260 likes | 434 Views
Emergence and self-organisation: Informal definitions. Emergence ….
E N D
Emergence … Although each effect is the resultant of its components, we cannot always trace the steps of the process, … , we propose to call the effect an emergent... instead of adding measurable motion to measurable motion, or things of one kind to other … of their kind, there is a cooperation of things of unlike kinds ... The emergentis unlike its components … these are incommensurable, and it cannot be reduced to their sum ... Lewis 1875,first use of the term “emergence” Fire, life, magnetism, heat … were all once thought … to be due to their own dedicated substances — phlogiston, vital fluid, magnetic fluid, caloric, and so on — but are now understood as emergent phenomenaof … natural processes. Bickhard 2002
Recap: some introductory ideas • Emergence • Behaviour observed at one scale is not apparent at other scales • Self-organisation • Structures that emerge without systematic external stimuli • Explore these informally … • Key issue: is emergence a natural phenomenonor an artefact of observation? • Can we answer this?
Emergence is notsurprise • Some early work defined emergence as surprise • The surprising effects that emerge when a lot of agents come together • when the football crowd does a Mexican wave • that single-cell amoebae can operate as a multi-cell organism • that quantum physics gives rise to Newtonian laws • OK – I wassurprised the first time • Surprise is too rooted in personal experience • If my only experience of a crowd produced a Mexican wave, my crowd definition includes pulsating surface behaviours
Who studies emergence? • Philosophers of mind • how the mind emerges in the physical brain • how intelligence emerges from unintelligent matter • Biologists (philosophers of biology) • how life emerges from inanimate matter • Computer scientists (ALife community) • how properties analogous to mind or life might emerge on non-biological substrates (computers) • Amongst others …
Defining Emergence • They agree on just two things • We need a consistent definition of emergence • We don’t have one • “The whole is other than the sum of its parts” • Metaphysics (Aristotle, Ancient Greece) • Phenomenology (Jung, Hegel, late 19th century) • Applied in solid state physics (Anderson, 1970) • Recognition that parts of science are resistant to understanding through reductionism
What’s wrong with reductionism? • The basics are there: • Quantum phenomena give rise to physics • Physical phenomena give rise to chemistry • Chemical phenomena give rise to biology, geology, etc. • Biological phenomena give rise to society • At which point, humans observe, and see patterns • Abstraction allows some prediction and replication • But only up to a point • Cannot model with sufficient precision • Heisenberg uncertainty, the mathematical limit on what can be known about a physical system • Non-determinism (e.g. in quantum physics) • Impossibility of capturing the precise start state
Reductionism ignores dynamics • Consider some examples: • Growth • phenotype emerges from structure and dynamics of growth rules • Intelligence • emerges from structure and dynamics in the nervous system • Sociology • emerges from structure and dynamics of social organisms • Keep checking as we look at new examples • Can you explain the emergent behaviour by reduction?
Reductionism versus phenomenology • Reductionism dominated science to 19th century • Despite Aristotle’s ideas and legacy • Non-human animals could be reductively explained as automata — (Descartes: De homine, 1662) • Matter from fundamental particles (Dalton, c1803) http://www.anyalarkin.com/alblog/wp-content/uploads/2012/04/Anya-automaton1.jpg • Observation and theory challenged reductionism • e.g., many new fundamental particles • Phenomenology in science • Empirical observations are related in ways consistent with fundamental theory but not directly derived from it • Monte Carlo modelling, PDEs, etc. • Used in biology, particle physics, etc. • http://www.edc.ncl.ac.uk/assets/graphics/montecarlo.jpg
Phenomenology and emergence • Phenomenology in science focuses on modelling to mirror observed behaviour • Guess key components • A surrogate for full understanding of observed behaviour • Cannot say what a model means in terms of natural phenomena • Estimate some rates, feed into equations, guess what it means • Some support for prediction • Often later verified by observation • Like Aristotle, our sense of emergenceis more fundamental • System properties and behaviours are an inherent property of collections of components over time and space
How can anything new emerge? Importance of process • Reductionism is founded on a “metaphysics of substance” • Static particles that just divide or combine • Process is vital to emergence (and scientific understanding) • Temporal and physical context and scale are vital • It is point-particles or entities that are artificial • persistent instances of organisations Bickhard, in Downward Causation, 2000 e.g., A vortex does not exist without flow http://earthobservatory.nasa.gov/Newsroom/NewImages/Images/Australia_AMO_2006156.jpg
Levels in emergence • Emergent properties are irreducible • No reductionist explanation • Systems theory • e.g., Checkland, 1981 • System level languageis meaningless at component level • Cannot derive system descriptionfrom component description • Each level has its own structure and dynamics • Longer time scales reveal relatively stable high-level patterns • Larger scale reveals patterns with extent and movement • Such as vortices Ryan: http://arxiv.org/PS_cache/nlin/pdf/0609/0609011.pdf
An observation on time-bands Burns et al, 2005: http://www.cs.york.ac.uk/ ftpdir /reports/YCS-2005-390.pdf • Time scale of emergence • Within the context of any particular band: • Activities within lower (faster) bands are instantaneous • Activities within higher (slower) bands are static curriculum Lecture New slide
Resolution and scope Emergent properties are simply a difference between global and local structure. • Instead of layers / levels, consider Resolution … • Characteristic of representation of system • Different properties apparent at different scales • … and Scope • How / where the system boundary is drawn • New properties arise if system encompasses many components • Time defines dynamics Ryan, 2006 http://arxiv.org/PS_cache/nlin/pdf/0609/0609011.pdf
Levels, Resolution and Scope • Resolution and scope are useful concepts • Macro-state is either wider (scope) or coarser (resolution) than the component state • Levels are also useful • Clear discontinuity in descriptions of system and components • “Macro-state” implies scale difference • Level, scope and resolution are just views • Observing properties or behaviours at a coarse resolution • Observing more of the system (wider scope) • This is an open academic discussion • We’ll return to it after entropy!
Types of emergence • Much discussion of types of emergence • Weak, strong, intrinsic, extrinsic • Often not very useful • e.g., intrinsic emergence (Crutchfield): • No external observation needed • “the system itself capitalises on patterns that appear” • e.g., strong emergence (Bedau) • Allied with downward causation • Weaker forms admit those who don’t like downward causation see, e.g., Stepney et al, ICECCS 2006
Causality among levels, scopes, resolutions • It is ‘obvious’ that coarser, higher level (… ) patterns are caused by finer, lower level (… ) dynamics • Upwardcausation • Downward causality is more controversial • At some temporal or spatial scale, global patterns affect local behaviours • Context is vital for emergence • To some researchers, downward causation is intrinsic • To other researchers it is too inexplicable for credibility • But some can’t cope with the ideas of emergence and complexity … • … full stop! Stepney et al, ICECCS 2006
Self-organisation “Is the Mexican wave really a ripple of excitation?” Cartwright, http://www.europhysicsnews.com/full/41/article3.pdf
Examples of self-organisation • Social activities • Construction by social insects, flocking, crowd dynamics • Dissipative structure • e.g., a thermodynamically-open system operating far from equilibrium in an environment with which it exchanges energy • e.g., BZ reaction, hurricanes, turbulence, convection • Some CAs and evolutionary computations • eg cyclic CAs, swarms • Some authors include phase transitions, turbulence, ecosystems, adaptation, natural design principles … Shalizi: http://www.cscs.umich.edu/~crshalizi/notebooks/self-organization
Self-Organisation • Idea probably from Descartes (1637) • Before that, order arises by chance, given time and space, • “Self-organisation” coined by Ashby (1947) • Ashby considers organisation to beinvariant • Organisation fderives from the functional dependence of a current state Sc on a past state Sp and some inputs I f:Sp×ISc W. Ross Ashby, 1949, 1962: reproduced at csis.pace.edu/~marchese/CS396x/Computing/Ashby.pdf
Self-Organisation • Ashby states that self-organisation is apparent if: • in two regions of state space, f is approximated by organisations g and h g:Sg×IgSg h:Sh×IhSh • system dynamics drive the system from g to h • Self-organisation is observed locally in a system with globally-invariant organisation • Self-organisation is thus an emergent property due to the scope of consideration of the larger system • Much subsequent work ignores Ashby • Preference for vague, informal definitions
Variants on self-organisation • Claims and counterclaims on whether systems self-organise • A few contribute to understanding: • Hypercycles to explain co-operation among competing individuals • Winfree’s study of rhythms and oscillation in biological systems • Computer science use in unsupervised learning Shalizi, 2001, http://cse.ucdavis.edu/~cmg/compmech/pubs/CRS-thesis.pdf
Self-organisation and emergence • Self-organising systems display emergent properties • Patterns at a higher level, coarser resolution or wider scope • Dicty amoebae self-organise [later] • an emergent slug and emergent fruiting behaviour • Social insects self-organise • to achieve construction, effective navigation, foraging • Crowd behaviours in higher species If this does not convince you that dynamics are essential… http://www.news.com.au/common/ imagedata/0,,5381235,00.jpg
A working definition of emergence • A system with levels • Scales, resolutions … • At each level, granularity of space and time are different • Levels have different languages • The concepts needed to describe each level are distinct • System is neither random, nor in a steady state • Constant flows of energy, matter… • Dynamics essential to emergence • At lower level, components may tend to self-organise • Try looking at the later examples in these terms…
Putting it all together Prokopenko et al, An information theoretic primer, 2007