INFORMS/CORS Montreal 1998 Sponsor: Decision Analysis Society

MA23.1 Probability Forecasting: The Legacy of Allan Murphy Robert L. Winkler --- Duke Univ., Fuqua Sch. of Bus., Durham, NC 27708-0120, (rwinkler@mail.duke.edu)

Allan Murphy was instrumental in moving forward probability forecasting, forecast verification and the use and value of forecasts. His interdisciplinary contributions spanned conceptual issues, methodology development, empirical studies, behavioral experiments, development of decision-making models and applications. We highlight his work and discuss important current research issues in probability forecasting.

MC22.1 Binary Probabilities & the Linear Signed Order Polytope Peter C. Fishburn --- AT&T Bell Labs., Rm. 2C 354, 180 Park Ave., Florham Park, NJ 07932 , (fish@research.att.com)

Signed orders are orderings of items and their anti-items that reflect relative likes and dislikes for the items. We consider relationships between binary choice probabilities on the set of items and anti-items, and probability distributions on the set of linear signed orders.

MC22.2 Classes of Binary Relations as Media & Submedia Jean-Claude Falmagne --- Univ. of CA, MBS Sch. of Sciences, Irvine, CA 92697 , (jcf@aris.ss.uci.edu)

Subjects are often asked to produce repeatedly binary relations (semiorders, quasi orders, etc.) representing judgments concerning commodities or individuals. A particular semigroup of transformations has been proposed by Falmagne as a tool for analyzing this type of temporal evolution. We pursue the study of this semigroup and review some applications.

MC22.3 Cumulative Discrete Choice Amit Pazgal, Itzhak Gilboa --- Washington Univ., One Brookings Dr., CB 1133, St. Louis, MO 63130 , (pazgal@wuolin.wustl.edu)

We present a discrete choice model in which consumer's impression of each alternative is based on her memory of past experiences, and is stochastically updated after consumption. Furthermore, the consumer may sometimes be 'dormant,' selecting the previously chosen option out of inertia. It is proven that both 'active' and 'dormant' cases lead to convergence of the choice frequencies with probability 1.

MC22.4 Random Utility Models & Social Welfare Michel Regenwetter --- Duke Univ., Fuqua Sch. of Bus., Durham, NC 27708-0120, (regenwet@mail.duke.edu)

In subset voting respondents pick a subset containing any number of alternatives from a fixed set. We merge 'approval voting' and the 'single transferable vote' from political science, the 'weak utility model' from mathematical psychology and 'social welfare orderings, Borda and Condorcet winners' from social choice theory.

MC23.1 Framing & Assessing Models of the Needs of Computer Users Eric Horvitz --- Decision Theory Group 9S/1, Microsoft Research, One Microsoft Way, Redmond, WA 98025 , (horvitz@microsoft.com)

The Lumiere project centers on the construction of Bayesian models of the goals and needs of users of computers. Lumiere prototypes served as the basis for the Office Assistant in the Microsoft Office '97 suite. I discuss the framing of distinctions and relationships and the assessment of probabilities in the Lumiere models.

MC23.2 Experiences with Probability Elicitation Robert F. Bordley --- General Motors, R&D Ctr., Bld. 30500 Mound Rd., Warren, MI 48090-9055, (robert_bordley@notes.gmr.com)

Much of the success of DA arises from the logical structuring it brings to a problem. However, in some contexts, quantifying subjective beliefs can also be of significant value. We focus on practical experiences associated with assessing subjective probabilities.

MC23.3 Imputing Uncertainty Distributions from Expert Judgments Stephen C. Hora --- Univ. of HI, 200 W Kawili St., Hilo, HI 96720-4091, (shora@pahuleka.uhh.hawaii.edu)

Probability elicitation should be conducted on values that are conceptually observable. When parameters are not observable, indirect methods of probability elicitation may be used. Some methods are discussed that have been used to quantify nuclear accident models.

MC23.4 Expert Assessment of Flammable Gas Risks Thomas Eppel, Detlof von Winterfeldt --- Decision Insights Inc., 2062 Business Center Dr., Ste. 110, Irvine, CA 92612 , (TomEppel@aol.com)

Some high-level nuclear waste tanks at DOE's Hanford site have the potential for gas releases and possible detonations or burns. We summarize the methodology and results of an expert elicitation about these risks and demonstrate the benefits of using specialized software in the elicitation process.

MD22.1 Nash Equilibria Are Superficial Correlated Equilibria Robert F. Nau, Sabrina Gomez, Pierre Hansen --- Duke Univ., Fuqua Sch. of Bus., Durham, NC 27708-0120, (rnau@mail.duke.edu)

It is well known that the set of correlated equilibrium distributions of a noncooperative game is a convex polyhedron that includes the Nash equilibria. We prove a remarkably simple yet surprising result: that the Nash equilibria must all lie on the surface of the correlated equilibrium polyhedron.

MD22.2 Nash & Correlated Equilibria of 2- & 3-Persons Games: An Empirical Study Sabrina Gomez, Pierre Hansen, Brigitte Jaumard, Robert F. Nau --- GERAD, , , ()

The sets of all Nash, correlated, Pareto optimal correlated and coalition proof correlated equilibria of bimatrix and of 3-persons games are studied. It is known that solutions in the last set can give substantially larger payoffs than Nash equilibria.

MD22.3 An Algorithm for All Extreme Nash Equilibria of Bimatrix Games Charles Audet, Pierre Hansen, Brigitte Jaumard, Gilles Savard --- GERAD & HEC, Polytech. of Montreal, 3000 ch. Cote-Ste-Catherine, Montreal, Quebec, H3T 2A7 , Canada (charlesa@crt.umontreal.ca)

The extreme equilibria of bimatrix games are enumerated in finite time by a branching scheme that exploits the KKT conditions of 2 pairs of parameterized LP s.

MD22.4 Correlated Equilibrium & Unilaterally Competitive Games Olivier De Wolf --- Univ. Catholique de Louvain, 25 Cours du Bia Bouquet, Louvain-la Neuve, 1348 , Belgium (dewolf@core.ucl.ac.be)

We propose to investigate some properties of correlated equilibria in some class of games commonly qualified as competitive. In particular, we show that in these kinds of games (including the n person case), there generally is no incentive for the players to correlate their strategies.

MD22.5 Extreme Nash Equilibria of Bimatrix Games Are Extreme Correlated Equilibria Pierre Hansen, Sabrina Gomez --- HEC, 3000 Ch Cote-Ste-Catherine, Montreal, Quebec, H3T 2A7 , Canada (pierreh@crt.umontreal.ca)

Polyhydral combinatorics are used to prove that in bimatrix games all extreme Nash equilibria are extreme correlated equilibria.

MD23.1 A Simple Approximation for Probabilistic Dependence Among Binary Events Donald L. Keefer --- AZ State Univ., Dept. of Mgmt., Box 4006, Tempe, AZ 85287-4006, (don.keefer@asu.edu)

We propose a simple model for approximating probabilistic dependence among binary success/failure events. This model requires assessment of just one conditional probability in addition to the n marginals. Extensive numerical tests indicate this model produces much more accurate joint probabilities, expected values and certainty equivalents than widely used approximations.

MD23.2 Market-Based Belief Aggregation David M. Pennock, Michael P. Wellman --- Univ. of MI, AI Lab., 1101 Beal Ave., Ann Arbor, MI 48109-2110, (dpennock@umich.edu)

We show how belief aggregation (combining the probabilistic beliefs of the members of a group) can be facilitated through a securities market. Each participant's investment decisions are individually rational; the market's resulting equilibrium represents the group's consensus belief. Similar market infrastructures may aid group decision making and coordination under uncertainty.

MD23.3 Bayesian Network Models of Portfolio Risk Catherine Shenoy, Prakash P Shenoy --- School of Bus., Summerfield Hall, Lawrence, KS 66045-2003, (cshenoy@ukans.edu)

We investigate the use of Bayesian networks for modeling portfolio risk, especially financial portfolios that are not well diversified. From the Bayesian network model we obtain a probability distribution of the portfolio return. We compare the variance of this distribution with the variance calculated from a traditional multi-factor model.

ME22.1 Expected-Utility Preference Reversals in Information Acquisition Gordon B. Hazen, Jay Sounderpandian --- Northwestern Univ., IE/MS Dept., MEAS, Evanston, IL 60208-3119, ()

Suppose you must choose between 2 pieces of information A and B. In the absence of cost, you would prefer to obtain A rather than B, and in fact would be willing to take more risk to obtain A than B. Nevertheless, you would pay more money for B than for A. Are your preferences consistent with expected utility? The answer is yes, they may very well be.

ME22.2 Decision Making & Participation Style Mehdi Mostaghimi, David K. Kreutter --- Pfizer Central Research, Portfolio Analysis Group, Groton, CT 06340 , (mehdi_mostaghimi@groton.pfizer.com)

The decision reached by a decision maker after consulting experts is influenced by his/her participatory style in the decision process. We use a Bayesian estimation method to show that the decision maker's optimal decision varies with his/her participatory style. Numerical examples are presented.

ME22.3 The Sensitivity of Value of Information to Analysis Assumptions Maxine E. Dakins --- Univ. of ID, 1776 Science Cntr. Dr., Idaho Falls, ID 83405 , (medakins@if.uidaho.edu)

We examine the sensitivity of the expected value of sample information in an environmental remediation decision problem to changes in important assumptions of the analysis including the structure of the model, the form and parameters of the loss function and the form and symmetry of the likelihood function.

ME22.4 Biases in the Estimation of the Expected Value of Information Charles D. Linville --- , 2917 Keswick Rd., Baltimore, MD 21211 , (clinville@worldnet.att.net)

We examine 2 mechanisms by which estimates of the expected value of information (VOI) can be biased. Expert overconfidence can lead both to underestimates and overestimates of VOI. Also, simulation variance can lead to positive and negative biases in estimates of VOI based on Monte Carlo simulation.

ME23.1 Discovering Options James E. Smith, Gardner Walkup --- Duke Univ., Box 90120, Durham, NC 27708-0120, (jes9@mail.duke.edu)

While the DA process frequently generates creative alternatives, the process often misses potentially valuable options associated with a project. We describe methods that we have found useful for thinking about project dynamics and discovering embedded options. We illustrate the ideas with examples drawn from the oil and gas industry.

ME23.2 Creating Alternatives When Tradeoffs Are Taboo Robin Gregory --- Decision & Value Scope Res., 1124 West 19th St., N Vancouver, BC, V7P 1Z9 , Canada ()

We describe recent work with private sector and government clients who are reluctant to acknowledge value tradeoffs made as part of both routine and strategic business decisions. We discuss approaches used to help identify these multiple objectives and assist in defining acceptable project or program alternatives.

ME23.3 Creating Customer Value by Creating Alternatives Ralph L. Keeney --- Univ. of Southern CA, 101 Lombard St., Ste 704W, San Francisco, CA 94111 , ()

A general model of individual customer values is presented. Assessments from representative customers indicate the distribution of specific customer values. Proposed new products are evaluated against the set of existing products to indicate the increase in value for different customers and the market share for the new product.

TA23.1 Benthamite Utility for Decision Making Rakesh K. Sarin, Peter P. Wakker --- UCLA, Anderson Sch. of Mgmt., 110 Westwood Plz. Box 951481, Los Angeles, CA 90095-1481, (rsarin@anderson.ucla.edu)

Modern decision theory is based exclusively on observable preferences with hedonistic and introspective aspects of utility being irrelevant for the analysis of decisions. We argue that experienced utility provides insight into the formation of preferences which determine decision utility and that, in some situations, experienced utility may serve as a better guide to decisions.

TA23.2 Decision Processes that Maximize Experienced Utility Joel Huber, Baba Shiv --- Duke Univ., Fuqua School of Bus., Durham, NC 27708-0120, (jch8@mail.duke.edu)

We define maximizing experienced utility as optimizing how good one expects to feel about the outcomes of one's decisions. We examine the effectiveness of various kinds of processing strategies in achieving this goal, summarize biases in each and identify the contexts in which they are most appropriate.

TA23.3 Sources of Value: The Links Between Prescriptive Negotiating Techniques & Experienced Utility Susan E. Brodt, Robert T. Clemen --- Duke Univ., Fuqua Sch. of Bus., Durham, NC 27708-0120, (brodt@mail.duke.edu)

Using Brodt & Clemen's (1997) framework for thinking about a stream of value associated with negotiation, we show how to define and assess the effectiveness of prescriptive negotiation techniques, especially by identifying the psychological processes that link elements of a technique to assessments of experienced utility.

TB23.1 Predicting Choices from Emotions Barbara Mellers --- OH St. Univ., Psych. Dept., 1885 Neil Ave., Columbus, OH 43210 , (mellers.1@osu.edu)

People often imagine how they will feel about the outcomes of decisions and those imagined feelings serve as guides to choice. We provide an account of experienced emotions called decision affect theory. Despite these differences between emotional experiences and utilities, there is considerable overlap between maximizing subjective expected pleasure and subjective expected utility.

TB23.2 Quality Adjusted Life Years Utility Assessment Under Non-EU Assumptions John Miyamoto --- Univ. of WA, Box 351525, Dept. of Psych., Seattle, WA 98195 , (jmiyamot@u.washington.edu)

The QALY utility model is usually applied in medical DA under EU assumptions and the assumption of linear utility of survival duration. We present work on QALY utility assessment under the less restrictive assumptions of rank-dependent utility and nonlinear utility of survival duration.

TB23.3 Predicting Health Decisions: The Application of Subjective Expected Utility & Remaining Challenges Young-Hee Cho, Robin L. Keller, M. Lynne Cooper --- Univ. of Missouri, Dept. of Psych., Columbia, , ()

We critically examine the application of the SEU to modeling risk-taking behaviors and review the studies which investigate the predictive validity of SEU for risk-reducing or preventative measures. We then discuss challenges in modeling health decisions, including the complexity of health decision making and difficulty in eliciting the decision components.

TB23.4 Reasons for Rank-Dependent Utility Evaluation Elke U. Weber --- Ohio State University, 1885 Neil Avenue, Dept. of Psych. & Mngmnt. & hr, Columbus, OH 43210 , (weber.211@osu.edu)

Three reasons for evaluating utility in a rank-dependent fashion have been suggested: perceptual biases, individual predispositions in weighting and rational adaptation to an asymmetric loss function. We present evidence for all 3 processes which have different implications for successful debiasing procedures.

TC23.1 A Tutorial on Life and Death Decision Making Ronald A. Howard --- Stanford Univ., Dept. of EES & OR, Stanford, CA 94305-4025, (rhoward@sdg.com)

Major life risks arise in health crises; minor ones in everyday life. A model incorporating life quality and quantity treats both. Simplification for minor risks: assess the risk in micromorts, 1/1,000,000 chances of death, multiply by the model's micromort value. Extensions cover major health risks and risks to a couple.

TD23.1 Web-Based Financial Decision Analysis Terry Bresnick, Dennis M. Buede --- Innovative Decision Analysis, Inc., 3011 Weber Pl., Oakton, VA 22124 , ()

Over the past 2 years we have designed web-based applications for Quicken Financial Network and InsureMarket. The applications include personal check-ups, rule-based decision aides and DA-based decision aides. We summarize the applications, demonstrate the applications if a web connection is available and provide an update on the usage and success of the applications.

TD23.2 Environmental Risk Management Daniel G. Brooks --- AZ State Univ., Dept. of Mgmt., Tempe, AZ 85287 , (daniel.brooks@asu.edu)

Petro-chemical companies have been transporting their products over aging, risk-prone pipeline systems for years. We present how we developed a consistent system-wide process for assessing and managing risk for Chevron Pipeline (CPL) Company that has been institutionalized throughout CPL.

TD23.3 Manufacturing Risk Management Jeffrey S. Stonebraker --- Applied Decision Analysis, Inc., 2710 Sand Hill Rd., Menlo Park, CA 94025 , (jeff@adainc.com)

We discuss our 1st application of risk management in manufacturing. The schedule of this pilot study was very tight having been compressed from months to weeks. We also discuss the organizational impacts of this study along with our lessons learned.

TD23.4 withdrawn - author request of 2/26 James C. Felli --- Naval Postgrad. Sch., DRMI (64FL), Monterey, CA 93943-5201, (jcfelli@nps.navy.mil)

TE23.1 Portfolio Management in an Upstream Oil & Gas Organization Mazen A Skaf --- SDG, 2440 Sand Hill Rd, Menlo Park, CA 94025 , (mskaf@sdg.com)

We describe a portfolio management process and system we have developed and successfully implemented in the oil & gas industry. We discuss our approach to portfolio management, the process, the system architecture and benefits of applying the process to multi-billion dollar asset portfolios in the deepwater Gulf of Mexico.

TE23.2 Analysis of Indirect Investment Effects Thomas R. Varner --- , PO Box 1261, Lafayette, CA 94549 , (tvarner@earthlink.net)

An analytical framework is presented to analyze the effects that current investment decisions have on future prospects. Existing strategic models typically consider these 'indirect effects' only on a qualitative or intuitive basis. Specific issues addressed with this analytical framework include financing, learning, project spawning and market effects.

TE23.3 Models & Strategy: An Arranged Marriage Patrick S. Noonan --- Emory Univ., Goizueta Bus. Sch., Atlanta, GA 30322 , (patrick_noonan@bus.emory.edu)

Despite their many surface incompatibilities, what OR/MS professionals consider 'modeling' and what planning professionals consider 'strategy' can be combined effectively. Numerous case examples illustrate - and scholarly research on individual and organizational cognition and decision explains - the critical prescriptive issues in approaching strategic decision processes with OR tools.

TE23.4 A Decision Analysis Approach for Strategic Capital Allocation in Health Care Don N. Kleinmuntz, Catherine E. Kleinmuntz --- Univ. of IL, 1206 S 6th St., Rm. 360, Dept. of Bus. Admin., Champaign, IL 61820 , (dnk@uiuc.edu)

We develop a methodology for allocating capital across organizational units, service lines and facilities. It balances critical financial, strategic and quality goals, and is particularly well suited to complex integrated healthcare organizations. We describe implementation at a large urban hospital with a capital budget in excess of $10 million.

WA23.1 Defining Team Processes in Decision Analyses James McCuish --- Amoco, 3700 Bay Area Blvd., Houston, TX 77058 , (jdmccuish@amoco.com)

In applications of DA, teams often have difficulty in following through after the initial framing. At Amoco, we utilize approaches from both value engineering and project management tools to augment the team's planning and define steps to apply the appropriate DA s. We present the methods used and discuss their relative successes.

WA23.2 Implementing DA in a Large Company: The Conoco Model Ellen M Coopersmith --- Conoco Inc, 600 North Dairy Ashford, P O Box 2197, Houston, TX 77252-2197, ()

Many companies attempt to implement DA in their organizations, all using different implementation strategies and results. Conoco has implemented the process in an attempt to make better decisions and to change the decision culture of the company. We discuss Conoco's business environment, implementation strategy, results and critical success factors associated with their success to date.

WA23.3 Experience of Applying Decision Analysis to Equipment Maintenance Planning Daniel I. Ash, David A. Mauney --- Southwest Research Inst., Materials & Structures Div., 6220 Cuebra Rd., San Antonio, TX 78238-5166, ()

We illustrate maintenance planning for a variety of equipment in the power generation, petroleum refining, petroleum production and gas pipeline industries. Also, the unique application of DA techniques is discussed.

WA23.4 withdrawn - author request of 3/8 Phil Kerig --- Conoco, Inc., POB 2197, Rm. MC 3075, Houston, TX 77252-2197, ()

WB23.1 Case Studies of a Real Options Approach to Asset Valuation in the Energy Industry Gardner Walkup --- Applied Decision Analysis Inc., 2710 sand Hill Rd., Menlo Park, CA 94025 , (gardner@adainc.com)

We present a practical approach to asset valuation which combines the strengths of Option Pricing Theory (OPT) with DA. Three real case studies from the Power Industry are presented. The results demonstrate that explicitly valuing managerial flexibility, corporate synergies and market opportunities result in significantly different valuations and optimal policies.

WB23.2 Flexibility in the Natural Gas Regulatory Framework: A Case Study Alexsandr Rudkevich, Daniel Smith --- Tellus Institute, 11 Arlington Street, Boston, MA 02116 , (arudkevi@tellus.org)

Gas distribution companies entering long-term contracts with pipelines must justify their decisions before state regulators. In reviewing these decisions, regulators apply 3 standard criteria - service reliability, cost and flexibility. Unlike the 1st 2 criteria, flexibility is hard to quantify. An approach to quantifying flexibility is tested in a litigated case.

WB23.3 withdrawn - author request of 3/6 Tom Hoff, Howard Wenger, Christy Herig --- Pacific Energy Group, 10 Glen Ct., Napa, CA 94558 , (tomhoff@leland.stanford.edu)

WB23.4 withdrawn - author request of 3/6 Anne Ku --- Enron Europe, 4 Millbank, London SWIP, 3ET , UK (aku@enron.co.uk)

WC23.1 Real Options & Product Life Cycles Nicolas P. Bollen --- Univ. of Utah, Eccles Sch. of bus., Salt Lake City, UT 84112 , (finnb@business.utah.edu)

We investigate the value of projects that produce goods for markets with changes in demand. If demand is cyclical, then standard real options techniques can result in significant valuation error. There are important implications for investment decisions, especially in high-technology industries that feature regular introductions of new products.

WC23.2 Strategic Aquisition of Information Technology Nalin Kulatilaka --- Boston Univ., Dept. of Mgmt., 595 Commonwealth Ave., Rm. 536, Boston, MA 02215 , (nalink@acs.bu.edu)

IT investment decisions are difficult because of elusive benefits and diffuse costs. This is a case study of how an IT investment strategy is formulated and executed using the real options approach. We also include the timing of investment decisions in our analysis.

WC23.3 A Real Options Valuation of Electric Generation Capacity Martha Amram, Jesse Langston --- Analysis Group Economics, Two Embanadero Center, Suite 1160, San Francisco, CA 94111 , (mamram@ag-inc.com)

We demonstrate how the real options approach can value the flexibility of peaking capacity in a volatile electricity market. Our approach quantifies the value of both the operating and investment options imbedded in the asset.