Iscram summerschool12 decisions

1,084 views

Published on

Tina Comes' (complete!) presentation on scenario-based decision support for the ISCRAM summerschool 2012

Published in: Education, Technology
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
1,084
On SlideShare
0
From Embeds
0
Number of Embeds
544
Actions
Shares
0
Downloads
20
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Iscram summerschool12 decisions

  1. 1. Decision Making and Scenario Planning2012 ISCRAM Summer School on Humanitarian Information ManagementTina ComesResearch Group: Risk ManagementInstitute for Industrial Production (IIP)KIT – University of the State of Baden-Wuerttemberg andNational Research Center of the Helmholtz Association www.kit.edu
  2. 2. Risk Management?Aim: support decision-makers in complex and uncertain situations bridge the gap between formal models and transparent, ready-to-use evaluations collaborative and distributed decision support tools based on modern ICT systemsTina Comes Decision Making and Scenario Planning 2Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  3. 3. Making decisions… What is the current situation? How will the future unfold? Yes NoTina Comes Decision Making and Scenario Planning 3Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  4. 4. How to improve the crystal ball?Each action has consequences Which of them are relevant? How do they evolve? How to compare different consequences? 200 60 people, %, beca because use … …Tina Comes Decision Making and Scenario Planning 4Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  5. 5. Making decisions1. Identify objectives System disaster what would you ideally achieve? • environment2. Describe the system • actors and their decisions what are the constitutent elements? how are they related?3. Derive relevant consequences from the higher- level objective Actions Consequences how to compare consequences? • supply water • number of and food casualties4. Find actions to improve • number of • evacuate the consequences people evacuated • ... what can be done?5. Compare and analyze what to do? improve actions and iterate make decisionTina Comes Decision Making and Scenario Planning 5Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  6. 6. ... but this is difficult in emergencies! Multiple stakeholders and decision makers Heterogeneous information on various aspects of the situation Uncertainty: unforeseen events and reactions Limited time to make a decision and pressure Actors possibly geographically dispersed Bounded availability of experts Risk of information overload and lack of informationTina Comes Decision Making and Scenario Planning 6Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  7. 7. Strategic decisions 60 %1. Multiple goals, diverse actors 200  how to make trade-offs people explicit?  how to build 100 consensus? people2. Uncertainty and complexity  what could the consequences of a decision be? 50 %  what can go wrong?  why?3. How to integrate uncertainty into the decision-making? what is the best option given limited knowledge?Tina Comes Decision Making and Scenario Planning 7Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  8. 8. An approach for scenario-based decisionsCollecting information:a distributed system with heterogeneous expertsHuman and artificial  different skills, backgrounds and knowledgeScenario-Based Multi-Criteria Decision Analysis Orchestrate distributed scenario generation Generate relevant, consistent, plausible and coherent scenarios Use the decision-makers‟ and experts‟ information needs as rationale for information filtering and sharing Provide understandable decision analyses and evaluationsTina Comes Decision Making and Scenario Planning 8Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  9. 9. Challenges1. Improving the crystal ball: objectives and information needs2. How to get relevant information?3. How to combine and process information?4. How to manage the combinatorics?5. Supporting decision makers: how to analyse, interpret and communicate the results?Tina Comes Decision Making and Scenario Planning 9Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  10. 10. More concretely... http://www.bbc.co.uk/news/world-asia-pacific-12149921 http://www.theaustralian.com.au/in-depth/queensland-floodsTina Comes Decision Making and Scenario Planning 10Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  11. 11. Example Situation Flood currently controlled by levee Risk: quick flooding if water rises higherThreat current uncertain situation developments Time 1. Do nothing? What to do? 2. Protect buildings, provide supplies? 3. Evacuation? The Kia Ora Levee http://www.crikey.com.au/2011/02/28/levees- and-the-lack-of-regulation-that-could-cost- millions/Tina Comes Decision Making and Scenario Planning 11Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  12. 12. What is best decision ?5 Groups 1. Residents 2. Local industry and infrastructure providers 3. EM staff (fire fighters, health care, police, ...) 4. Political authorities (responsible to make the decision) 5. ModeratorsYour aim: Establish a consensus about what to do!1. Preparation and analysis of options2. Discussion and consensus building  one member per teamTina Comes Decision Making and Scenario Planning 12Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  13. 13. CHALLENGE #1 Improving the crystal ball: objectives and information needsTina Comes Decision Making and Scenario Planning 13Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  14. 14. Determining possible futures Relevant consequences Situation information What goes here? Ranking of Alternatives alternatives for actionTina Comes Decision Making and Scenario Planning 14Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  15. 15. http://www.theaustralian.com.au/news/nation/queenslands-flood-disaster-a- long-way-from-over-warns-anna-bligh/story-e6frg6nf-1225979264551Tina Comes Decision Making and Scenario Planning 15Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  16. 16. What are the relevant consequences?Discuss in your team:1. From your perspective, what the relevant consequences? health and safety, avoid economic losses, efficiency of operations, ...2. Which of them are the most relevant for you?3. How can the consequences be measured? Use indicators that quantify the consequences, such as “duration of business interruption” for economic losses!Tina Comes Decision Making and Scenario Planning 16Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  17. 17. How are the consequences related? Aim: structured evaluation of a decisions consequences taking into account the decision makers preferences modelling the problem by an attribute tree # people evacuated per day health 1. do nothing # people exposed to flood 2. protection and supplies total performance firefighters [man-h] 3. evacuation effort police [man-h]Tina Comes Decision Making and Scenario Planning 17Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  18. 18. Back to the example In your team, structure the problem by an attribute tree 1. do nothing 2. protection and supplies total performance 3. evacuationTina Comes Decision Making and Scenario Planning 18Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  19. 19. Determining the consequences?Decision tables specify the consequences for all alternatives withrespect to each attribute # people # people firefighters police evacuated exposed [man-h] [man-h] per day to flood 1. do nothing 2. protect 3. evacuate How to fill in the blanks? 1. collect information 2. manage uncertaintyTina Comes Decision Making and Scenario Planning 19Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  20. 20. An example from chemical emergencymanagement # pp unshelt & police [manh] # pp shelt & firefighters losses [k€] alternative economic [manh] exp exp E&S1 15 0 0 247,50 123,75 S1 7 0 0 165,00 82,50 DN 0 0 0 0,00 0,00Tina Comes Decision Making and Scenario Planning 20Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  21. 21. An example from chemical emergencymanagement – determining the basicinformationWhat information is required to determine the attributes? variables indicators variables ATTRIBUTES affected* (GVP/d, affected* (GVP/d, population registry # pp unshelt & exp firefighters [manh] economic losses # pp shelt & exp firms indirectly critical objects infrastructure* transportation infrastructure police [manh] firms directly source term* population alternative presence* leak size* chemical weather* building registry plume [k€] k€) k€) E&S NW none Cl_2 none none 750 0 5 0 0,33 5 0,67 15 0 0 247,5 123,8 1 S1 NW none Cl_2 none none 500 0 5 0 0,33 5 0,67 7 0 0 165 82,50 0 DN NW none Cl_2 none none 0 0 5 0 0,33 5 0,67 0 0 0 0Tina Comes Decision Making and Scenario Planning 21Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  22. 22. CHALLENGE #2 Collecting Information: Getting Experts to CooperateTina Comes Decision Making and Scenario Planning 22Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  23. 23. How to determine a decision’s consequences?Monolithic SystemSeems like a good idea Built exactly to system specification Quick simulation of results Artificial intelligence techniques are mature …However Vendor lock-in Specification changes over time as problem changes Artificial Intelligence techniques are expensive …Tina Comes Decision Making and Scenario Planning 23Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  24. 24. An alternative approachIn your team discuss:1. Which information do you need to determine the best alternative from your perspective?2. Who can provide it?3. How to combine it?Tina Comes Decision Making and Scenario Planning 24Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  25. 25. Using a Hybrid Heterogeneous Distributed SystemNetwork of experts Hybrid: both human and artificial experts Diverse backgrounds, skills and expertise  breaking down complex problems into manageable sub-problemsExperts cooperate… … to determine a set of possible futures: scenarios … via a standardized communication „engine‟Tina Comes Decision Making and Scenario Planning 25Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  26. 26. Cooperating experts? What goes here?Tina Comes Decision Making and Scenario Planning 26Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  27. 27. A distributed problem solving approachCooperation structure Distributed information processing workflow Workflow setup: combined top-down bottom-up approach Based on information need („backwards‟): request for information Based on event („forwards‟): information available  further processingMatching the experts‟ processing capabilities Based on profiles per expert Match based on information types (input & output) expertise (e.g., location, capabilities)Tina Comes Decision Making and Scenario Planning 27Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  28. 28. Orchestrated information processingTina Comes Decision Making and Scenario Planning 28Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  29. 29. Experts in workflow for the chemicalemergency exampleTina Comes Decision Making and Scenario Planning 29Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  30. 30. Another distributed system Summer of extreme weather - sbs.com.au/news http://maps.google.com.au/maps/ms?ie=UTF8&hq=&hnear=Bundarra+New+South+Wales&gl=au&t=h&so urce=embed&oe=UTF8&msa=0&msid=216305641036137584677.000498fa830661a4cbafb .Tina Comes Decision Making and Scenario Planning 30Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  31. 31. Summer of extreme weather - sbs.com.au/news http://maps.google.com.au/maps/ms?ie=UTF8&hq=&hnear=Bundarra+New+South+Wales&gl=au&t=h&so urce=embed&oe=UTF8&msa=0&msid=216305641036137584677.000498fa830661a4cbafb .Tina Comes Decision Making and Scenario Planning 31Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  32. 32. Local informationhttp://www.rockhamptonregion.qld.gov.au/Council_Services/News_and_Announcements/Latest_News/Evacuation_Centre_open_8am_Friday_31_DecemberTina Comes Decision Making and Scenario Planning 32Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  33. 33. Tina Comes Decision Making and Scenario Planning 33Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  34. 34. Trying it outEstablish a rationale for the negotiations referring to the goals andobjectives you identified!- where would you enforce evacuation?- recommend evacuation?- recommend sheltering?- other?Some sources you may find usefulhttp://www.qldreconstruction.org.au/maps/aerial-imaging-and-mapping-pdfshttp://highload.131940.qld.gov.au/#11http://maps.google.com.au/maps/ms?ie=UTF8&hq=&hnear=Bundarra+New+South+Wales&gl=au&t=h&source=embed&oe=UTF8&msa=0&msid=216305641036137584677.000498fa830661a4cbafbTina Comes Decision Making and Scenario Planning 34Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  35. 35. CHALLENGE #3 Keeping track of the futureTina Comes Decision Making and Scenario Planning 35Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  36. 36. Why information is not perfect Uncertainty Ambiguity Incomplete and uncertain information in consequences and evaluation Constraints in Time Constraints resourcesTina Comes Decision Making and Scenario Planning 36Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  37. 37. Robust Decision-Making Aim: Find the alternative that performs satisfactory in many (all) scenarios. ScoreScore Satisfactory threshold Time Time Considering one scenario per Considering multiple scenarios per alternative results in one scoring. alternative results in spread of scoring. Tina Comes Decision Making and Scenario Planning 37 Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  38. 38. Considering several futures… A £ A’ $ B B’ E 1.2 C 2.5 C’ 25 512 E’ D D’Tina Comes Decision Making and Scenario Planning 38Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  39. 39. The flood?Tina Comes Decision Making and Scenario Planning 39Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  40. 40. Media CoverageAt the scene: Nick Bryant BBC News,RockhamptonAlmost completely encircled by muddy floodwaters,Rockhampton risked being entirely cut off if those rose muchfurther, but they peaked slightly lower than the authoritieshad feared, enough to keep the one highway thats open frombeing inundated. Many of the citys low-lying suburbs willremain flooded for more than a week, but a local official saidthe city as a whole had "dodged the bullet".Longer term consequencesNow attention is shifting to the economic http://www.bbc.co.uk/news/world-asia-pacific-12116919impact of the flooding on Australias two most vital sectors, mining and agriculture.Operations at some 40 mines have been interrupted and many of the railway lines thattransport coal to the ports have been severed. Queensland is responsible for more thanhalf of the countrys coal exports. With farms flooded and crops ruined, the price of freshfruit and vegetables is also forecast to rise, by as much as 50%.State Premier Anna Bligh predicted this disaster could have a global impact, partly becauseQueensland supplies half of the worlds coking coal for steel manufacturing. At least onesenior economist here thinks this could be Australias most costly natural disaster, largelybecause of the impact on exports.Tina Comes Decision Making and Scenario Planning 40Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  41. 41. Trying it outRevisit your recommendation and rationale- is it optimal?- is it robust?- which are the most important scenarios you want to use in the discussions? why?Tina Comes Decision Making and Scenario Planning 41Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  42. 42. Managing the experts’ work in distributedreasoning framework Old situation New situation What goes here? Information flowTina Comes Decision Making and Scenario Planning 42Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  43. 43. Keeping track of (partial) scenarios Scenarios capture uncertainty Requirements Consistency and comparability  Not mixing scenario values Coherence:  Keeping track of the scenario constructionTina Comes Decision Making and Scenario Planning 43Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  44. 44. Consistency in the example Combination of information Combination of informationabout independent variables about related variables Changing the workflow mechanisms to … keep track of partial scenarios … correctly merge partial scenariosTina Comes Decision Making and Scenario Planning 44Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  45. 45. An extract from the chemical emergencymanagement example variables indicators variables FOCUS transportation police [manh] infrastructure infrastructure source term* (GVP/d, k€) (GVP/d, k€) # pp shelt & # pp unshelt population firefighters losses [k€] population alternative presence* leak size* affected* economic indirectly weather* affected* chemical registry registry directly building objects critical [manh] plume & exp firms firms exp * E&S1 NW none Cl_2 none none 750 0 5 0 0,33 5 0,67 15 0 0 247,50 123,75 E&S1 NW none Cl_2 none none 750 0 5 0 0,33 5 0,85 18 0 0 247,50 123,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0 0,25 40 0,67 72,00 925,00 4262,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0 0,25 50 0,67 90,00 925,00 4262,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0 0,25 40 0,85 72,00 1375,00 2687,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0 0,25 50 0,85 90,00 1375,00 2687,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0 0,6 40 0,67 72,00 925,00 4262,50 1050,00 525,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0 0,6 50 0,67 90,00 925,00 4262,50 1050,00 525,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0,1 0,6 40 0,85 72,00 1375,00 2687,50 1056,00 528,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 20 0,1 0,6 50 0,85 90,00 1375,00 2687,50 1056,00 528,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0 0,25 48,00 0,67 86,40 925,00 4262,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0 0,25 60,00 0,67 108,00 925,00 4262,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0 0,25 48,00 0,85 86,40 1375,00 2687,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0 0,25 60,00 0,85 108,00 1375,00 2687,50 437,50 218,75 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0 0,6 48,00 0,67 86,40 925,00 4262,50 1050,00 525,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0 0,6 60,00 0,67 108,00 925,00 4262,50 1050,00 525,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0,1 0,6 48,00 0,85 86,40 1375,00 2687,50 1056,00 528,00 E&S1 NW med Cl_2 Big Area-big-1 2500 2 22 0,1 0,6 60,00 0,85 108,00 1375,00 2687,50 1056,00 528,00 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0 0,25 50 0,67 90,00 590,00 3935,00 312,50 156,25 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0 0,25 80 0,67 144,00 590,00 3935,00 312,50 156,25 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0 0,25 50 0,85 90,00 950,00 2675,00 312,50 156,25 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0 0,25 80 0,85 144,00 950,00 2675,00 312,50 156,25 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0 0,6 50 0,67 90,00 590,00 3935,00 750,00 375,00 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0 0,6 80 0,67 144,00 590,00 3935,00 750,00 375,00 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0,1 0,6 50 0,85 90,00 950,00 2675,00 756,00 378,00 E&S1 NW large Cl_2 Big Area-big-2 2000 3 30 0,1 0,6 80 0,85 144,00 950,00 2675,00 756,00 378,00... and this is just a small extract...Tina Comes Decision Making and Scenario Planning 45Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  46. 46. CHALLENGE #4 Handling combinatoricsTina Comes Decision Making and Scenario Planning 46Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  47. 47. Too many possible futures… Given Limited time, effort, available expertise Need for a decision Aim: exploring the space of possible developments Combinatorics… Too many scenarios! What to do?Tina Comes Decision Making and Scenario Planning 47Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  48. 48. Scenario Management During the construction Selection of the most relevant partial scenarios Pruning of invalid scenarios Update to take into account relevant new information Evaluation: Partial scenario Selection of the most relevant scenarios Selected partial Aggregation of results scenario Updated partial scenarioTina Comes Decision Making and Scenario Planning 48Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  49. 49. Which scenarios are the most relevant? Most scenario similarity measures: distance of the variables‟ values Our aim: Explore the space of evaluations  Making risks and chances transparent  Robustness Definition of Scenario classes  Based on the similarity of the evaluation  Selection of a representative per classTina Comes Decision Making and Scenario Planning 49Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  50. 50. Impact on exploration of scenario space exploitingthe network structures 1 0.9 UPDATED 0.8 0.7 ORIG Evaluation 0.6 SEL 0.5 0.4 0.3 0.2 0.1 0 ScenarioTina Comes Decision Making and Scenario Planning 50Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  51. 51. Scenario Updates: Efficiency 400 Upper Bound of Duration [min] 350 Duration of update from indicator variables to FOCUS 300 250 Duration of update to indicator variables 200 150 100 50 0 Complete update Partial update all Partial update of scenarios selected Approach to updateTina Comes Decision Making and Scenario Planning 51Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  52. 52. How a distributed system can work in chemicalemergencies Video available on: http://www.pdc.dk/diadem/Video/DiademVideo.wmvTina Comes Decision Making and Scenario Planning 52Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  53. 53. CHALLENGE #5 Supporting decision makersTina Comes Decision Making and Scenario Planning 53Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  54. 54. How to develop good alternatives? MCDA: workshops serve Define the - for the identification of Recommendation Problem decision criteria and feasible countermeasures Sensitivity Analysis n Con Identify the ctio - as exercises Attributes clus odu ing her ion Intr Pla - for the identification of Mea nning su Gat ics top be t res to responsibilities and authorities Choose an ake n Se le to implement a rapid response Alternative c to tin pi g Specify Performance top g the ndlin c a ic Measures Ha How to support decision makers in building better Weight Criteria Identify the alternatives and establish Analyse the Alternatives consensus in very Alternatives uncertain situations?Tina Comes Decision Making and Scenario Planning 54Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  55. 55. How to handle trade-offs?Preference models represent the preferences and value judgements of adecision maker by1. A model that scores each alternative against each individual attribute  concerns all attributes2. A model that compares the relative importance among the criteria to obtain a ranking of alternatives a. Elicitation of the relative importance (weights) of the criteria b. Aggregation  concerns the complete attribute treeTina Comes Decision Making and Scenario Planning 55Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  56. 56. Back to the example attribute trees How to compare the attributes? 1. do nothing 2. protection and supplies total performance 3. evacuationTina Comes Decision Making and Scenario Planning 56Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  57. 57. Some technical details: Value functions allow to scoreeach alternative against each individual attributeScores si(a) of the alternatives are measured in different units for thedifferent attributesto make comparisons, map these scores to a scale ranging from 0 to 1(where the “worst” and “best” possible outcomes correspond to 0 and 1respectively) by defining value functions si a : score of alternative a relative to attribute i vi vi si a : value of the score of alternative a relative to attribute i si a min si a # people protected a , if max si a highest value max si a min si a a a a vi max si a si a a , if max si a lowest value max si a min si a a a a work effort (# workers)Tina Comes Decision Making and Scenario Planning 57Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  58. 58. Weights – Inter-criteria preferencesDifferent weighting procedures The simplest way is the DIRECT weighting In the SWING procedure, 100 points are first given to the most important attribute; then, less points are given to the other attributes depending on the relative importance of their ranges The SMART method is similar, but the procedure starts from the least important attribute (assigning 10 points to it) keeping it as the reference In SMARTER, the weights are elicited directly from the ranking of the alternatives In AHP, the weights are determined by pairwise comparisonsTina Comes Decision Making and Scenario Planning 58Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  59. 59. Trying it out...Go back to the attribute tree and the rationales you have developed.- which are the most important criteria for you?- can you establish clear preferences within your group (for weights and value functions)?Tina Comes Decision Making and Scenario Planning 59Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  60. 60. Scenario selection: Exemplary results Selected sources of uncertainty: success of chlorine transfer residual amount of chlorine in tank weather Evaluation of Scenarios 1 Health Effort 0.9 Society 0.8 results for best and worst Evaluation R(s) 0.7 0.6 scenarios Evaluation R(s) 0.5 0.4 0.3 0.2 0.1 0 E S N E S N E S N E S N E S N E S N E S N E S N E S N E S N Scenarios for Alternatives Evacuation (E), Sheltering (S) and Do nothing (N) Scenarios for Alternatives Evacuation (E), Sheltering (S) and Do Nothing (N)Tina Comes Decision Making and Scenario Planning 60Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  61. 61. Aggregation of results:how important is each scenario?Definition of weights – but how? direct elicitation from the decision-makersAccording to the Evaluation Goal Attainment  Trying to satisfice overall or partial goals (Simon, 1979)  Deviation from equal weighting if these goals are not attained: penalty functions According to risk aversion  Risk aversion: relative importance of scenarios evaluated worst/best (Yager, 2008)  Determination of weights according to the scenarios„ rankingTina Comes Decision Making and Scenario Planning 61Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  62. 62. Example: Results for varying levels of risk aversion 1 1 Evacuation 0.9 Sheltering 0.9 Do Nothing 0.8 Aggregated weights 0.8 0.7 aggregated weight of worst evaluated scenarios 0.6 aggregated weight of Result(alternative) 0.7 best evaluated scenarios 0.5 0.4 0.6 0.3 0.5 0.2 0.1 0.4 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.3 Risk level 0.2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Risk levelTina Comes Decision Making and Scenario Planning 62Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  63. 63. Interpreting the results: scenario reliabilityNumber of scenarios increases with growing uncertainty risk of overemphasizing some scenarios‟ results for structural reasonsScenario Reliability Modelling the relative uncertainty of scenarios: uncertainty of the situation: comparison to other scenarios uncertainty of the specific scenario preferences of the decision makers easily manageable measure enables decision-makers to adapt scenario weights and overcome cognitive biasesTina Comes Decision Making and Scenario Planning 63Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  64. 64. How to make alternatives better1. How is the quality of an alternative measured? MCDA!2. What can go well and what can go wrong? SBR!An iterative approach1. Identification of key weaknesses per alternative2. Identification of better alternatives to address these weaknessesAnalysis: how can these alternatives be combined?So, all information is there. But... ... large numbers of scenarios and results ... visualisations not easy to interpret need for a clear and transparent explanation of resultsTina Comes Decision Making and Scenario Planning 64Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  65. 65. Making sense of what you seeTina Comes Decision Making and Scenario Planning 65Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  66. 66. Communicating decisions under uncertainty Evaluation of Scenarios 1 Health Effort 0.9 Society 0.8 0.7 0.6 Evaluation R(s) 0.5 0.4 0.3 0.2 0.1 0 E S N E S N E S N E S N E S N E S N E S N E S N E S N E S N Scenarios for Alternatives Evacuation (E), Sheltering (S) and Do nothing (N)Tina Comes Decision Making and Scenario Planning 66Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  67. 67. Generation of natural language reports1. Content determinationInformation about what? Type of report and variables: alternatives, outcomes, drivers, ... informationQuestions that should be addressed? requirements relations: causes and effects, better or worse, ...2. Discourse planning3. Sentence generationTina Comes Decision Making and Scenario Planning 67Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  68. 68. Generation of natural language reports1. Content determination• variables Type of report and• relations information requirements2. Discourse PlanningWhat can be said about the entities and their relations? determine types of individual messages ArgumentationHow to combine the messages into an argumentation? relate and cluster messages into a tree structure3. Sentence generationTina Comes Decision Making and Scenario Planning 68Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  69. 69. Generation of natural language reports1. Content determination• variables Type of report and• relations information requirements2. Discourse Planning• types of individual messages Argumentation• tree structure structure3. Sentence generationHow to express the message? choose of adequate text patterns Template SystemWhat is the argument for this case?completion of statementsTina Comes Decision Making and Scenario Planning 69Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  70. 70. From numbers to verbal expressions:Semantic quantifiersAim: describe the quality of a decision“substantially better”, “slightly worse”, ...Alternative <name of alternative> performs <semantic quantifier> on<objective> in the context of all available scenarios.A relative approach1. set of evaluated scenarios and relevant objectives2. determine mean μ and standard deviation3. set SQs Alternative evacuation performs very poor on effort in the context of allavailable scenarios.A benchmark approach: goal programming and satisfaction levelsAlternative evacuation has an acceptable performance with respect tohealth in most scenarios.Tina Comes Decision Making and Scenario Planning 70Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  71. 71. Key weaknesses1. What do the worst scenarios for an alternative have in common? statistical approach: worst % for each alternative benchmark approach: scenarios that violate threshold  identify variables var1, ..., varn and their values Alternative <name of alternative> performs <semantic quantifier> on <objective> for all scenarios that assume <value of var1> for <var1>,..., <value of varn> for <varn>.2. How do other alternatives perform for the same / similar scenarios?3. Identify better alternatives and describe significance in an SQ Alternative <name of alternative2> performs <semantic quantifier> on <objective> than <name of alternative> for the identified scenarios.Tina Comes Decision Making and Scenario Planning 71Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  72. 72. Finally...Prepare for thediscussion, collectthe material youneed and choose therepresentative...... and then, find asolution:which strategicmeasures shouldbe implementedand where?Tina Comes Decision Making and Scenario Planning 72Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  73. 73. REFLECTIONS AND CONCLUSIONSTina Comes Decision Making and Scenario Planning 73Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  74. 74. ConclusionIntegrated Scenario-Based MCDA Distributed processing of relevant information Consideration of interdependencies Formalization using set and graph theory Ensuring comparability Scenario management: updating, selection, pruning Respecting constraints and requirements in emergency managementDecentralised vs. centralised: Orchestrating emergence Decentralised experts involved in workflow Decision-centric management with overviewTina Comes Decision Making and Scenario Planning 74Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  75. 75. Reflections1. What were the main challenges in your team? in the discussion?2. Social media applications?Tina Comes Decision Making and Scenario Planning 75Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  76. 76. Thank you!ContactTina Comescomes@kit.edu Questions?Tina Comes Decision Making and Scenario Planning 76Institute for Industrial Production (IIP) ISCRAM Summer School 2012
  77. 77. References Comes, T., Wijngaards, N. & Schultmann, F. (2012): Efficient Scenarios Updating in Emergency Management. 9th International Conference on Information Systems for Crisis Response and Management Comes, T., Wijngaards, N., Maule, J., Allen, D. & Schultmann, F. (2012): Scenario Reliability Assessment to Support Decision Makers in Situations of Severe Uncertainty. 2012 IEEE Conference on Cognitive Methods in Situation Awareness and Decision Support Comes, T., Hiete, M., Wijngaards, N. & Schultmann, F. (2011): Decision Maps: A framework for multi-criteria decision support under severe uncertainty. Decision Support Systems, 52(1), 108-118. Comes, T., Conrado, C., Hiete, M., Wijngaards, N. & Schultmann, F. (2011): A distributed scenario-based decision support system for robust decision-making in complex situations. International Journal of Information Systems for Crisis Response and Management, 3(4), 16-35. Simon, H. (1979): Rational Decision Making in Business Organizations, The American Economic Review, 69(4), 493-513. Ronald R. Yager, “Using trapezoids for representing granular objects: Applications to learning and OWA aggregation,” Information Sciences 178(2), 363-380.Tina Comes Decision Making and Scenario Planning 77Institute for Industrial Production (IIP) ISCRAM Summer School 2012

×