Black swans
in a risk context
Terje Aven
JRC, ISPRA
University of Stavanger
21 June 2013
Talking about black swans
• Creates a lot of enthusiasm
• Hard negative words from some
researchers
Aven (2013) On the meaning of a black swan in a risk
context. Safety Science
Professor Dennis Lindley
Taleb talks nonsense
He lampoons Taleb’s distinction between the lands of Mediocristan
and Extremistan, the former capturing the placid randomness
as in tosses of a coin, and the latter covering the dramatic
randomness that provides the black swans
No need to see beyond probability
•Mediocristan (Normalistan)
•Extremistan (black swans)
Nassim N. Taleb
Lindley example
A sequence of independent trials with a
constant unknown chance p of success (white
swan)
Lindley shows that a black swan is almost
certain to arise if you are to see a lot of swans,
although the probability that the next swan
observed is white, is nearly one.
Prior density for
p: the chance of a
white swan
1
1
Prior density for
p: the chance of a
white swan
1
What is the
probability that p=1?
1
Prior density for
p: the chance of a
white swan
1
What is the
probability that p=1?
It is zero!
1
Prior density for
p: the chance of a
white swan
1
There is a positive
fraction of black swans
out there !
1
The probability-based approach
to treating the risk and
uncertainties is based on a
background knowledge that
could hide critical assumptions
and therefore provide a
misleading risk description
Prior density for
p: the chance of a
white swan
0.8
x
x
0.2
0.99
1
Prior density for
p: the chance of a
white swan
0.8
x
x
0.2
the probability of a
black swan occurring is
close to zero
0.99
1
Depending on the
assumptions made,
we get completely different
conclusions about the
probability of a
black swan occurring
Lindley’s example also fails to reflect
the essence of the black
swan issue in another way
In real life the definition of a probability
model and chances cannot always be
justified
P(attack)
Main problems with the probability
based approach
1
Assumptions can
conceal important
aspects of risk and
uncertainties
3
The probabilities
can be the same
but the
knowledge they
are built on
strong or weak
2
Presume
existence of
probability
models
4
Surprises occur
15
Risk perspective
Probabilitybased
Historical data
+
Knowledge
dimension
+
Surprises
P(head) = 0.5
P(attack) = 0.5
Strong
knowledge
Poor
knowledge
John offers you a game: throwing
a die
• ”1,2,3,4,5”:
• ”6”:
What is your risk?
6
-24
Risk
(C,P):
• 6
5/6
• -24 1/6
Is based on an important
assumption – the die is fair
“Background knowledge”
Assumption 1: …
Assumption 2: …
Assumption 3: …
Assumption 4: …
…
Assumption 50: The platform jacket structure will withstand
a ship collision energy of 14 MJ
Assumption 51: There will be no hot work on the platform
Assumption 52: The work permit system is adhered to
Assumption 53: The reliability of the blowdown system is p
Assumption 54: There will be N crane lifts per year
…
Assumption 100: …
…
Model: A very crude gas dispersion model is applied
Risk perspective
Probabilitybased
Historical data
+
Knowledge
dimension
+
Surprises
Black swan (Taleb 2007)
• Firstly, it is an outlier, as it lies outside the realm of
regular expectations, because nothing in the past
can convincingly point to its possibility.
• Secondly, it carries an extreme impact.
• Thirdly, in spite of its outlier status, human nature
makes us concoct explanations for its occurrence
after the fact, making it explainable and predictable.
22
Aven (2013) questions whether a
black swan is
1. A surprising extreme event relative to the
expected occurrence rate
2. An extreme event with a very low probability.
3. A surprising, extreme event in situations with
large uncertainties.
4. An unknown unknown.
Black swan (Aven 2013)
A surprising extreme event relative to the present
knowledge/beliefs.
Hence the concept always has to be viewed in relation
to whose knowledge/beliefs we are talking about, and at
what time.
Unforeseen/surprising events:
A. Events that were completely unknown to
the scientific environment (unknown
unknowns)
B. Events that were not on the list of known
events from the perspective of those who
carried out a risk analysis (or another
stakeholder)
C. Events on the list of known events in the
risk analysis but found to represent a
negligible risk
Government building Oslo 22 July 2011
Threats
Known
unknowns
(A’, C’, Q, K)
Unknown
unknowns,
black swans
It is not about assigning correct
probabilities
• But to provide
– a proper understanding of the total system
– means to identify many of these B and C
events
– measures to me meet them, in particular
resilient measures
– means to read signals and warnings to
make adjustments
28
Statfjord A
Do we have black swans
here?
How to confront black swans
• Improved Risk Assessments
• Robustness
• Resilience
• Antifragility
How to confront black swans
• Improved Risk Assessments
• Robustness
• Resilience
• Antifragility
Taleb: propose to stand our current approaches
to prediction, prognostication, and risk
management
PETROMAKS project:
Improved risk assessments
- to better reflect the knowledge
dimension and surprises
Unforeseen/surprising events:
A. Events that were completely unknown to
the scientific environment (unknown
unknowns)
B. Events that were not on the list of known
events from the perspective of those who
carried out a risk analysis (or another
stakeholder)
C. Events on the list of known events in the
risk analysis but found to represent a
negligible risk
• Not seeing what is coming, when we should
have seen it
- Preoccupation with failure
- Reluctance to simplify
- Sensitivity to operations
Commitment to resilience
Deference to expertise
2
Mindfulness
(Collective)
2
Quality
management
New way of
thinking about
risk
1
Risk analysis
and
management
1
Concepts and
principles
Aven and Krohn (2013) RESS.
Analysis
Risk analysis
Describing
uncertainties, …
Management
Management
review and
judgment
Risk-informed decision
making
Decision
• Extra
Risk
(A,C,U)
A: Event,
(C,U)
C: Consequences
U: Uncertainty
Risk description
Q
(A,C,U)
(C,U)
K
C’
Q: Measure of uncertainty (e.g. P)
K: Background knowledge
C’: Specific consequences
Subjective/knowledge-based
probability
• P(A|K) =0.1
• The assessor compares his/her uncertainty (degree
og belief) about the occurrence of the event A
with drawing a specific ball from an urn that
contains 10 balls (Lindley, 2000. Kaplan and Garrick 1981).
K: background knowledge
Analysis
Risk analysis
Cost-benefit analysis,
Risk acceptance criteria
…
Management
Management
review and
judgment
Risk-informed decision
making
Decision
Download

Electronic Presentation Guide - 2004 HPS Topical Meeting