"Is Artificial "Systems" Research Possible?"
Artificial “life” research was an early specialty that contributed to the current explosion of work on complex systems. This talk will explore the possibility of a new, related, but unique specialty called Artificial “Systems” research (ASR). ASR would be based on a general model of systems that consists of a network of 75 systems mechanisms and hundreds of “linkage propositions” that describe how they influence each other. The propositions and their influences appear to be programmable in Prolog or LISP. The resulting dynamic network is an expression of what is common to many natural and social systems. ASR would place this network in computer space and systematically add, alter, or remove specific linkages to measure across computer time how the change affects overall measures of systems stability and/or performance. This talk will present arguments for the need for Artificial Systems Research programs and explore its initial tenets and assumptions. The talk will explore the similarities between Artificial Life Research and Artificial Systems Research. For example, both exist in computer space and require considerable, often dedicated, computer resources and time. The talk will also explore the differences between Artificial Life Research and Artificial Systems Research. For example, ASR programming would model general systems process interactions and not depend solely on characteristics of bio-systems as ALR does. The former depends on genetic algorithms using “mutation” and “recombination” of information sequences, while the latter does not. Also, selection, if it exists at all in ASR, would not be based on evolutionary selection. ASR would model a process of “emergence” that is distinct from the process of “evolution” as programmed in ALR. The talk will also list problems or obstacles inhibiting the development of ASR including identification of what might be optimized in systems structure and process when modeled as a general system rather than a specific, real system. How could ASR runs measure that optimization? What would be a generalizable systems output? The relationship between, or usability of ASR for work in Systems Pathology will be examined. Important challenges will be posited, such as, what would be the expected results of ASR, or what are the correspondence principles between Artificial Systems research and its results and the real world? The talk will also explore practical organizational needs for doing ASR and its prospects for funding.