# Emergence ### and exciting networked products
## What is emergence? [T]he arising of novel and coherent structures, patterns and properties during the process of self-organisation in complex systems. [Goldstein](#49/0) Locally acting rules can combine to create unexpected large-scale emergent behaviour. Emergence is everywhere.
# Examples Culled from Wikipedia, Google Maps, Facebook…
# Shiny!
## Why emergence is cool Emergent biological systems have lots of desirable properties - Fault tolerance - Robustness - Adaptability
## Flocking [Reynolds / RED3D](http://www.red3d.com/cwr/boids/)
## Rules for flocking Alignment – go at same velocity as neighbours Cohesion – move towards other Boids Separation – but don’t get too close
## Theories of emergence The “how many angels can dance on the head of a pin?” bit
## Are emergent phenomena real? Can emergent models be causal, or are they all epiphenomena? **Epiphenomenon** – [a] phenomenon that can be described independently of the underlying phenomena that bring it about. [Abbot](#49/0) **Physicalism** – science[_n_ + 1] is just applied science[_n_] [Anderson](#49/0) **Vitalism** – there is a ‘vital spark’ or ‘élan vital’, which some equate with the ‘soul’ (dualism)
## Physicalism & causal drain Causal drain – epiphenomena all the way down? Quantum theory – where have the particles gone???
## Causation and supervenience **Downward causation** – genuine emergent phenomena, can produce behaviour at lower levels **Supervenience** – One level supervenes on another when there can only be a change at the higher level if there is also a change at the lower level (not necessarily causal) Seems like bollocks to me **Downward entailing** – properties of high levels can tell us something about low levels [Abbot](#49/0)
## Complexity and timescales Longer timescales? Timebands – high level band constant for low level band, low instantaneous for high [Burns](#49/0) Complexity – [a]t each level of complexity entirely new properties appear [Anderson](#49/0) Emergent simplicity – planet orbiting a star [Bar-Yam](#49/0) ([Kepler](http://en.wikipedia.org/wiki/Kepler%27s_laws_of_planetary_motion)), climate vs weather? Shalizi – emergent process as one that has a greater predictive efficiency than the process it derives from. [Shalizi](#49/0) (remember this :)
## Changing scope or resolution Levels seem rather vitalist; ontological or epistemological? Resolution – Rh ≤ Rl Scope – Sh ≥ Sl
(Rh, Sh) ≠ (Rl, Sl)
See [Ryan](#49/0); will continue to use level even though define emergence in terms of scope and resolution
## Weak emergence
Sh = Sl ⇒ Rh < Rl
Examples: temperature, patterns on animal skins, traffic jams (probably) [Bar-Yam](#49/0) Epistemic? – Once we understand the mapping it’s no longer emergent [Ryan](#49/0), [Goldstein](#49/0). Confuses mapping and path ([see later](#24)) Also limitations on predictability due to nonlinearity of complex systems, e.g. strange attractors, [Life](http://en.wikipedia.org/wiki/Conway%27s_Game_of_Life) [Goldsein](#49/0)
## Strong emergence
Sh > Sl
Only exists at the emergent level, cannot be identified without looking at all parts of system together Example – secret scheme: scope must cover all pieces to know secret [Ryan](#49/0)
## Dynamics and emergence Dynamics is one of the common properties that identify [systems] as emergent. [Goldstein](#49/0) Emergent characteristics: nonlinearity, self-organisation, non-equilibrium systems and the presence attractors Crutchfield’s ε-machines [Crutchfield](#49/0) Shalizi’s ε-machines [Shalizi](#49/0)
# My view of emergence Fits within weak emergence Subjective emergence Mappings Quantified emergence Neutral emergence
## Subjective emergence It’s all relative – there’s no right emergent phenomenon Depends on the mapping and language chosen Different view → different emergent properties Some phenomena are less useful than others
## Independence and lossy emergence Emergent properties appear to add something new and distinct But they’re merely a (carefully chosen) subset of the underlying behaviour Emergence doesn’t add anything; it removes things, leaving a coherent core of behaviour – the high level phenomenon – that appears emergent
## Discontinuities and lossy emergence A discontinuity must exist between high and low levels of emergent system Responsible for much of power – can use emergent model to predict future behaviour Simpler – info lost between levels
## Mappings in emergence Mappings matter – without a good mapping impossible to use emergent model to predict system behaviour
## Quantitative emergence Possible to put a number on how good an emergent model is (at least sometimes) Information theoretic model using Kolmogorov complexity
I(S : L) = H(S) - H(S|L) ≡ H(L) - H(L|S)
Now have an automatically developed emergent system See also [Shalizi](#49/0)
## Gaining MI over time
## Neutral emergence Cf. neutral evolution – indirection, evolvability Discontinuity in emergent systems → incomplete mapping → some behaviours indistinguishable An emergent property exhibits neutral emergence when a change in the microstate L does not change the macrostate S, or vice versa. [Me](#49/0)
## Coarse graining as emergence I used elementary cellular automata Rule 128
Rule 102
Rule 110
## Coarse graining
## Subjective emergence Coarse graining 140 to 136
## Subjective emergence Coarse graining 140 to 204
## There is nothing magic about emergence, or developing emergent systems
# Emergence is easy

…but developing useful emergent systems is considerably more difficult

## Engineering emergence Traditional engineering can’t really model, tries to eliminate emergence Emergent engineering manages novel behaviour – include unexpected event phase Essentially how [GAs](http://en.wikipedia.org/wiki/Genetic_algorithm) work – and only need to know if answer is correct, avoiding discontinuity problem
## Robustness and neutral emergence The best solutions are right and robust – not brittle Allow system to explore its environment before ‘committing’ to a model Stress the model during development
## Evolving flocking Discontinuity between levels → developing incrementally not possible Fish evolved dispersal and aggregation, but schooling behaviour was degenerate [Zaera](#49/0) Fitness function proved inadequate: rewarded correct schooling, but also favoured many other behaviours Schooling emergent → impossible to provide indicator of progress towards schooling in terms of lower level behaviour
## Evolving flocking indirectly Reasons for schooling: reducing risk being eaten, easier to find food, mating efficiency, learning environment, reducing drag [Werner](#49/0), [Partridge](#49/0) Add predators to model → schooling may emerge as selectively advantageous Flocking has evolved in response to predation at least twice: Reynolds against simple predator, Ward against more complex co-evolved fish with nnets [Reynolds](#49/0), [Ward](#49/0)
## More complex, yet simpler to do Predator-prey models definitely more complex models But objective is very simple: don’t get eaten And this is where the fitness function operates That’s much easier to understand and specify, even if the solutions are more complex Also more robust: flocking is an effective response to many predator models, so system doesn’t have to be exactly ‘right’
## Getting emergence by solving a different problem

One that’s useful every step of the way

## Mendeley Mendeley is a free reference manager and academic social network that can help you organize your research, collaborate with others online, and discover the latest research. Useful even if you–re the only person using it – store papers, make notes, generate bibliographies More useful if group uses it – share papers, discuss ideas Even more useful if everyone uses it – top papers in field, connect other researchers [Mendeley’s] real power lies in what it does with the collective data from users BBC
## Epicurely Connecting people through food. Epicurely is a platform for organizing and discovering local culinary experiences Social dining, AirBnb for food (kinda) Predicting food trends, location trends Dinner parties → popup restaurants → traditional restaurants? Islington → Soho → Wandsworth?
## Songkick Predicting band popularity, who will like a band Look at who likes band over time – movement of likes (also engagement?) Compare to band’s (un)stated influences Reverse → predict who will like what Genre, venue, location popularity?
## Hang on… A lot of this sounds like data analysis Using GAs works, but many EAs seem to have a complexity ceiling Trend → prediction → rule Think there’s a lot more to dig out here – these ideas seem rather limited [Forest fire models](http://en.wikipedia.org/wiki/Forest-fire_model), [Duncan Watts](http://research.microsoft.com/en-us/people/duncan/) on social influence – cascades _MI : extra entropy_ – map to general model for _signal : noise_? [Me](#49/0)
## This is just a start There’s much more here


## More There’s more about most of this in [my thesis](http://meloncholy.com/thesis) if you’re interested Or [send me a message](http://meloncholy.com/contact) – be lovely to hear from you
## References - _Abbot_ – R. Abbott, “Emergence explained: getting epiphenomena to do real work.” Complexity 12, pp. 13-26, 2006 - _Anderson_ – P. W. Anderson, “More is different.” Science 177, pp. 393-396, 1972 - _Bar-Yam_ – Y. Bar-Yam, Dynamics of Complex Systems. Westview Press, 1997 - _Burns_ – A. Burns, I. J. Hayes, G. Baxter, C. J. Fidge, “Modelling temporal behaviour in complex socio-technical systems.” Technical Report YCS-2005-390, Department of Computer Science, University of York, 2005 - _Crutchfield_ – J. P. Crutchfield, “The Calculi of Emergence: Computation, Dynamics, and Induction.” Physica D 75, pp. 11-54, 1994 - _Goldstein_ – J. Goldstein, “Emergence as a construct: History and issues.” Emergence, pp. 49-72, 1999
## References - _Partridge_ – B. L. Partridge, T. J. Pitcher, ‘Evidence against a hydrodynamic function of fish schools.’ Nature 279, pp. 418-419, 1979 - _Reynolds_ – C. Reynolds, ‘An Evolved, Vision-Based Behavioral Model of Coordinated Group Motion.’ From Animals to Animats 2: Proceedings of the Second International Conference on Simulation of Adaptive Behavior, MIT Press, pp. 384-392 - _Ryan_ – A. J. Ryan, “Emergence is coupled to scope, not level.” Complexity 13, pp. 67-77, 2007 - _Shalizi_ – C. R. Shalizi, “Causal Architecture, Complexity and Self-Organization in Time Series and Cellular Automata,” PhD thesis, University of Wisconsin at Madison, 2001 - _Ward_ – C. R. Ward, F. Gobet, G. Kendall, ‘Evolving collective behavior in an artificial ecology.’ Artificial Life 7, pp. 191-209, 2001
## References - _Werner_ – G. M. Werner, M. G. Dyer, ‘Evolution of herding behavior in artificial animals.’ From animals to animats 2: Proceedings of the Second International Conference on Simulation of Adaptive Behavior, MIT Press, pp. 393-399, 1992 - _Zaera_ – N. Zaera, D. Cliff, J. Bruten, ‘(Not) Evolving Collective Behaviours in Synthetic Fish.’ Animals to Animats 4: Proceedings of the Fourth Internatinal Conference on Simulation of Adaptive Behavior, MIT Press, pp. 635-644, 1996 - _Me_ – A. Weeks, “Neutral Emergence and Coarse Graining Cellular Automata,” PhD thesis, University of York, 2010