New Theories and Methodologies

On this page, ONCE-CS coordinators propose a synthesis of the discussions in the CS community concerning Theories and Methodologies in Complex Systems. For the first synthesis, ONCE-CS coordinators asked Mark Buchanan to do a summary of the discussions of the first brainstorming meeting after the reports of the several working groups. These reports will soon be available in the corresponding working group of the brainstorming meeting.

This synthesis has been made after the reports of the several sessions of the brainstorming meeting, held in Paris, september 2005

• New Theories and Methodologies
• Key Methods
  • Organization
  • 1. What are the key methods of complex systems science?
  • 2. What kinds of new methods do we need?
    • Analysis and visualization of “difficultâ€? data
    • Modeling
  • 3. Important general points/questions on methods.
• Killer Applications
  • 1.General categories:
  • 1. Solving Social or Technological Problems
    • A. In General Social Science
    • B. ICT-related Social systems
  • 2. Solving “difficultâ€? problems
  • 3. Changing minds about Complexity Science
    • A. General
    • B. Education
  • 4. Plus many others!

Key Methods

Organization

1. What are the key methods of complex systems science? 2. What kinds of new methods do we need? 3. Important questions and issues on methods.

1. What are the key methods of complex systems science?

We can list a number of key methods (or general method areas) currently in use:

• Modeling (physical models as well as computational models)
• Multi-agent systems and distributed systems in general
• Simulations (agent-based and otherwise)
• Category theory
• Game theory
• Markov analysis
• Bio-inspired methods (evolutionary computation, etc.)
• Graph and network theory
• Linear and non-linear system identification (time-series analysis)
• Bifurcation theory/dynamical systems theory
• Nucleation theory
• Market research, neuro-economics (economics)
• Focus groups (social science)
• Interviews (psychology)
• Nano-technology (materials science)

Also, don’t forget:

• Doing experiments. Good method!

2. What kinds of new methods do we need?

Needs in several areas:

Analysis and visualization of “difficult� data

• Methods for analyzing and visualizing large data sets
• New mathematical theories for understanding large (and possibly noisy) data sets
• Methods for analyzing qualitative data.
• Methods adapted for short time-series or data sets with missing data

Modeling

• Formal models that lend insight into the micro-to-macro transition; that might even make predictions (of some sort) for specific problems
• Theories for phase transitions in non-equilibrium systems
• Accurate dynamic models for the temporal evolution of complex networks
• Methods for simulating multi-agent systems with open phase space

Simulation

• Efficient parallel architectures and distributed algorithms for simulation

Experiments

• Methods for testing/confirming/falsifying CS theories in (especially) social-economic sciences
• Methods for collecting, organizing and making public data for social-economic experiments

3. Important general points/questions on methods.

1. Are there general ways map the features of a proposed model to the relevant features of some phenomenon? 2. It is crucial to finding the right level of abstraction in building a model (not only in CS, but generally). Are there general methods for doing this? 3. How does one choose between different techniques (e.g. discrete computation versus dynamical modeling with differential equations. When are such methods interchangeable or complementary? 4. Why do we focus on computational methods? It is important to remember that physical models are also useful. 5. Computational models have inherent limits (theoretically, as there are “non-computational� processes, and practically, by hardware/software limts). 6. How can we validate/invalidate simulations? What are the criteria? Comparison with experiment? Are there more formal processes?

Killer Applications

1.General categories:

1. Solving Social or Technological Problems 2. Solving especially “difficult� problems 3. Changing minds about Complexity Science 4. Some other stuff

In more detail:

1. Solving Social or Technological Problems

A. In General Social Science

• Tools for enabling intelligent “CS managementâ€? of organizations – avoiding "surface-interpretations" of problems, for example. Applications in business and government
• Tools to support collaboration and help deal with an “increasingly complexâ€? environment (if it really is).
• Tools for distributed decision-making (for example, decisions on adopting a new technology, such as GM crops, at European level)
• Curing cancer, etc. with artificial forms of life (maybe achieve quantum computing while your at it, while avoiding accidental destruction of planet).
• CS tools for enabling local/global sustainable development
• CS tools to help people deal with disasters. Participatory simulations help people learn by collaboration for their own and their communities’ benefit.
• Tools for gathering/monitoring that would identify when patterns – either undesirable or unexpected – arise. Would allow intervention to avoid undesirable events (development of cancer, outbreaks of epidemics, episodes of “creative accountingâ€? (Enron)), or to take advantage of unusual opportunities.
• Tools for aiding human negotiations

B. ICT-related Social systems

• tools for winning the fight against cyber-dysfunction and criminality (congestion-free traffic, wiping out viruses, SPAM, etc.
• networks of wireless sensors, collective robots

2. Solving “difficult� problems

• develop new programming languages/paradigms
• mobile machine language translation that works
• modelling brain and mind
• Tools (chaos theory, etc.) for dealing with limited data

3. Changing minds about Complexity Science

A. General

• Tools that demonstrate the need to use a complexity approach in different domains
• Systems to support the transition from centralised mindset to a more natural ability to think about and engineer useful and (possibly) crucial Complex Systems (including educational applications)

B. Education

Intelligent simulation tools with tutoring component built into them; application that would build a bridge towards new ways of teaching science that include Complex Systems methods

4. Plus many others!

• Context-aware software
• Self-adapting PC: a PC adapting to the user. It learns (from) the user and adapts to him (what to store, what to use, what to delete etc)
• Create ‘individualised’ drugs.
• Weather control (intelligence and actuation)
• Collective brainstorming
• Reducing human hierarchy/bureaucracy

---