Open System Science

Founder’s Message
Since its establishment, Sony Computer Science Laboratories has had “open systems” as its consistent theme. When the institute was first established, we had a vague apprehension about using conventional scientific methodologies in conducting research on artificial intelligence and distributed systems. Over time, I came to understand that the reason for this uneasiness was that the subject of our research was open systems.

Open systems are to closed systems (closed systems). In the past, science and technology have solved problems by defining the domain of the problem, dividing it into simple subproblems, solving each, and reconstructing the results. In recent years, however, we have reached a situation where we must solve problems that cannot be defined in isolation and cannot be easily divided into subproblems. Common examples are problems of the global environment, sustainability, economic phenomena, and life. These problems are complex and interrelated with a wide variety of elements, and the state of the subsystems containing those elements changes simultaneously.

For artificial systems, we can mention huge systems connected by the Internet and man-machine interaction. In such large systems, the boundaries of the system and the services provided change from moment to moment, and it is almost impossible to have sufficient knowledge of the elemental systems needed to control the entire system. In addition, to provide a truly user-friendly environment, one must know humans well. Humans, however, are extremely multidimensional, and their behavior is strongly dependent on circumstances and time. We felt that it would be impossible to understand and attempt to solve these problems solely on the basis of reductionism.

In 2008, Sony Computer Science Laboratories proposed a new scientific methodology to solve these major problems, which is called “Open Systems Science.” In addition to “analysis” and “synthesis,” this new methodology includes the concept of “management” of temporal changes in systems. This methodology was subsequently expressed in the following concrete steps.

If the behavior of the model is inconsistent or deviates from the behavior of the real system over time, the model is modified. If there are unacceptable inconsistencies or deviations, the model is modified, and if necessary, the domain of interest is changed. Whereas the methodology of conventional science gives us a way to dig deeper, the methodology of open systems science gives us a way to understand and solve problems in a constant interrelationship. We believe that the complementary use of these two methodologies provides the means to solve the problems that have confronted us in recent years.

The practice of open systems science cannot be done in the confines of a laboratory. In the laboratory, we may be able to explore problems in depth by working hard in a given area of research. However, unless you look at the field, you will not be able to solve the problem with an understanding of the interrelationship between that problem and other problems. This is where Sony CSL’s “cross-border, take-action” research attitude comes into play. In other words, crossing borders and taking action are the essence of practicing open systems science.

Mario Tokoro
Sony Computer Science Laboratories, Inc.
founder