dynamic facial expression and speech, which is transmitted electronically to enable animation of a realistic computer-generated image that represents, in real time, the facial expression and speech of the person at a remote location (Goto, Kshirsagar, and Magnenat-Thalmann 2001). Improvement in such systems is measured by the growing bandwidth necessary for the transmission, which reflects our inability to anticipate system behavior from prior information.

To advance this objective more broadly, developments in systematic approaches (including quantitative languages, multiscale representations, information capture, and visual interfaces) are necessary, in conjunction with a set of related complex systems models. For example, current computer-based tools are largely limited to separated procedural languages (broadly defined) and databases. A more effective approach may be to develop quantitative descriptive languages based on lexical databases that merge the strength of human language for description with computer capabilities for manipulating and visually representing quantitative attributes (Smith, Bar-Yam, and Gelbart 2001). Such extensible quantitative languages are a natural bridge between quantitative mathematics, physics, and engineering languages and qualitative lexicons that dominate description in biology, psychology, and social sciences. They would facilitate describing structure, dynamics, relationships, and functions better than, for example, graphical extensions of procedural languages. This and other core complex systems approaches should be used in the description of a set of key complex systems under a coordinating umbrella.

For each system, an intensive collection of information would feed a system representation whose development would be the subject and outcome of the project. For example, in order to develop a representation of a human being, there must be intensive collection of bio-psycho-social information about the person. This could include multisensor monitoring of the person's physical (motion), psycho-social (speech, eye-motion), physiological (heart rate), and biochemical (food and waste composition, blood chemistry) activity over a long period of time, with additional periodic biological imaging and psychological testing. Virtual world animation would be used to represent both the person and his/her environment. Models of biological and psychological function representing behavioral patterns would be incorporated and evaluated. Detailed studies of a particular individual along with comparative studies of several individuals would be made to determine both what is common and what is different. As novel relevant convergent technologies become available that would affect human performance or affect our ability to model human behavior, they can be incorporated into this study and evaluated. Similar coordinating projects would animate representations of the earth, life on earth, human civilization, a city, an animal's developing embryo, a cell, and an engineered system, as suggested above. Each such project is both a practical application and a direct test of the limits of our insight, knowledge, and capabilities. Success of the projects is guaranteed because their ultimate objective is to inform us about these limits.

Design Highly Complex Engineering Projects as Evolutionary Systems

The dramatic failures in large-scale engineering projects such as the Advanced Automation System (AAS), which was originally planned to modernize air traffic control, should be addressed by complex systems research. The AAS is possibly the largest engineering project to be abandoned. It is estimated that several billion dollars were spent on this project. Moreover, cost overruns and delays in modernization continue in sequel projects. One approach to solving this problem, simplifying the task definition, cannot serve when the task is truly complex, as it appears to be in this context. Instead, a major experiment should be carried out to evaluate implementation of an evolutionary strategy for large-scale engineering. In this approach, the actual air traffic control system would become an evolving system, including all elements of the system, hardware, software, the air traffic controllers, and the designers and manufacturers of the software and hardware. The system context would be changed to enable incremental changes in various parts of the system and an evolutionary perspective on population change.

The major obstacle to any change in the air traffic control system is the concern for safety of airplanes and passengers, since the existing system, while not ideally functioning, is well tested. The key to enabling change in this system is to introduce redundancy that enables security while allowing change. For example, in the central case of changes in the air traffic control stations, the evolutionary process would use "trainers" that consist of doubled air traffic control stations, where one has override capability over the other. In this case, rather than an experienced and inexperienced controller, the two stations are formed of a conventional and a modified station. The modified station can incorporate changes in software or hardware. Testing can go on as part of operations, without creating undue risks. With a large number of trainers, various tests can be performed simultaneously and for a large number of conditions. As a particular system modification becomes more extensively tested and is found to be both effective and reliable, it can be propagated to other trainers, even though testing would continue for extended periods of time. While the cost of populating multiple trainers would appear to be high, the alternatives have already been demonstrated to be both expensive and unsuccessful. The analogy with paired chromosomes in DNA can be seen to reflect the same design principle of redundancy and robustness. These brief paragraphs are not sufficient to explain the full evolutionary context, but they do resolve the key issue of safety and point out the opening that this provides for change. Such evolutionary processes are also being considered for guiding other large-scale engineering modernization programs (Bar-Yam 2001).

0 0

Post a comment