2. Raphael Sanzio “The school of Athens”, 1510-11, Apostolic palace, The Vatican
3. 1694 1980-90s
Bank of Wynne, Ravetz
1690 England Post-normal science
Locke
Darwin
The argument from Design 1859 Complexity
Reducible uncertainty Indeterminacy
1543
Copernicus Energy
SDI Resources Modern-
1637 ism
Descartes Enlight-
enment
1430-50
Rationalist Science
Papal authority
Medieval Renaissance
1100 1300 1500 1900 2100
1417 1632 1687
Lucretius Galileo Newton
rediscovered 1770-80
Boulton and Watt
4. The precepts of natural law
• The succession of events or phenomena that
we perceive is not entirely arbitrary or
whimsical: there are causal relations manifest
in the world
• The relations posited above are, at least in
part, capable of being perceived and grasped
by the cognitive self
• .....Otherwise no science, no development.......
5. A THEORY OF MODELS: Casti (1992)
decoding Syntax
Semantics
I
C N
A F
U E
R
S N F E
A
L N
C
E
observables theory
encoding
When we do this we make choices and abstractions
Rosen, 1991
Newton made a choice!
6. The Groucho Marx paradox
• To quote Groucho: "I wouldn't want to belong to any
club that would accept me as a member“
• John D Barrow used this to say: "A universe simple
enough to be understood is too simple to produce a
mind capable of understanding it.“
• i.e. the closer we get to a description of reality, the
more complex and incomprehensible the description
becomes
7. THE FIRST PROBLEM
Engineering
Physics Known Unknown
Knowns Unknowns
Directives complexity
Scientific Adaptive Robust Precautionary
management management Decision Principle
Making
Frequentist Bayesian
statistics statistics Causal thickets
Ecology “Black Swans”
Climate
Complete Rivers Complete
certainty Uncertainty
INFRASTRUCTURE
8. Engineering
Physics Known Unknown
Knowns Unknowns
Directives complexity
Scientific Adaptive Robust Precautionary
management management Decision Principle
Making
Frequentist Bayesian
statistics statistics Causal thickets
Ecology “Black Swans”
Climate
Complete Rivers Complete
certainty Uncertainty
THE FRAMING PROBLEM
9. Aristotelian causes
• Material: weapons that soldiers use in battle
• Efficient or Mechanical: soldiers who swing the
swords or pull the triggers
• Formal: battle tactics, the role of the generals
• Final: the ultimate context; socio-economic or
political causes of the state or government
• Science is only concerned with Material and
(maybe) Efficient or Mechanical causes
10. Newtonian physics, “Senior science”
• Universal, equilibrium • “Received” science
– Axiomatic, time reversible • Deductive (induction)
• Ergodic – Formal logic
– Homogeneous • Value free
– Statistical stationarity – About “is”
• Rationalist mathematics • Abstraction, externalities
• Realist, reductionist – Time invariance
• (Post-)positivist – “frozen fields”
• Materialist (Newton)
• Mechanism, “clockwork” • So development since the
– Atom driven by forces
Enlightenment is due to
picking soluble problems!
– Prediction possible
11. METAPHOR
decoding Syntax
Semantics
I
C N
A F
U E
R
S N F E
A
L N
C
E
observables theory
encoding
Rosen, 1991; Casti, 1992
12. FUNCTIONAL (IDEAL) MODELLING
Deduction and strong inference in the scientific method
Laws provide predictions, refutation is possible
The diagram commutes, inverse modelling is possible
SEMANTICS SYNTAX
abstraction
REAL T2 ABSTRACT
Time
prediction
computation
causality
LAWS
abstraction
REAL T1 ABSTRACT
“of things” ALGEBRA
Semantics works in “real” states, objective reality
Hauhs & Widemann, 2010
17. New problems
• What to do about spatially and temporally
extensive, heterogeneous, adaptive (evolving),
non-linear, complex, contingent, emergent
systems with people (life)?
• Infrastructure, companies, ecosystems??
• Non-stationary, systemic risks, network failures,
super-transients: cultures, beliefs, values (norms)
• No controls, no replication, inability to “control”
variances – “cause-effect” unclear
• But we still use “received” modernist science and
management – lack of critical thinking
20. EXACT SPECIAL
SCIENCES SCIENCES
Engineering
Physics
Correlations
Environmental Nomological Humanism
Mastery complexity Sociology
Power Scientific Bridge laws Psychology
Precautionary
Markets management History
Principle
Innovation Machines Environmentalism
Low risk Statistics ModelsCausal thickets
Ecology “Black Swans”
Climate
Complete Complete
Rivers
certainty Uncertainty
Enlightenment values Post-enlightenment values
Reason, rationalism Uncertain, complex
Reducible uncertainty Causal thickets, indeterminacy
Predict-act, risks known Cause-effect unclear
Positivism, realism Non-equilibrium, contingent
Reductivism
Materialism Bounded rationalism
Reification
Philosophy for limited beings
Physics envy in the social science literature (e.g. Ergodic economics)
21. Instrumental reasoning
• The explication and reduction of value
judgements, the “rational” pursuit of goals,
and the adaptation of models or systems to
the goal attainment – as far as environmental
or other constraints permit” (M: p. 317)
• Whatever works –
– “anything goes” Feyerabend (1975) “Against
Method” regarded as anarchic by science
22. Pragmatism (Peirce, Dewey)
• Theories are instruments for making revisable
empirical claims about reality
• Information = evidence (data) by which
expectations are changed with regard to specific
statements (Hypotheses) – (M: p.230)
• Knowledge = law-like statements sufficiently
supported by evidence acceptable as
(provisionally) true by a certain branch of science
• Inductive inference – genuine science not
possible if indeterminism holds (M: p. 234)
– But remember the precepts of natural law!!
23. THE SECOND PROBLEM
Complexity, emergence (1)
• “A process undergoes emergence if at some
time the architecture of information
processing has changed in such a way that a
distinct and more powerful level of intrinsic
computation has appeared that was not
present in earlier conditions”.
• Crutchfield (1994) The calculi of emergence,
Physica D.
24. Complexity, emergence (2)
• Knowledge of lower level agents and
interactions (pattern and process) is a
necessary, but not sufficient, condition for
prediction at higher level
• Reflexive, recursive interactions
– Non-linear, contingent, local, initial conditions
• “Reductionist nightmare” (Cohen & Stewart)
– Predict-act fails: SURPRISE!!
25. Meta-physics: the philosophical basis
of natural law
All things by immortal power
Emergence
Near or far
Fungibility
Hiddenly
Equifinality
To each other linked are
That thou canst not stir a flower
Without troubling of a star
Francis Thompson “The Mistress of Vision”
Complicity
Cohen & Stewart, 1994
26. Not “atoms” but components
• Components have reflexive relations ships
with other components
– Therefore there are both external and internal
(system) drivers
• Causes beyond (above) the Material (Horrors!)
– Purpose, meaning, intention??
• Non-ergodic and non-stationary
– Systemic risks – non-Normal statistics
28. AN APPROACH TO A SOLUTION
Systems methodology
• Complexity and emergence means
reductionist nightmare – therefore choices...
– Choices mean values, norms, semantics, purposes
• Abstractions, levels, processes
– Empiricism, weak inference, equifinality
• Bounded rationalism (not relativism)
– We know a lot but not everything (Pragmatism)
• Not “anything goes” – because there are
biophysical constraints, design limits
29. Systems methodology
• Mattessich (1978) argued that science –
particularly the applied science of systems – is
“structural-holistic, dynamic as well as
instrumental” because it “not only
emphasises the recorded insights of science
but also stresses the entire process of doing
science, as well as the holding and using of
theories, of elaborating and eventually
replacing them by better ones (his italics,
Mattessich, p. 250).
30. EXACT SPECIAL
SCIENCES SCIENCES
Engineering
Physics
Correlations
Environmental Nomological Humanism
Mastery complexity Sociology
Bridge laws
Power Scientific Psychology
Precautionary
Markets management History
Principle
Innovation Machines Environmentalism
Low risk Statistics ModelsCausal thickets
Ecology “Black Swans”
Climate
Complete Complete
Rivers
certainty Uncertainty
Enlightenment values Post-enlightenment values
Reason, rationalism Uncertain, complex
Reducible uncertainty Causal thickets, indeterminacy
Predict-act, risks known Cause-effect unclear
Positivism, realism Non-equilibrium, contingent
Reductivism
Materialism Bounded rationalism
Reification
Philosophy for limited beings
Physics envy in the social science literature (e.g. Ergodic economics)
31. INTERACTIVE MODELLING
This delves into computational theory
interface SEMANTICS SYNTAX
“REAL” assessment
ACTUAL T2 VIRTUAL
“Reductionist computation Time
Nightmare” or
regulation
empirics Components
Emergence STRATEGIES
NORMS
“REAL” assessment
ACTUAL T1 VIRTUAL
“of actions” COALGEBRA
Semantics works in observed behaviour not “real” states
Equifinality is likely – many potential equivalent models
Hauhs & Widemann, 2010
32. Modelling - reflexively
• Any models must inevitably be meta-models and any
theory must be a meta-theory whose referents are not
merely measured facts or data but also 1st and 2nd
order models of these data and the relation between
both e.g. the process of model building or learning
(Atmanspacher & Wiedenmann, 1999).
• The normative view of science – especially applied,
instrumental systems science – is that of an “ongoing
cultural activity with all its dynamical, dialectical and
normative aspects” (his italics, Mattessich, 1978, p.
261).
33. The world of “is” Hume (1739) The world of “ought”
A systems methodology combining the realist and the normative
THEORY METATHEORY
Emergence
“atoms” “Components”
Information
bisimulation
Realist, rationalist Normative
Strong inference, “science” Strategies, rules and norms
Can we find a “language” to do this? Dynamic, reflexive, inclusive?
34. A quite different “science”
• Theory and meta-theory: meta-models
– Constraints plus norms and values
• Meta-analysis via meta-statistics (2nd order)
including realist and normative factors
• “Naive” observations of “pure” facts (1st order
statistics) neither confirm or refute meta-models
• “Experiments” have to include the relationships
between data and models – value judgements
• Multiple models (various norms) reproduce
different aspects of complex systems – data,
model, people, norms – recursive relations
35. Computational mechanics
• Non-stationary analyses of complexity
– Beyond statistics, towards structure
• Crutchfield’s work on ϵ - machines
– Information theoretic measures (not stats)
– Complexity metrics for detecting structure and
quantifying emergence
– Hidden order parameters; 1st and 2nd order stats
• Analysis of information in time series
– Objective methods – prediction, errors
36. 2nd order statistic
1st order statistic
Characteristics of hydrographs: “life” as a complex, fractal filter
37. Epistemological pluralism
• In a paper in Bioscience in 2006, Kevin Rogers and
colleagues (Mills et al., 2006) wrote that is was time to
develop a “fundamentally new course” for tackling these
kinds of complex, heterogeneous.....problems. They likened
the usual scientific approach to a horse race in which most
of the horses were shot at the starting gate before the race
got underway. Instead they advocated a more pluralistic
approach in which “all horses (even those which appear
lame at the first appearance) are nurtured and coaxed to
their full capacity.”
• Mitchell (2009) has similarly argued for a new “integrated
pluralism” in our epistemology and methodology:
traditional reductionist, materialist approaches do not work
sufficiently well.
38. Recursion and complexity
• “Scientific”, modernist management fails
– Local interaction, CA, GA, A(R-D) “noise”
– Debatable evidence, recursion, “nudge”
• Process network in real time
– Extreme events, tipping points?
• Information, transmission, storage,
architecture: upward and downward causes
– Meta-architecture (Douglass North), institutions
39. Uncertainty
• Both aleatory and epistemic, weak induction
• “systems methodology” is a new kind of
uncertainty principle; new kind of risk
• Not “normally” distributed; power laws
• “Anti-fragility” (Taleb), redundant, evolving,
prone to failure (“super-transients”), options,
keep the upsides, “fast failure” innovation,
investment strategies
40. Power laws and 2nd order stats
• Power laws give finite probabilities of events
of any size; require generating mechanism
• Use of 1st order statistics in complex situations
underestimates risk
• 2nd order (non-stationary) stats, trajectories
• Science, economics, engineering practically
and institutionally reduces risk as something
to be “controlled” and eliminated by more
research (Wynne)
41. INDUCTIVE FALLACIES
Especially in complex situations, we have a problem with induction: there
is no “logically safe procedure for obtaining nontrivial universal empirical
truths” (Goodman, 1955; Hauhs & Widemann, 2010).
42. Biology vs infrastructure
• Biology is “bottom up”, anti-fragile (Taleb)
– CA, GA, A(R-D) – cascading failures, redundancy,
diversity, evolution, keep “upsides” options
– Recursive, adaptive in real time
– BUT unstable, non-equilibrium, supertransients
• Infrastructure is “top down”, fragile
– Designed in advance, less diverse, low redundancy
– Fixed structure, rigid networks, super-transients
– Stable except to (small) unexpected shocks
44. Graham Harris
Prof of Infrastructure and Env. Systems
Graham_Harris@uow.edu.au
0417 463 158
45.
46. So what do we do??
• Pete and Paul’s problem....
– Stop thinking in terms of entities
– Accept that knowledge is partial (limited beings)
– Accept that “evidence” will be partial also
– Look for new measures: information storage and
transmission; hidden order
– Think levels, emergence, 2nd order stats
– Align meta-models with meta-architecture
– Expect surprises, look for SMS
47. The myth of models
Uncertainty
Prediction or
Prophecy?
Read Gregory Bateson, 1970s, Robert Rosen, Keith Beven
Life is different – physics won’t do!
Either cybernetic, “systems” gigantism or multi-agent models:
either way cal/val issues and prediction problems
48. ABSTRACTION
The set of all models
Newtonian MAS, Swarms
Information
theoretic
GEP Box “All models are wrong but some are useful”
49. Epistemological uncertainty
Complexity implies confuses
Knightian uncertainty
Or “unknown unknowns”
Complexity
Model
Conceptual Science
error Result
Input Output
error error
Reification Data
Data
Realism and rationalism Aleatory
Aleatory
Culture and values uncertainty
uncertainty
50. Hydrological models: Dmitri Kavetski
Process
True inputs Responses
descriptions
Independent Independent
Aleatory Aleatory
Uncertainty Input errors Parameter and Uncertainty
Output errors
Structural errors
Conceptualised
Obs inputs Model result
model
Potentially Potentially
Strong Strong
Priors Priors
Identifiability
Problem Epistemic uncertainty Beven GLUE
Standard or Exogenous errors Model
Weak priors
Weighted Non-stationary Performance
Inability to
Least squares Non-linear
distinguish
Perform poorly Epochs in time
“Monstrous” input errors
51. Complexity and “received” science
• Bottom up emergence and evanescent
structure defeats Newtonian modelling
– Reductionist nightmare, pragmatism
– Laws are well scrubbed Sherlock Holmes stories
(Cohen & Stewart)
• Structural errors, uncontrolled variances,
“monstrous” statistical errors
• Cal/val issues (Oreskes), prediction?
54. A systems methodology combining the positivist and the normative
MODELS META-MODELS
THEORY METATHEORY
LEVELS
EMERGENCE
STATISTICS META-STATISTICS
Positivist, rationalist Normative
RECURSIVE DEVELOPMENT
55. Explanation, prediction
• Statistical limits – more information in “big
and slow” (contexts) than in “small and fast”
• Ability to control contexts, bifurcation points
– Safe minimum standards, robust decision making,
MaxiMin etc
• Levels, predictability, explanation