Event Schedule:

Start Date:    24th October 2017End Date:    24th October 2017
Start Time:    09:30 AMEnd Time:    12:30 PM

Event Agenda

This year, when celebrating its 60th birthday, the EU made a bold claim about its future: it will become a global player on the next stage of high-performance computing, known by the label of exascale computing. This raises the question in which fields Europe may choose to develop a comparative advantage in future high-performance computing – in which fields to form new communities, know-how, infrastructures and markets. One such field could be the understanding and mastering of global challenges.

The conference will focus on four example global challenges:

  • Developing a sustainable and resilient global financial system
  • Addressing the daunting risks of pandemics
  • Transforming the fossil-fuel based global mobility system
  • Creating forms of democracy adequate to the age of digitalization

In global systems, large numbers of human agents interact in complex network structures embedded in a shared environment.

The system dynamics arises from these interactions – as in models of molecular dynamics, where many particles interact to give rise to the system's behaviour.For the latter, HPC methods, tools and codes exist; moreover, there are equations that describe the system at the macro-level.

For global systems, no macro-level equations exist that describe the complex systems dynamics. These dynamics can be explored with the help of agent-based simulation models, which differ from molecular dynamics models in several respects:

  • heterogeneity of different types of agents needs to be accounted for, a synthetic population of agents provides a "social coordinate system", in which sets of heterogeneous data are attached to agents
  • social networks between agents are not regular grids; they go beyond spatial proximity relations and come with a set of network properties such as heavy tailed degree distributions, hierarchical structures and assortativity; not only the agents, but also the links between them may be of different types, and the networks themselves co-evolve over time with the agents' actions
  • the environment in which agents interact is spatially differentiated and also co-evolves with the agents' actions; for example there may be an interplay between agents' actions and niches the agents are located in

In those cases where networks between agents have an important influence on the model outcome, where effects to be modelled concern a small subgroup of a population (as for example an initially small group of adopters in addressing the question whether a certain "green" innovation can trigger a sustainability transition), or where path dependency can lead to differing outcomes resulting from small changes at the micro-level of the system, models at the individual level are necessary in order not to "aggregate away" important elements of the evolution of a global system.

Further, the process of GSS model development is fundamentally based on iterative loops of model definition, implementation and analysis. Thorough exploration and analysis of stochastic model dynamics throughout large parameter spaces is necessary in this process. One main aim is to identify the "turbulent zones" of interest, where rapid changes occur in a social system. Such change may be desired in some cases, as for example when looking for policy measures that can induce a transformation to sustainable mobility. In other cases, the goal will be to steer the system in a safe distance from turbulent zones, as for the example of an epidemics taking off.

Some fields in GSS are already using this new type of models – for example, epidemiology, to represent statistically correct dynamics of encounters between individuals for analysing the spread of a virus – for other topics, such models are being developed.

This conference wants to explore how GSS and HPC can join forces in further development of large scale simulation modelling for addressing societal challenges

  • by investigating, for the four examples, applications in GSS that show a need for HPC support to enable them to advance beyond the current state
  • by looking into directions of research and development at the intersection between HPC and GSS that can accomodate GSS structures in HPC-efficient simulation models and can lead to HPC-efficient simulation analysis methods for these
  • while grounding the discussion in the practical dimensions of addressing global challenges, that is, with a focus on the type of questions decision makers have and informed by their expertise

To this end, the conference shall bring together three groups of experts: scientists from the fields of GSS and HPC and practitioners working on these challenges.

In the longer run, agent-based models with refined representations of human individuals, that may be developed by merging GSS and HPC, may lead to a larger need for computational power than in molecular modelling, since the complexity of humans and their interactions, including decision making procedures, vastly exceeds that of molecules. Typical structures in global systems may then also help shape future developments in HPC and HPDA.

   Event Venue: San Francesco Complex, Piazza San Francesco 19, 55100 Lucca, Italy

Join our Newsletter

Receive your dose of sustainable news
signing up for our newsletter

   Global e-Sustainability Initiative (GeSI) c/o Scotland House Rond Point Schuman 6 B-1040 Brussels Belgium

   info@gesi.org, press@gesi.org

   +32 2 282 84 42