Complex Systems May be More Controllable Than They Appear

 Complexity  

Researchers have developed a theoretical framework for analyzing individual controls within networks based on numbers of sources and sinks for information flow. By this method, the number of controls required by a network can be predicted and direct comparisons for the basis for control across networks of differing size, structure, and function can be made.

Researchers have found a new way of decomposing the structures that give rise to the need to control complex systems at various points. Using this model, they developed a new network statistic, called the control profile, that both highlights significant differences between real networks and commonly-used random models.

The two brothers, Profs. Justin and Derek Ruths, from Singapore University of Technology and Design and McGill University respectively, have suggested, in an article published in Science, that all complex systems, whether they are found in the body, in international finance, or in social situations, actually fall into just three basic categories, in terms of how they can be controlled.

"While our framework does provide insights into the nature of control in these systems, we're also intrigued by what these groupings tell us about how very different parts of the world share deep and fundamental attributes in common."

Related articles Chris Eliasmith Discusses SPAUN's Artificial Intelligence  System That Merges Human and Artificial Intelligence Awarded US Patent  Researchers Connect Entropy To Intelligent Behaviour

They reached this conclusion by surveying the inputs and outputs and the critical control points in a wide range of systems that appear to function in completely different ways.

"When controlling a cell in the body, for example, these control points might correspond to proteins that we can regulate using specific drugs," said Justin Ruths. "But in the case of a national or international economic system, the critical control points could be certain companies whose financial activity needs to be directly regulated."

One grouping, for example, put organizational hierarchies, gene regulation, and human purchasing behavior together, in part because in each, it is hard to control individual parts of the system in isolation. Another grouping includes social networks such as groups of friends (whether virtual or real), and neural networks in the brain, where the systems allow for relatively independent behavior. The final group includes things like food systems, electrical circuits and the internet, all of which function basically as closed systems where resources circulate internally.

Referring to these groupings, Derek Ruths commented, "While our framework does provide insights into the nature of control in these systems, we're also intrigued by what these groupings tell us about how very different parts of the world share deep and fundamental attributes in common – which may help unify our understanding of complexity and of control."

"What we really want people to take away from the research at this point is that we can control these complex and important systems in the same way that we can control a car," says Justin Ruths. "And that our work is giving us insight into which parts of the system we need to control and why. Ultimately, at this point we have developed some new theory that helps to advance the field in important ways, but it may still be another five to ten years before we see how this will play out in concrete terms."



SOURCE  PhysOrg

By 33rd Square

Embed