Behavioral Aspects in Decision Analysis Raimo P Hmlinen
Behavioral Aspects in Decision Analysis Raimo P. Hämäläinen Systems Analysis Laboratory Aalto University, School of Science Finland SA” L Systems Analysis Laboratory
Why should we consider behavioral aspects? The models in Decision Analysis relate to the preferences and utilities of the decision maker Models are free of behavioral effects As soon as we start using models in practice the behavioral elements are present What is the human impact on model based decision and systems analysis?
Howard Raiffa A pioneer in decision analysis When teaching statistics, Raiffa learned that more is needed to support decision making in practice This led to Decision Analysis (1968) • How to structure decision problems • Use of subjective probabilities • Decision trees Economic Expansion Expand Economic Contraction Economic Expansion Do nothing Economic Contraction +$5, 000 -$3, 000 $2, 000 -$1, 000
Decision Analysis: A Personal Account of How It Got Started and Evolved (Howard Raiffa, Operations Research 2002) Acknowledged behavioral aspects already early • 60 s challenge: How to obtain reliable judgments from experts? • Research on improving elicitation procedures and framing Behavioral perspective is essential in Decision Analysis
Ralph Keeney and Howard Raiffa (1976) ”These psychological insights will undoubtedly help analysts design better assessment protocols in the future. ” Early suggestion: Ask elicitation questions in multiple ways • Check consistency and discuss with DM
Descriptive psychological research in judgment and decision making Axioms of rationality not followed Perceived value Decision behavior Value function Gains Cognitive and motivational biases • Anchoring, scope insensitivity, confirmation bias etc. Heuristics Systems 1 and 2 thinking
Human behavior drives the Decision Analysis process Social interaction: Engagement, dialogue, communication Behavioral effects are present in all the steps • Problem framing • Choice of criteria • Uncertainty modelling, etc. Biases influence elicitation of subjective values and parameter estimates • Weighting • Estimation of consequences and probabilities
Textbooks discuss biases, decision traps and how to deal with them Decision Analysis and Behavioral Research (Von Winterfeldt and Edwards 1986) Making Hard Decisions (Clemen and Reilly 1999, 2014) Structured Decision Making (Gregory et al. 2012) Montibeller and von Winterfeldt (2015) review: • 175 references to papers related to biases in DA • 30 biases and ideas for debiasing There is still a lot to be done!
Examples of biases in the DA process Montibeller and von Winterfeldt (2015). Bias Explanation Debiasing Anchoring A numerical value is based on an initial value (anchor), which is then insufficiently adjusted to provide the final answer. Avoid anchors. Use different experts who use different anchors Gain-loss bias Descriptions of a choice and its outcomes either as gains or as losses and may lead to different answers Clearly identify the status quo. Myopic problem representation Oversimplified problem representation is adopted Explicitly encourage to think based on an incomplete mental model of the about more objectives, new decision problem. alternatives. Splitting biases The way the objectives are grouped in a value Use hierarchical estimation of tree affects their weights or the way a fault tree is weights or probabilities. pruned affects the probabilities placed. Proxy bias Proxy attributes receive larger weights than the respective fundamental objectives. Avoid proxy attributes Range insensitivity bias Weights of objectives are not properly adjusted to changes in the range of attributes. Make attribute ranges explicit. Use multiple elicitation procedures.
Splitting bias Higher weight if attribute is split into more detailed lower 1/6 Variation in water level attributes 1/6 2/6 2/6 1/6 Recreation Nature 1/6 Economy 1/6 Recreational fishing Reproduction of fish Dense bay vegetation Shoreline vegetation Economy Occurs e. g. when people give equal weights to all attributes
No splitting bias vs. systematic bias Systematic bias The effect of splitting the environmental attribute
Splitting bias is difficult to eliminate Students with debiasing guidance: no splitting bias Stakeholders: systematic bias, guidance did not help Stakeholders Hämäläinen and Alaja (2008)
Ideas to deal with biases still need to be tested in practice Guidance • Did not help to avoid the splitting bias (Hämäläinen and Alaja 2008) • Pre-exposure to attribute levels can reduce decision context related biases (Carlson and Bond 2006) Estimate bias coefficients and calibrate judgments • To reduce scale compatibility bias (Anderson and Hobbs 2002) • To reduce loss aversion bias (Bleichrodt et al. 2001) Design procedures so that the effects of biases cancel out (Lahtinen and Hämäläinen 2015) • Does not require additional training or calibration
Decision Analysis for the general public: Smart Choices (1999) To ”bridge the gap between how people actually do make decisions… and how they should make decisions” • Pr. OACT framework Problem: Define it Objectives: Identify and clarify them Alternatives: Develop good alternatives Consequences: Describe alternatives Trade-offs: Make tough compromises • Even Swaps method • Avoiding psychological traps discussed
The Even Swaps process Eliminate attributes by irrelevance and alternatives by dominance until one alternative remains • Even Swap: Alternative replaced by a preferentially equivalent one that differs in two attributes • The process can be technically challenging • Smart Swaps software helps to identify efficient swaps (Mustajoki and Hämäläinen 2007) 25 78 Practically dominated by Montana An even swap Commute time removed as irrelevant Dominated by Lombard
Behavioral phenomena in Even Swaps The process allows different paths: Sequences of swaps taken Accumulation of biases is possible and can depend on the path followed Path dependence when different paths lead to different outcomes DM chooses A? DM chooses B?
The measuring stick effect can create path dependence (Lahtinen and Hämäläinen 2015) Apartment related case: Four alternatives, pricing path and two reference paths Pricing path Make all attributes but cost irrelevant. Use money (cost) as measuring stick in all swaps. Explanation Money as measuring stick in all swaps: Alternatives with low cost are favored Debiasing Avoid using measuring stick in which alternatives differ much
Behavioral challenges in group decision making Risk of biases is high Strategic behavior • Stakeholders can emphasize factors that are important to them • Strategic representation of preferences This is the right model Groupthink (Janis 1972) Yes Yes Facilitator skills important: Understand manage behavior in the social system Yes Yes
Decision Analysis used to solve conflicts Established by president Martti Ahtisaari, Nobel Peace Price 2008 Policy action prioritization in Egypt (2012) • DA workshop facilitated by CMI (Finland) • Overall goal: Mitigation of violent conflicts • Participants: Government officials DA tools also used in regional conflicts in the Middle - East
Howard Raiffa A pioneer in negotiation science Active in the negotiation on IIASA in the cold war period Negotiation course at Harvard in 1970 s • Demonstrated that student experiments can be valuable for developing theory ”Collectively we could test what worked…and …discuss whether our heuristic insights would be applicable in the real world” (Raiffa 2002) ⇒ The Art and Science of Negotiation (1982) • Analytical models and Behavioural insights
Negotiation in conflict resolution Camp David negotiations (1970 s) • Between Egypt (Anwar El Sadat) and Israel (Menachem Begin) • Mediated by the US (Jimmy Carter) • A sequence of single negotiation texts with joint gains (Roger Fischer) Begin, Carter and El Sadat
The ART of negotiation emphasizes behavioral aspects Reservation value Egypt Joint gains process: Searching for solutions where each player gains and need not make trade-offs Behavioral challenges: • Effect of the starting point • Strategic misrepresentation of preferences Reservation value Israel The idea of post-settlements: • Negotiations after initial settlement • Mediator suggests a post-settlement with joint gains
Behavioral perspective in modeling Behavioral Decision Analysis can inform general modeling studies What is the human impact on the modeling process? Modeler biases, communication, group interaction etc. Man and the hammer – syndrom Behavioral Operational Research is an emerging area (Hämäläinen et al. 2013) Upcoming special issue in (Franco and Hämäläinen 2015)
Emotions are needed in decision making People with damaged emotion related brain areas have difficulties in decision making (Damasio 1994) Negative emotions can lead to avoiding decision making, sticking with status quo (Luce 1998) Frustration and anger increases risk taking behavior (Leith and Baumeister 1996)
Future? Decision neuroscience Does it help to understand the neural processes? Different areas of the brain involved in different tasks Striatum related to reward evaluation • Striatum is a subcortical part • In the brain area that was developed very early in human evolution Striatum activation (Niv and Montague 2009) Insula related to risk evaluation • Insula is located on brain cortex • Developed later in evolution Insula activation (Preusschoff et al. 2008)
Summary Howard Raiffa acknowledged the importance of behavioral aspects early on Today: Wide awareness of biases within the decision analysis community • Challenge to spread awareness to practitioners in different areas, e. g. in environmental management and policy decision making • Debiasing methods need to be tested and taken into practice Behavioral perspective is important in the entire modelling and systems analysis process
References Bleichrodt, H, J. L. Pinto, and P. P. Wakker. Making descriptive use of prospect theory to improve the prescriptive use of expected utility. Management science 47(11), 2001: 1498 -1514. Carlson, K. A. , and S. D. Bond. Improving preference assessment: Limiting the effect of context through pre-exposure to attribute levels. Management Science 52(3), 2006: 410 -421. Damasio, A. R. Descartes' error and the future of human life. Scientific American 271(4), 1994. Delquié, P. Inconsistent trade-offs between attributes: New evidence in preference assessment biases. Management Science 39(11), 1993: 1382 -1395. Franco, L. A. and R. P. Hämäläinen. Behavioural operational research: Returning to the roots of the OR profession. European Journal of Operational Research, In Press, 2015. http: //www. sciencedirect. com/science/article/pii/S 0377221715009601 Hammond, J. S. , R. L. Keeney, and H. Raiffa. Smart choices: A practical guide to making better decisions. Vol. 226. Harvard Business Press, 1999. Hämäläinen R. P. , J. Luoma and E. Saarinen: On the Importance of Behavioral Operational Research: The Case of Understanding and Communicating about Dynamic Systems European Journal of Operational Research, 228 (3), (2013): 623 -634. Hämäläinen R. P. : Behavioural issues in environmental modelling - the missing perspective. Environmental Modelling & Software, 73, 2015: 244 -253. Hämäläinen R. P. and T. J. Lahtinen: Path Dependence in Operational Research - How the Modeling Process Can Influence the Results. 2015. Manuscript: http: //sal. aalto. fi/publications/pdf-files/mham 15 c. pdf Hämäläinen, R. P. , S. Alaja: The Threat of Weighting Biases in Environmental Decision Analysis Ecological Economics, 68, 2008: 556 -569. Janis, I. L. Groupthink: Psychological studies of policy decisions and fiascoes. Boston: Houghton Mifflin. 1982. Keeney, R. L. , and H. Raiffa. Decisions with multiple objectives: preferences and value trade-offs. Cambridge university press, 1993.
References Lahtinen T. J. and R. P. Hämäläinen: Path Dependence and Biases in the Even Swaps Decision Analysis Method. European Journal of Operational Research, 2015, In Press. http: //sal. aalto. fi/publications/pdf-files/mlah 15. pdf Leith, K. P. , and R. F. Baumeister. Why do bad moods increase self-defeating behavior? Emotion, risk tasking, and selfregulation. Journal of personality and social psychology 71(6), 1996. Liesiö J. , P. Mild, and A. Salo: Robust Portfolio Modeling with Incomplete Cost Information and Project Interdependencies European Journal of Operational Research, 190, 679 -695. Luce, M. F. Choosing to avoid: Coping with negatively emotion-laden consumer decisions. Journal of consumer research 24(4), 1998: 409 -433 Montibeller, G. , and D. Winterfeldt. Cognitive and Motivational Biases in Decision and Risk Analysis, 2015. Mustajoki J. and R. P. Hämäläinen: Web-HIPRE: Global decision support by value tree and AHP analysis INFOR, 38, no. 3, 2000, 208 -220 Mustajoki J. and R. P. Hämäläinen: Smart-Swaps - A decision support system for multicriteria decision analysis with the even swaps method Decision Support Systems, Vol. 44, No. 1, 2007, pp. 313 -325. http: //smart-swaps. aalto. fi/ Niv, Y. , and P. R. Montague. Theoretical and empirical studies of learning. Neuroeconomics: Decision making and the brain, 2008: 329 -50. Preuschoff, K. , S. R. Quartz, and P. Bossaerts. Human insula activation reflects risk prediction errors as well as risk. The Journal of neuroscience 28(11), 2008: 2745 -2752. Raiffa, H. Decision analysis: a personal account of how it got started and evolved. Operations Research 50(1), 2002: 179 -185. Raiffa, H. The art and science of negotiation. Harvard University Press, 1982. Raiffa, H. Decision analysis: introductory lectures on choices under uncertainty. 1968. Raiffa, H. Post-settlements. Negotiation Journal 1(1), 1985: 9 -12. Rilling, J. K. , and A. G. Sanfey. The neuroscience of social decision-making. Annual review of psychology 62 (2011): 2348.
- Slides: 28