S ystems Analysis Laboratory Helsinki University of Technology Decision Conferencing in Nuclear...

S ystemsAnalysis LaboratoryHelsinki University of Technology

Decision Conferencing in Nuclear Emergency Management

by

Raimo P. Hämäläinen

Mats Lindstedt

Kari Sinkko


Contents of presentation:

• background of the study

• decision conferences at STUK

• results and conclusions


RODOS project

• a Real-time On-line DecisiOn Support project to develop a group support system for nuclear emergency management

• sponsored by the European Commission and started in 1990

• in Finland STUK (Radiation and Nuclear Safety Authority) participates in the project

• the decision conferences were part of the RODOS project and organized by STUK


Objectives

• to study decision conferencing and its suitability in RODOS

• to study the use of the RODOS software

• to study the incorporation of uncertainties


Decision conferencing• refers to intensive, computer supported

meetings

• a group of people develops a shared understanding of a common problem

• develops a decision analysis model, assisted by a facilitator

• originally a two-day meeting


Decision conferencing

• here a faster type of decision conferencing was used (a few hours)

• prestructured value trees or separate decision making groups

• decision analysis interviews


Conferences at STUK

• early phase countermeasures (a few hours after the accident)

• iodine tablets, sheltering, and evacuation

• the RODOS software was used to calculate accident data and impacts of countermeasures

• first phase of the conferences: values and attributes

• second phase of the conferences: uncertainties

• participants from STUK and from the Finnish nuclear power companies


First series of decisionconferences in 1997

Second meeting:Decision conference

First meeting: Problemstructuring session

First phase of the decisionconferences: attributes and values

Second phase of the decisionconferences: uncertainties

Second meeting:Decision conference

First meeting:Preconference

Second series of decisionconferences in 1998

Conferences at STUK


Conferences at STUK

Facilitators

Decision Makers

Screen used for projecting the computerscreens and slides

Analyst for thedecision aiding

software

Analyst for theRODOSsoftware


Objectives of the first conference

• to define the factors and attributes important when deciding on countermeasures

• no uncertainties included


Problem structuring

session

Preliminary value tree

Overall

Health

Economics

Political

Industry

State/Atom pool

Technical feasibility

Socio-Psychological

Hereditary effects

Non-radiation

Cancer

Deterministic

Adults

Children

Immediate costs

Citizens

Industry

Society

Social

Health related

Safety related

Terratogenic

Int. confidence

Public confidence

National safety

Value of property

Economy

Loss of income

Immediate costs

Agriculture

Tourism

Production

Protection

Compensation

Ability to recover

Citizens

National property


First decision conference

Finalvalue tree

Overall

Health

Economics

Political

Losses duringrecovery

State/Atompool

Technicalfeasibility

Socio-Psychological

Non-radiation& other

Rescueworkers

Othercancers

Thyroid cancer

Social

Health andsafety


First decision conference-the strategies and their impacts

Attribute Unit No action Reluctant Medium RigorousHealth Thyroid cancer number of incidents 1000 900 700 0 Other cancers number of cancer deaths 400 300 220 0 Rescue workers number of deaths 0 1 1 2 Non-radiation and other

number of deaths 0 3 7 10

Socio-Psychological Health and safety

very neg-very pos(100 - 0)

80 80 70 40

Social very neg-very pos(100 - 0)

0 5 40 80

Technical feasibility very low-very high(100 - 0)

0 1 10 100

Economics State/ Atom pool MFIM 250 240 500 20000 Losses during recovery

MFIM 1000 1000 1000 10000

Political very neg-no change(100 - 0)

100 30 30 80


First decision conference- weights given by participants

Attribute LeastPreferred

Level

MostPreferred

Level

DM 1SMARTER

DM 2SMART

DM 3SMART

Thyroid cancer 1000 0 0.337 0.250 0.235Other cancers 400 0 0.212 0.250 0.188Rescue workers 2 0 0.013 0 0.047Non-radiation & other 10 0 0.013 0 0.012Health and safety 100 0 0.076 0 0.235Social 100 0 0.030 0 0.118Technical feasibility 100 0 0.107 0 0.047State/ Atom pool 20000 0 0.149 0.250 0Losses during recovery 10000 0 0.013 0.250 0Political 100 0 0.051 0 0.118


First decision conference- rankings

AlternativeNo actionReluctantMediumRigorous

Utility 0.337 0.446 0.536 0.610

State/Atom pool Losses d. recovery Health and SafetySocial Technical feasib. PoliticalThyroid cancer Other cancers Rescue workersNon-radiation&other


Utility 0.472 0.559 0.656 0.500



Utility 0.271 0.390 0.448 0.612


1 2

3


• uncertainties included

• it was known that an accident had happened, but it was not known how severe it had been

• 5%, 50%, and 95% fractiles used

three accident scenarios

Second decision conference


Second decision

conference

Overall

Health

Costs

Political cost

Socio-Psychological

Othercancers

Thyroid cancer

Negativeeffects

Positiveeffects

Finalvalue tree


Second decision conference-impactsAttribute Unit Strategy

0Strategy

1Strategy

2Strategy

3Strategy

4Health Thyroid cancer nr of cancer 5% 0 0 0 0 0

incidents 50% 20 5 2 2 495% 240 50 20 20 40

Other cancers nr of cancer 5% 0 0 0 0 0incidents 50% 22 20 20 20 12

95% 320 286 288 286 204Socio-Psychological Positive effects no change- 5% 0 100 10 10 0

very pos 50% 0 75 50 45 40(0 - 100) 95% 0 50 90 80 80

Negative effects no change- 5% 40 0 90 80 50very neg 50% 70 40 50 45 35(0 - 100) 95% 100 80 10 10 20

Costs MFIM 5% 0 9.7 13.5 13.5 960.2MFIM 50% 11.8 18.7 22.5 22.5 962.9MFIM 95% 166.2 143.1 145.3 144.5 1056.2

Political cost no change- 5% 30 0 0 20 80very neg 50% 65 40 40 30 50(100 -0) 95% 100 80 80 40 20


Second decision conference- eliciting utility functions

Lottery question:

Please select the number of cancer incidents (L) that would make you indifferent between getting that amount for sure and a fifty-fifty chance of getting 250 cancer incidents or 0 incidents.

Please set Level L and probability P so that certain Alternative Aand lottery Alternative B are equally preferred

A B

L = 50

P = 0.500250

1 - P = 0.5000

Risk Premium = 75, Local Risk Aversion (r) = 0.0131251


Second decision conference- utility functions

Utility

Thyroid cancer (nr. of incidents)

1

0

0 250

Selected Point -- Level:50 Utility:0.5

Utility

Costs (MFIM)

1

0

0 1100

Selected Point -- Level:1000 Utility:0.5


Second decision conference- weights

Attribute LeastPreferred

Level

MostPreferred

Level

SMART Tradeoff SMART

95%

Thyroidcancer

250 0 (20) 0.328 0.210 0.400

Othercancers

400 0 (200) 0.262 0.105 0.120

Positiveeffects

0 100 0.016 0.030 0.040

Negativeeffects

100 0 0.098 0.101 0.080

Costs 1100 0 (100) 0.033 0.050 0.040Politicalcost

100 0 0.262 0.504 0.320


Second decision conferenceAlternativeStrategy 0Strategy 1Strategy 2Strategy 3Strategy 4

Utility 0.566 0.743 0.753 0.783 0.722

Costs Other cancers Political costSoc.-Psych Negative Soc.-Psych Positive Thyroid cancer

AlternativeStrategy 0Strategy 1Strategy 2Strategy 3Strategy 4

Utility 0.474 0.697 0.694 0.748 0.628


Rankingwith

SMART

Rankingwith

Tradeoff


Second decision conference- ranking with SMART (95% fractile)

Alternative

Strategy 1Strategy 2Strategy 3Strategy 4

Utility 0.043 0.431 0.636 0.762 0.781


Strategy 0


Observations

• the decision conferencing format was successful

• a lot of progress was made in just a few hours, with more training this method could be used in a real situation

• using prestructured value trees or benchmarks seems a promising way forward

• brainstorming was a good way to get the process started

• the participants were able to agree on the value trees

• provides a common framework from which to discuss the situation


Problems

• the choice of strategies was too limited, the best choice was too obvious

• the attributes need to be better defined

• the terminology used needs to be clearer

• the case assumed a single decision point, in reality sequential decision making would be used

• the participants did not feel that the weighting of the attributes was very appropriate


Understanding uncertainties

• this was found to be very difficult

• the participants rather focused on the worst case scenario (95% fractile) and ignored the probabilities

• there was no uncertainty about the accident, if there had been the situation would have been even more difficult


Conclusions - RODOS software

• still a prototype, but could have worked better

• problems with presenting the data using thematic maps

• does not yet allow what-if analyses

• the software was not used very much during the conferences

• the participants felt RODOS could be used to provide data on the accident and to calculate impacts

• they did not feel RODOS could help in the actual decision making


References

Hämäläinen, R. P, Leikola, O. 1996. Spontaneous Decision Conferencing with Top-level Politicians. OR Insights Vol 9, pp. 24-28.

Hämäläinen R.P., Sinkko K., Lindstedt M. 1999. Multi-Attribute Risk Analysis in Nuclear Emergency Management. Risk Analysis, 2000.

Hämäläinen R.P., Sinkko K., Lindstedt M., Ammann M. and Salo A. 1998. RODOS and decision conferencing on early phase protective actions in Finland, Radiation and Nuclear Safety Authority, STUK-A159, December, pp. 1-76. Downloadable at http://www.hut.fi/Units/SAL/Publications/

Hämäläinen R.P., Sinkko K., Lindstedt M., Ammann M. and Salo A. 1999. Decision analysis interviews on protective actions in Finland supported by RODOS system. Radiation and Nuclear Safety Authority, STUK-A173, February 2000, pp. 57. Also RODOS Report - Decision Support for Nuclear Emergencies, RODOS(WG7)-TN(99)-04.

Hämäläinen R.P., Lindstedt M. and Sinkko K. 2000. Decision analysis interviews in nuclear emergency management. Manuscript.

French, S., Finck, R., Hämäläinen, R.P., Naadland, E., Roed, J., Salo, A. and Sinkko K.. 1995. An exercise on clean-up actions in an urban environment after a nuclear accident, Nordic Decision Conference, Sweden, 20-31, August.

Hämäläinen R P. 1988. Computer Assisted Energy Policy Analysis in the Parliament of Finland. Interfaces; 4 (Vol. 18): 12-23.

S ystems Analysis Laboratory Helsinki University of Technology Decision Conferencing in Nuclear...

Documents

Transcript of S ystems Analysis Laboratory Helsinki University of Technology Decision Conferencing in Nuclear...