Craig W. Kirkwood

Ranking with Partial Information: A Method and an Application

A method is presented for ranking multiattributed alternatives using a weighted-additive evaluation function with partial information about the weighting scaling constants, the method is applied to evaluate materials for use in nuclear waste containment. The paper derives conditions to determine whether a pair of alternatives can be ranked given the partial information about weighting constants, and presents an algorithm that partially rank-orders the complete set of alternatives based on the pairwise ranking information.

Decision Analysis Applications in the Operations Research Literature, 1970-1989

In order to provide a guide to source material for practitioners interested in applying decision analysis methods, this paper surveys applications of decision analysis published from 1970 through 1989. In addition, it presents references for useful decision analysis methods that are often omitted from introductory textbooks. As used in this article, the term decision analysis refers to a set of quantitative methods for analyzing decisions which use expected utility as the criterion for identifying the preferred decision alternative. To be included in this survey, an application had to explicitly analyze alternatives for a decision problem using judgmental probabilities and/or subjectively assessed utility functions. The paper classifies the applications into five areas: energy, manufacturing and services, medical, public policy, and general. It further subclassifies energy applications into bidding, product and project selection, regulation, site selection, and technology choice. Those in manufacturing and services are subclassified into budget allocation, product planning, strategy, and miscellaneous. Applications in public policy are subclassified into standard-setting and miscellaneous. The paper notes articles that present significant detail about methodological and implementation issues, including problem structure/formulation, decision trees, probability and utility assessment, communication/facilitation, and group decision making.

Perspective on Decision Analysis Applications, 1990-2001

This article identifies, and provides perspective on, trends and developments in decision analysis applications, based primarily on an exhaustive survey of decision analysis applications published in the period 1990-2001 in major English-language operations research and closely related journals. It serves as a guide to those interested in recent applications in specific areas or in applications that illustrate the use of particular methods. We compare the characteristics of the applications articles surveyed here with those of applications articles appearing in a similar set of journals between 1970 and 1989 and conclude that the overall rate of publication of decision analysis applications has increased. In addition, we find that both the mix of application areas and the specific aspects of decision analysis that are emphasized in applications publications have shifted somewhat. We also identify and discuss noteworthy trends in, and developments affecting, published applications, including those in computer software and software-related tools, decision conferencing, stochastic trees, value-focused thinking, normative systems, organizational processes, and real options. We highlight several award-winning decision analysis applications and discuss formation of a new practitioner-oriented professional group. Finally, we present some concerns and thoughts on future needs for advancing decision analysis practice.

Multiattribute Preference Analysis with Performance Targets

This paper develops an approach based on performance targets to assess a preference function for a multiobjective decision under uncertainty. This approach yields preference functions that are strategically equivalent to conventional multiattribute utility functions, but the target-oriented approach is more natural for some classes of decisions. In some situations, the target-oriented preference conditions are analogous to reliability theory conditions for series or parallel failure modes in a system. In such cases, reinterpreting the conditions using reliability concepts can be useful in assessing the preference function. The target-oriented approach is also a generalization of common forms of goal programming. The approach has particular applicability for resource allocation decisions where the outcome of the decision is significantly determined by the actions of other stakeholders to the decision, such as new product development or decision making in a controversial regulated environment.

Water Planning Under Climatic Uncertainty in Phoenix: Why We Need a New Paradigm

The uncertainties associated with global climate models pose substantial hurdles for urban water planning. Despite growing consensus among climatologists that the American Southwest is headed for a warmer and drier future, water planners in metropolitan Phoenix and elsewhere are reluctant to consider long-term climate change as a significant factor in increased risk of future water scarcity. A new paradigm for climate research and water planning is needed—one that is based on an assumption of uncertainty and a vision of multiple plausible futures, managing risk, and adaptive behaviors. To this end, we downscaled global climate models from the Intergovernmental Panel on Climate Change Third and Fourth Assessment Reports for the watersheds north of Phoenix and estimated changes in runoff using a hydrological model. Results then were used as inputs to WaterSim, an integrated simulation model of water supply and demand in Phoenix. The model simulated “what if” scenarios under varying policy decisions and future climates. Results of simulation experiments suggest that (1) current levels of per capita water consumption cannot be supported without unsustainable groundwater use under most climate model scenarios, (2) feasible reductions in residential water consumption allow the region to weather the most pessimistic of the climate projections, (3) delaying action reduces long-term sustainability of the groundwater resource under some climate scenarios, and (4) adaptive policy with appropriate monitoring to track groundwater provides warning that the need for use restrictions is approaching and avoids the need for drastic, ad hoc actions.

Vulnerability assessment of climate-induced water shortage in Phoenix

Global warming has profound consequences for the climate of the American Southwest and its overallocated water supplies. This paper uses simulation modeling and the principles of decision making under uncertainty to translate climate information into tools for vulnerability assessment and urban climate adaptation. A dynamic simulation model, WaterSim, is used to explore future water-shortage conditions in Phoenix. Results indicate that policy action will be needed to attain water sustainability in 2030, even without reductions in river flows caused by climate change. Challenging but feasible changes in lifestyle and slower rates of population growth would allow the region to avoid shortage conditions and achieve groundwater sustainability under all but the most dire climate scenarios. Changes in lifestyle involve more native desert landscaping and fewer pools in addition to slower growth and higher urban densities. There is not a single most likely or optimal future for Phoenix. Urban climate adaptation involves using science-based models to anticipate water shortage and manage climate risk.

Preference Conditions for Multiattribute Value Functions

This paper examines conditions on preferences that simplify the assessment of multiattribute value functions for use in the analysis of multiobjective decision problems. It is shown that when these conditions hold the value function must have a simple analytic form. A procedure is presented for testing whether the conditions hold and determining the value function when the conditions are found to be valid.

WaterSim: a simulation model for urban water planning in Phoenix, Arizona, USA

WaterSim, a simulation model, was built and implemented to investigate how alternative climate conditions, rates of population growth, and policy choices interact to affect future water supply and demand conditions in Phoenix, AZ. WaterSim is a hierarchical model that represents supply from surface and groundwater sources and demand from residential, commercial, and agricultural user sectors, incorporating the rules that govern reservoirs, aquifer use, and land-use change. In this paper we: (1) report on the imperative for exploratory modeling in water-resource management, given the deep uncertainties of climate change, (2) describe the geographic context for the Phoenix case study, (3) outline the objectives and structure of WaterSim, (4) report on testing the model with sensitivity analyses and history matching, (5) demonstrate the application of the model through a series of simulation experiments, and (6) discuss the models use for scenario planning and climate adaptation. Simulation results show there are significant challenges to Phoenixs water sustainability from climate change and rapid growth. Policies to address these challenges require difficult tradeoffs among lifestyles, groundwater sustainability, the pace of growth, and what is considered to be an appropriate level of risk of climate-induced shortage.

Parametrically Dependent Preferences for Multiattributed Consequences

The assessment of multiattribute cardinal utility functions can be a difficult experimental problem. To aid in this assessment, we define the concept of parametric dependence of preferences for multiattributed consequences. We show the usefulness of this concept for the assessment of multiattribute utility functions by proving several representation theorems that simplify the assessment problem when parametric dependence holds.

Improving Supply-Chain-Reconfiguration Decisions at IBM

We developed a decision-support system for IBMs supply-chain-configuration decisions. Managers and analysts used this prepackaged multiobjective decision-analysis procedure in facilitated workshops to analyze mid-level supply-chain configuration decisions based on 22 considerations covering cost, quality, customer responsiveness, strategic issues, and operating constraints. These multiattribute utility analyses incorporated uncertainty through expert estimates of probabilities and were implemented in a spreadsheet environment. We applied the approach to five IBM supply-chain decisions, and the results satisfied internal stakeholders that the analysis correctly included financial and nonfinancial considerations along with the associated risks and provided a useful audit trail for executive management. IBM now views this decision-support system as a potential template for future supply-chain decisions.