Believe in the Model: Mishandle the Emergency

1
Believe in the ModelMishandle the Emergency

• Simon French and Carmen Niculae
• Manchester Business School

2
Issues

• Is our view of emergency management too
  scientific and technocratic?
• Are our models appropriate for supporting
  emergency management?
• Do we understand the decision context fully?

3
Decision contexts

• Emergency management is essentially the process
  of taking a sequence of decisions as the
  situation evolves
• Thus to understand emergency management we need
  to understand decisions and their evolving
  contexts
• Snowden has classified decision contexts in an
  informative way

4
Cynefin model of decision contexts
Knowable Cause and effect can be determined with
sufficient data The realm of scientific inquiry
Chaos Cause and effect not discernable
Known Cause and effect understood and
predictable The realm of scientific knowledge
5
Different styles of decisions for different
contexts
Knowable Cause and effect can be determined with
sufficient data Assess, learn and respond
Chaos Cause and effect not discernable
Known Cause and effect understood and
predictable Categorise and respond
Act, reflect, act
6
The context expected by current emergency
management DSS
7
and then as data accumulates
8
What can we learn from the past?
9
Chernobyl
10
TMI
11
Non nuclear events/issues

• Outside the Nuclear domain
• BSE
• GMOs
• MMR
• Challenger and Columbia Shuttle Disasters
• All indicate
• Change of role of science and public trust of
  science
• Difficulty of communication between scientists,
  managers and the public

12
An accident is an event in a human society
Long term
Late Phase
Early Phase
13
Decisions are not independent

• What we do in the early phase sets the context
  for the later phase and the long term
• in fact,
• what we say in the early phase sets the context
  for the later phase and the long term
• and
• the early phases responses can constrain the
  flexibility we may need in the long term

14
The domains of scientific models
Value based thinking important
QuantitativeScientific Models are usedto encode
Knowledge
15
Knowledge and Decision Making
16
Tacit and Explicit Knowledge

• Explicit Knowledge can be codified
• codified ? can be stored in a computer
• Tacit knowledge cannot be codified
• skills, expertise, flair and, above all, values,
  empathy,

17
Players
Decision Makers
18
Knowledge Management Systems

• use flexible data/information management systems
  to allow storage and access to data and model
  based information and forecasts
• explicit knowledge
• collaborative working tools to share and work
  together, allowing softer skills and expertise to
  be deployed
• tacit knowledge

19
Cynefin model of decision contexts
20
Decision Support and Models

• Scientific Models encode our understanding of the
  past.
• Models for decision support need to provide
  requisite predictions of the future.

21
Information sought by decision makers

• Evatech project
• nuclear emergency management processes
• workshops
• 9 workshops across Europe supported by DSS
  (RODOS, CONDO)
• clean-up actions
• systems offered much complex information and
  forecasts
• decision makers used simple summaries of dose,
  cost, waste
• moreover, what data they did used needed to be
  re-summarised using spreadsheets

22
Key issues in using Scientific models

• Models poorly calibrated for emergency management
• too few datasets (thankfully!)
• Experts are overconfident in their models
• DMs do not understand the models
• DMs do not like uncertainty and conflicting views
• Different models

23
Ensemble Exercise 7
DK1 FI1NL1 DE1 Total Integrated
Concentration Analysed meteorology
24
But perhaps the key issue is

• Models poorly calibrated for emergency management
• too few datasets (thankfully!)
• Experts are overconfident in their models
• DMs do not understand the models
• DMs do not like uncertainty and conflicting views
• Different models
• Scientific models focus attention on the known
  and knowable domains

25
In summary

• The context will almost certainly become
• complex
• Thus we need a socio-technical decision support
  process
• which anticipates this complexity
• multi-disciplinary
• integrated
• Is it sometimes too partitioned into different
  groups of experts?