GSDP was coordinated by


The role of ICT in Global System Science

Below is a draft by Cezar and me about the role of ICT in Global System Science. Comments are welcome!

Kind regards,
  Patrik & Cezar

The role of ICT in Global System Science

It is a trivial observation that the emerging science of global systems relies essentially on computer models.  All science nowadays relies on computer models (for an interesting analysis of just how much, see ``Simulation and its Discontents'' by Sherry Tuckle, MIT Press, 2009).  However, a distinguishing feature is that in GSS these models are usually the only kind we have.  In physics, for example, computer models are usually secondary to mathematical ones.  It is the latter which embody the theories and serve as object of discussion between scientists: computer code rarely makes its way in scientific publications.  The computer model is judged to be correct if it faithfully implements the mathematical model.  While it might occasionally be difficult to formulate exactly what counts as ``faithful'', we have, at least in principle, a ``golden standard'' for correctness.

Consider a multi-agent model which is supposed to inform politicians about the effect of green building subsidies on the evolution of unemployment.  What is the ``golden standard'' against which to judge the correctness of such a model?  Usually, such models are accompanied by a narrative, an informal description of the ideas behind their development, but a narrative is too blunt an instrument to help us decide whether a computer implementation of it is correct or not, whether the results of a simulation are trustworthy or flawed by programming errors.  In fact, in absence of an external criterion for the correctness of the model, it is not clear that we can talk about programming errors at all.

Developing the traditional kind of mathematical models for the complex heterogeneous socio-ecological systems involved in GSS has not been very successful so far.  GSS is massively inter-disciplinary, and even when the various components of a model have mathematical representations within the disciplines involved, there is no clear way of coming up with a mathematical model of the interactions between the components.  The almost automatic way that scientists solve this problem is by implementing the components and their interactions in code.  This provides a formal description of the model and allows exploring its consequences by simulations, and *only* by simulations, because the description is too low level to allow us to reason about it the way we do with traditional mathematical models.

This unsatisfactory state of affairs can be summed up as follows: either we limit ourselves to informal narratives, or we use simulations of computer models which we do not understand and whose correctness we cannot guarantee.  Either way, the scientific status of GSS is in question.

The solution appears to be the creation of an intermediate, mathematical layer between narratives and simulations, similar to that which exists in other established sciences.  This mathematical layer cannot just be, say, the theory of partial differential equations underlying the physics of climate, or the functional analysis accounting for the general equilibrium models of economics.  The formal language of GSS is computer code, therefore the mathematical layer has to be part of the mathematics of general programs, that is, computer science.

To put it in a somewhat pointed fashion: computer science should play for GSS the same role that mathematics plays for physics.

More concretely, we need to start by writing *specifications* for the kind of models used in GSS, which will involve choosing, adapting, and extending one or more of the formal languages for specifying and reasoning about programs.  At the moment, the main candidate for such a formal language is that of constructive type theory, due to its ability to express both (functional) programs and classical mathematical results.  But, just as physics has influenced the development of mathematics, providing fruitful problems and intuitions, we expect that GSS will also influence the kind of formal languages and the results of computer science.

Specifications will allow checking the correctness of implementations, but in the long run, we can do better: we can implement high-level domain specific languages, such that the distance between specification and implementation will be as short as possible.  Ideally, the specification should be expressive enough that its compatibility with the narrative which motivated it can be seen ``by inspection'' and that it can serve for the communication of scientific ideas, and at the same time it should be part of the programming language used to implement it, so that the implementation is correct by construction.

The foundational role of computer science in GSS does not exhaust by any means the role of ICT in GSS.  For example, we have not touched so far on the empirical side of GSS.  In common with almost every other scientific discipline, in GSS we also have the problem of the massive amounts of data to be collected and analyzed, but here we also face the *heterogeneity* of the data.  As an integrative science, GSS must develop the concepts and techniques for dealing with data coming from a wide spectrum of scientific disciplines.  Efficiently dealing with this kind of data must lead to establishing correlations between disciplines, developing indicators to summarize otherwise unmanageable quantities of information, suggesting new concepts for the theory of GSS.  These concepts should also allow us to deal better with the new sources of data available from the new kinds of social networking tools, such as Facebook, LinkedIn, Twitter.

These social networking tools have been developed in an ad-hoc fashion, raising questions about their security, reliability, use and misuse.  We should expect a science of global systems to enable us to understand these issues better, and design a next generation of communication tools.  Also in view of e-Governance issues, we need scientific theories that can tell us about how to increase the reliability of information, how to counteract disinformation, how governments can encourage a broad democratic participation, enabling change and building trust.



22.08.2012 16:17

1 reply

Patrik Jansson


Re: The role of ICT in Global System Science

     It sounds like you are looking for a physical science of systems, to parallel the information modeling approaches we now have.  I've been working on one and making reasonable progress for a while.  It extends from the conservation laws, and the requirement that energy using systems have continuities of organizational development to produce continuities of energy use when beginning and ending uses of energy.   I call that "the continuity principle", and use it to investigate the ubiquitous pairs of 'S' curves (¸¸.·´ ¯ Â¯ `·.¸¸) found in our data. 

As a science it is less about creating better artificial worlds in a computer than learning to use computers to more skillfully observe how the systems of our environment work and are behaving,  Success in that  would clearly also improve modeling abilities too, especially by grounding them more firmly on verifiable behavioral certainties.

Part of the end result is a recognition that the environmental systems which animate change in an environment operate as cells of organization, that like any growth system, operate by dramatically expanding their control over resources from their environments.   That portrays nature as something of a "bomb thrower", in germinating seeds of viral processes right and left.   It also lets you identify them as units of organization, to track and study.  Having definable boundaries allows one to study their organic thermodynamics too, and discover their 'S' curve developmental boundary conditions implied by "the continuity principle" for developmental change.  

Most of my work of recent years has been on short subjects and a couple long papers.

Informal collection of intro's to open systems physics

Software for isolating continuities in noisy data



< back to main discussion