INTERACTIVE COMPUTER SUPPORT IN DECISION CONFERENCING: THE CASE OF NUCLEAR EMERGENCY MANAGEMENT Jyri Mustajoki, Raimo P. Hämäläinen and Kari Sinkko Helsinki University of Technology Systems Analysis Laboratory P.O. Box 1100, FIN-02015 HUT, Finland jyri.mustajoki@hut.fi, raimo@hut.fi, kari.sinkko@stuk.fi 1. INTRODUCTION We have applied different ways of decision conferencing as an approach to support complex decision making problems [see e.g. Hämäläinen and Leikola 1996; Hämäläinen et al. 1998, 2000a, 2000b, 2001]. In decision conferences, also called facilitated workshops, all the decision making parties are gathered to discuss and process the problem thoroughly. All the relevant alternatives are evaluated with a systematic approach. Computer support can be used to facilitate different phases of the process, e.g. modeling of the problem, evaluation of the alternatives and analysis of the results. In this paper, we study the use of a decision support system (DSS) in a decision conference for planning countermeasures to dairy products in a nuclear accident. A day long computer aided conference exercise was held to consider the problem from different perspectives. From the point of view of radiation safety, the objective of the conference was to consider different alternatives for the possible countermeasures. The experiences and the results can then be utilized in a real emergency situation. From the decision analytic point of view, the objective was to gather user experiences on the use of web-based group decision support systems in decision conferencing. 2. DECISION SUPPORT SYSTEM A DSS used to support the analysis consisted of seven portable computers, a wireless local area network, a projector and MCDA software on the Decisionarium site [Hämäläinen 2000, 2001] (Figure 1). The participants of the conference (14) were representatives of the Ministry of Agriculture and Forestry, the National Food Administration, the Finnish Food and Drink Industries' Federation, the Association of Rural Advisory Centres and the Radiation and Nuclear Safety Authority. They were divided into six groups according to their backgrounds, and each group was handed a portable computer. An essential difference between this and many earlier conferences was that the participants used the system by themselves. Most of them had no prior experience about DSSs, which induced requirements for the usability of the decision analytic methods as well as the supporting software. 279
Server Internet Projector Decision Makers Figure 1. Portable decision support facility The facility used is portable and easily installable in different locations. It also enables the use of web resources to support the analysis. For example, the backgrounds of the case were available for the participants on a web page during the conference (Figure 2). Three different software, Web-HIPRE [Hämäläinen and Mustajoki 1999, Mustajoki and Hämäläinen 2000], WINPRE [Hämäläinen and Helenius 1997] and Opinions-Online [Hämäläinen and Kalenius 1999], were used. Of these, Web-HIPRE and Opinions-Online can be operated via the Internet or a local area network also making remote participation possible. These software form a logical entity to support different phases of the analysis. Web-HIPRE was used for modeling the problem, evaluating the alternatives and analyzing the results. The SWING method [von Winterfeldt and Edwards 1986] was used in the attribute weighting. The individual results were aggregated with a group model via the local area network. Uncertainty around the problem attributes was modeled with the interval SMART/SWING technique [Salo and Hämäläinen 1992] using the WINPRE software. Opinions-Online was used to survey the approval of the alternatives. The progress of the conference can be divided into four phases: introduction to the decision conference and the methods, discussion and construction of the model, weighting process, and analysis of the results. Further details of the progress of the conference are available in [Sinkko et al. 2001]. 3. NUCLEAR ACCIDENT CASE The objective was to plan countermeasures to dairy products in a hypothetical nuclear accident in a nuclear power plant in western Finland. We assumed that the decision conference is arranged a week after the accident, when the fallout has been roughly mapped. For the first week after the accident the cattle were sheltered as a precautionary action. The attributes of the case were preliminary derived in the preparatory meetings with the experts. The experiences gathered in previous decision conferences on nuclear emergency [Hämäläinen et al. 1998, 2000a, 2000b] were also used when deriving the attributes. The final value tree (Figure 3) and the exact meanings of the attributes were defined in the conference. 280
Figure 2. WWW page of the case Figure 3. Value tree of the case The considered countermeasures were provision of uncontaminated fodder ( Fod ), processing of the milk ( Prod ) and banning the milk ( Ban ). In addition, an action where nothing ( --- ) is done was considered. These were also preliminarily 281
derived in the preparatory meetings. The actual six alternatives were combinations of the different countermeasures for the weeks 2 5 and 6 12 after the accident. The consequences of the actions were estimated with the RODOS system [Ehrhardt and Weis 2000]. The details of the case are available in [Sinkko et al. 2001]. 4. WEIGHTING METHODS AND RESULTS The primary weighting was conducted with SWING method using Web-HIPRE. It turned out to be an easy enough method to use by the participants. As Web-HIPRE models are saved on the server, these were easily retrieved and projected on the wall for the analysis of the results. For most groups, the top alternatives were the alternatives where clean fodder is provided for at least either of the periods. However, the ranking of these alternatives varied between the groups. Alternative no actions ( ---+--- ) was among the three worst alternatives for every group. Figure 4 presents an example of the values of the alternatives for one participant group. Figure 4. Results of one participant group The individual models were combined with a weighted arithmetic mean method (see e.g. Keeney 1976) using Web-HIPRE s group facility. We did not make a stand on the importances of the groups, but weighted the groups equally. The top three alternatives in the group model were Fod+Fod (0.879), Prod+Fod (0.809) and Fod+--- (0.776). We also tested the rank-based SMARTER method [Edwards and Barron 1994]. In four out of six groups the ranking of the alternatives was completely the same with SMARTER and SWING weights. In the other two groups only one pair of alternatives swapped ranks. Thus, at least in this case SMARTER appeared to be an 282
applicable method to quickly get preliminary results. However, no conclusions on the general applicability about SMARTER can be drawn from this study. The results were collectively analyzed and discussed. The sensitivity of the results was studied with single parameter sensitivity analyses on both the individual model and the group model. We also asked the participants to vote for the approval of the alternatives with Opinions-Online. Two alternatives ( Fod+Fod and Prod+Fod ) were approved by all the participants. Thus, although one clearly the best alternative could not be identified, the participants were able to reach a consensus by finding even two alternatives, which were feasible for all of them. We also carried out an interval sensitivity analysis (Lindstedt et al. 2000), where all the possible uncertainties in participants preference judgments can be simultaneously modeled with intervals. The participants were asked to model the uncertainties with the interval SMART/SWING method using WINPRE software. In most cases, alternative no actions was found dominated. Another interval sensitivity analysis based on the uncertainties given as error percentages for the parameters was also tested. For example, when adding error intervals to the example in Figure 4, three alternatives ( Fod+---, Fod+Fod and Prod+Fod ) were found non-dominated. This supports the earlier finding that at least some countermeasures should be done, but the type of countermeasures is not that substantial. 5. CONCLUSIONS The experiences of the conference strongly supported the format of the decision conference with interactive DSS. All of the participants considered the conference useful. On the basis of the analysis, the participants were able to reach a consensus about the possible countermeasures. Most of the participants saw computer supported decision conference as a potential approach to this kind of problems in general. However, if the participants themselves use the DSSs, easy-to-understand methods and software are necessities. The existing technique would also make the arranging of the conference on the web possible. The needed background information is on the web, the primary software used (Web-HIPRE and Opinions-Online) can be operated via the Internet and the existing conference tools could be used for conversation. Furthermore, the software used turned out to be easy enough to be used also by the decision makers themselves. REFERENCES EDWARDS, W., BARRON, F.H. (1994), SMARTS and SMARTER: Improved Simple Methods for Multiattribute Utility Measurement, Organizational Behavior and Human Decision Processes, 60, 306-325. EHRHARDT, J., WEIS, A. (Eds.) (2000), RODOS: Decision support system for off-site nuclear emergency management in Europe, European Commission, EUR 19144 EN. (www.rodos.fzk.de) HÄMÄLÄINEN, R.P. (2000), Decisionarium global space for decision support, Systems Analysis Laboratory, Helsinki University of Technology. (www.decisionarium.hut.fi) 283
HÄMÄLÄINEN, R.P. (2001), Decisionarium Global Support for Group Decisions and Negotiations, Group Decision and Negotiations 2001 Conference, La Rochelle, France, June 4-7. HÄMÄLÄINEN, R.P., HELENIUS J. (1998), WINPRE Workbench for Interactive Preference Programming, Computer Software, Systems Analysis Laboratory, Helsinki University of Technology. (Downloadable at www.sal.hut.fi/downloadables/winpre.html) HÄMÄLÄINEN, R.P., KALENIUS R. (1999), Opinions-Online Platform for Global Participation, Voting Surveys and Group Decisions, Computer Software, Systems Analysis Laboratory, Helsinki University of Technology. (www.opinions.hut.fi or www.opinions-online.com) HÄMÄLÄINEN, R.P., KETTUNEN, E., MARTTUNEN, M., EHTAMO, H. (2001), Evaluating a framework for multi-stakeholder decision support in water resources management, Group Decision and Negotiation. (to appear) HÄMÄLÄINEN, R.P., LEIKOLA, O. (1996), Spontaneous decision conferencing with top-level politicians, OR Insight, 9(1), 24-28. HÄMÄLÄINEN, R.P., LINDSTEDT, M., SINKKO, K. (2000a), Multiattribute Risk Analysis in Nuclear Emergency Management, Risk Analysis, 20(4), 455-467. HÄMÄLÄINEN, R.P., MUSTAJOKI J. (1998), Web-HIPRE Global Decision Support, Computer Software, Systems Analysis Laboratory, Helsinki University of Technology. (www.hipre.hut.fi) HÄMÄLÄINEN, R.P., SINKKO, K., LINDSTEDT, M., AMMANN, M., SALO, A. (1998), RODOS and decision conferencing on early phase protective actions in Finland, Radiation and Nuclear Safety Authority, STUK-A159. (Downloadable at www.sal.hut.fi/publications/pdf-files/rham98.pdf) HÄMÄLÄINEN, R.P., SINKKO, K., LINDSTEDT, M., AMMANN, M., SALO, A. (2000b), Decision analysis interviews on protective actions in Finland supported by the RODOS system, Radiation and Nuclear Safety Authority, STUK-A173. (Downloadable at www.sal.hut.fi/publications/pdf-files/rham00.pdf) KEENEY, R.L., RAIFFA, H. (1976), Decisions with Multiple Objectives, John Wiley & Sons, Inc. LINDSTEDT, M., HÄMÄLÄINEN, R.P., MUSTAJOKI, J. (2000), Using Intervals for Global Sensitivity Analyses in Multiattribute Value Trees, Systems Analysis Laboratory, Helsinki University of Technology. Manuscript. (Downloadable at www.hut.fi/units/sal/publications/pdf-files/mlin.pdf) MUSTAJOKI, J., HÄMÄLÄINEN, R.P. (2000), Web-HIPRE: Global Decision Support by Value Tree and AHP Analysis, INFOR, 30(3), 208-220. SALO, A., HÄMÄLÄINEN, R.P. (1992), Preference assessment by imprecise ratio statements, Operations Research, 40(6), 1053-1061. SINKKO, K., AMMANN, M., KOSTIAINEN, E., SALO, A., LISKOLA, K., HÄMÄLÄINEN, R.P., MUSTAJOKI, J. (2001), Evaluating measures to protect the milk pathway in accidental radioactive releases, Radiation and Nuclear Safety Authority, Finland. Manuscript. (in Finnish, an English translation will be available soon) VON WINTERFELDT, D., EDWARDS, W. (1986), Decision Analysis and Behavioral Research, Cambridge University Press, Cambridge, UK. 284