After the 3.11 earthquake many people realized that there are events that cannot be reasonably anticipated. These "unexpected" events occur as an outside of the anticipated envelope (e.g., Tsunami of 14m high vs the anticipated max of 5.7m), or something completely unheard of (e.g., Tokyo subway gas attack in 1995). We recognize that these unexpected events do happen, but because they are "unexpected," we cannot prepare for the event and protect our systems. The only thing we can do is to recover from the damage as quickly and as inexpensively as possible. This process may not be exactly a "recovery" because it may not restore the system into the original configuration; rather, the system can be a completely new configuration that is also acceptable, or even desirable, to the stake holders. The ability to do this generalized recovery is, in our definition, "resilience."
There are many natural and artificial systems that demonstrate resilience. Some examples are ecological systems, engineering systems such as computer networks, and financial and social systems. We will explore these systems and draw out the common characteristics of resilient systems, and develop tools and methodologies for designing resilient systems. Our research is truly trans-disciplinary -- we will pull together insights from biology, computer and information science, economics, and social science. We welcome any ideas that are worth sharing. Together we will be able to help build a safer and more sustainable world.
We currently plan to work on the following four subprojects:
1. Mathematics of unexpectedness
We cannot build a resilient system only by focusing on the resistance aspect of the system since a cost for building a system that resists against an enormous disaster, such as the 3.11 earthquake, would be unreasonably expensive. For example, it is probably not economically feasible to build a seawall high enough to withstand a tsunami following that earthquake.
We claim that we should take the perspective that system failures are inevitable and thus adopt recovery strategies for building systems that recover from various damages in a flexible way. Our goal in this subproject is to develop a mathematical theory for analyzing risks of rare, but significant events such that we can take a right balance between costs for resistance and those for recovery when building a resilient system.
2. Resilience in biological systems
Biological systems must possess features that confer robustness to survive both constant stress and periodic catastrophic events. The “Resilence in Biological Systems” subproject seeks to identify the characteristics of organisms, populations, and ecosystems that allow tolerance and recovery from stress.
- Population genetics of deleterious mutations:
Genetic mutation is unavoidable and large-‐scale population genetic analyses have revealed that the genomes of a wide variety of taxa are subject to a constant input of weakly deleterious mutations. Because population size is finite, such mutations can spread within populations and damage biological function over evolutionary time. A collaboration between Akashi (NIG) and Mano (ISM) will test whether “diminishing returns” fitness relationships contribute to the ability of small populations to recover from genome degradation.
- Ecological genetics and recovery from environmental stress:
Because environmental disturbances are rapidly changing the ecosystem, it is increasingly important to understand the genetic and ecological mechanisms of adaptation and speciation in natural animal populations. Kitano (NIG) investigates the effects of human disturbance (urbanization and pollution) as well as environmental catastrophes (tsunami) on the ecology, evolution, and conservation of natural stickleback fish populations.
- Biological systems in extreme environments:
The terrestrial environment in Continental Antarctica is extremely severe. Imura (NIPR) studies how populations of mosses, lichens or bacteria can establish and remain maintained in such environments. Risks of invasion by non-‐indigenous species are also under study.
3. Resilience in social systems
We argue that a social system whose rules or regulations are defined in a bottom-up way is more resilient than that whose rules are established by a central authority, since the best knowledge for solving a given social problem is owned by people involved in that problem.
We consider the concept of "co-regulation,’’ which has been increasing considered in European countries, and plan to study how we should build a legal system by allowing private parties to incorporate self-regulations into a legal system while reducing its risks and incompleteness with centralized regulations of public sectors. In this project, we consider legal systems for cyber security and privacy protections, and plan to develop methodologies for building a resilient society in a bottom-up way.
4. Computational theory of resilience
Resilience is a new concept of transdisciplinary science, and can be seen in many natural and artificial systems. To take a systems approach to resilience, we explore the basic theory of resilience from the computational viewpoint. The main objectives of this subgroup are to answer the following three questions:
- What are general computational principles of resilient or non-resilient systems?
- How can we compute new acceptable states in the face of new or unexpected events?
- How can we design resilient systems?
Then we evaluate and apply our theories to biological, engineering or social systems.
We are particularly concerned with network dynamics. In an increasingly inter-connected world, where each entity is related with other entities, small local perturbations in the world may cause complex effects on a global scale. Moreover, existence of positive and negative feedback loops makes analysis of network dynamics very hard. While the dynamics of networks is modeled by differential equations in physics, in informatics it is modeled in either discrete systems like automata, Boolean networks and constraint networks or probabilistic systems like Baysian/Markov networks and causal networks. Then theories of dynamic networks are considered in four aspects: (i) sensitivity of networks with respect to the initial values (chaos) or the modeling and parameters, (ii) (un)predictability of states, (in)completeness of the modeling, and (un)reachability in the possible worlds, (iii) syntactic and semantic distances between states of a network, and (iv) updating networks by satisfying rational postulates and integrity constraints and by maximally satisfying soft and dynamic constraints.