Does anything Emerge CSS Interaction Task

1
Does anything Emerge?CSS Interaction Task

• A review on the state of the art in the
  understanding, modelling and formal description
  of emergence

2
Why

• Aims
• to define the broad directions which a research
  program on Emergence should follow in order to
  deepen our theoretical understanding
• to bring Emergence concepts and tools into the
  arsenal of physical, natural and social
  scientists

3
What has been done

• Call for material/ideas/collaborations via
  CSS_Ideas mailing list
• Considerable reading
• Production of a working draft
• Preliminary session at the 2004 CSS workshop
• Submission of a short paper to the KES conference
  (Knowledge-Based Intelligent Information
  Engineering Systems, Melbourne Sept 2005)
• Main workshop, May 05
• Manuscript Preparation

4
Actual outcomes

• A better understanding of the concept
• I think so.. listen on..
• Document/Web site open to CSIRO and others,
  http//www.per.marine.csiro.au/staff/Fabio.Boschet
  ti/CSS_emergence.htm
• short paper to the KES conference
  http//www.per.marine.csiro.au/staff/Fabio.Boschet
  ti/3054CO/papers/emergence_kes_final.pdf
• Review Paper
• Research Program Paper under development

5
Interaction Outcomes

• crucial input from Cosma Shalizi and Daniel
  Polani
• close interaction between 4 members (FB, Mikhail
  Prokopenko, IM, AMG) in draft preparation
• close interaction between 16 members at main
  workshop
• Strongly multidisciplinary approach (biologists,
  chemists, computer scientists, applied maths,
  physicists, social scientists, and engineers)
• Exposure to the The Pacific Institute of
  Theoretical Physics (PITP) workshop on
  Emergence, 15-18 May 2005, Vancouver, BC
  http//pitp.physics.ubc.ca/archives/CWSS/emergence
  /talks.html

6
Review / Research Program Paper

• Approach
• Information Theoretic
• Cosma Shalizis The efficiency of prediction
• Difference between emergence, complexity and self
  organization
• How Cosmas definition could be implemented and
  tested on a set of real problems

7
Review / Research Program Paper

• Cosma Shalizis The efficiency of prediction
• The efficiency of prediction of a process is the
  ratio between its excess entropy and its
  statistical complexity, e E/C.
• EExcess Entropy
• the mutual information between two adjacent
  semi-infinite blocks of variables -gt
• measure of the apparent memory or structure in
  the system -gt
• we may think of it as the fraction of historical
  memory stored in the process which does useful
  work" in the form of telling us about the future.

8
Review / Research Program Paper

• Cosma Shalizis The efficiency of prediction
• e E/C ---- How to calculate E

Excess Entropy
Entropy Density
9
Review / Research Program Paper

• Cosma Shalizis The efficiency of prediction
• The efficiency of prediction of a process is the
  ratio between its excess entropy and its
  statistical complexity, e E/C.
• C Statistical Complexity
• Entropy of the process Casual States -gt
• We may think of it as a measure of the amount of
  memory required for optimal prediction.

10
Review / Research Program Paper
Cosma Shalizis Ease of predictability e E/C
---- How to calculate C

• Generate a causal state machine (CSSR algorithm,
  for example)
• Calculate entropy of the e-machine

11
Review / Research Program Paper
Cosma Shalizis The efficiency of prediction
12
Review / Research Program Paper
Cosma Shalizis The efficiency of prediction
Given 2 levels of representation, we say one
level emerges from the other if it has greater
predictive efficiency.
13
Review / Research Program Paper
Cosma Shalizis The efficiency of prediction

• What is good about it?
• It is non trivial
• It removes magic from the concept (cannot be
  explained, is not displayed, etc)
• it is computable (albeit in limited cases)
• it distinguishes Emergence from Complexity and
  Self Organisation

14
Review / Research Program Paper

• Complexity amount of information necessary for
  optimal prediction
• Self-organisation spontaneous increase of
  complexity in time
• Emergence E/C
• AdaptivityIncrease of Mutual Information between
  a system and the environment

15
Review / Research Program Paper
Cosma Shalizis The efficiency of prediction

• What might be missing?
• Not always computable
• Emergence/Complexity/Self-Organisation are
  different, but what is the relation between them?
• How do we account creation of novelty and
  diversity? (surprise effect)
• How does Emergence arise?
• Are there different classes of Emergence?

16
Review / Research Program Paper
Paper Organisation

• Motivation
• Justification for an information theoretical
  approach
• Ordering of emergence phenomena
• Drivers and conditions for Emergence
• Detection of Emergence
• Modelling of Emergence
• Research agenda
• how to construct emergence
• Designing desired emergence properties
• Mechanism for emergence
• Boundaries
• How we use it

17
What next?

• Finish the paper and submit
• Possible relation to UD, Memes and generation of
  follow up IT? (David Batten?)
• Results, tools, insights feeding into CSS
  projects
• Network Theory IT
• Model Structure
• Delineation of a research program
• Detection
• Design for emergence
• The role of system boundaries
• Mechanism for Emergence
• How to use it

18
Some Conclusions

• Information Theoretic approach
• Emergence as relation between levels of
  representations
• Emergence as increase efficiency of prediction
• Definition of a research program on the topic

19
Some Conclusions

• Did we interact? Yes
• Did we learn something? Yes
• Did we contribute something useful?
• We took a unpublished definition
• We challenged it
• We planned how to test it on real cases
• How to extend it to different disciplines

20
All material currently available at

• http//www.per.marine.csiro.au/staff/Fabio.Boschet
  ti/CSS_emergence.htm

21
(No Transcript)