In This Issue

A Quantum Mechanics Approach to Decision Making

Half a century ago, Herbert Simon, economics Nobel Laureate, in collaboration with Allen Newell, expressed that, “In dealing with the ill-structured problems of management we have not had the mathematical tools we have needed—we have not had ‘judgment mechanics’ to match quantum mechanics.” Such expectations regarding the necessity of understanding our own minds and a mechanics of the decision-making process are not yet fulfilled. In “Quantum Mechanics and Human Decision Making,” P. M. Agrawal and R. Sharda show that we are gaining momentum in understanding the human decision processes through quantum mechanics. This OR Forum paper describes essential concepts of quantum mechanics and reviews the recent progress to stimulate more exploratory research on applications of quantum mechanics concepts in decision making. The authors also propose several applications that might benefit from applying a quantum mechanics lens to understand the decision-making models.

Estimating the Data Quality of Query Results

In making critical business decisions, managers often rely on information gleaned from various data sources. Errors in data values at these sources may lead to costly errors in business decisions. Today's managers must, therefore, have a good estimate of the quality profile of a query's result before using it in a decision-making context. Previous studies have examined how the quality level of a database query output can be estimated based on the quality level of the input data. In “Data Quality of Query Results with Generalized Selection Conditions,” D. Dey and S. Kumar generalize this stream of research and allow a query to have general selection conditions involving multiple attributes with any combination of conjunction or disjunction of subconditions that may even include functions of multiple attributes. Results of their research can be easily implemented in real-world decision contexts.

Tackling the Velocity of Big Data: Refreshing Knowledge Discovered from Data

Knowledge extracted from consumer behavioral, social networking, business transaction, law enforcement, and clinical data has increased the value of behavior targeting marketing campaigns; improved the quality of health, customer, and government services; and facilitated supply chain and day-to-day management decision making. To sustain these data-driven values, as companies such as Google and IBM have indicated, the currency of knowledge must be maintained over evolving customer and transaction data. In “When Is the Right Time to Refresh Knowledge Discovered from Data?” X. Fang, O. R. Liu Sheng, and P. Goes address this issue from the perspective of deciding when to rediscover knowledge (namely, the knowledge refreshing problem). They show that an in-depth analysis of knowledge loss gives knowledge-driven organizations an understanding of and foundation for managing knowledge loss. They also present a practical, automated knowledge management tool for determining when to refresh discovered knowledge while minimizing knowledge loss and refreshment cost.

A Model-Based Analysis for Improving Airline Reliability

Flight delays reached an all-time high in recent years, with airlines' on-time performance at its worst level in 2007 since 1995. A recent report by the Joint Economic Committee of the U.S. Congress estimated that flight delays cost the U.S. economy as much as $41 billion in 2007. In “Building Reliable Air Travel Infrastructure Using Empirical Data and Stochastic Models of Airline Networks,” M. Arikan, V. Deshpande, and M. Sohoni utilize publicly available data to analyze the propagation of delays through air transportation networks. Their stochastic models allow the development of important robustness measures for airline networks. These robustness measures can serve as useful tools for policy planners, airline managers, and passengers. The authors' model-based analysis enables them to make policy recommendations regarding managing bottleneck resources in the air travel infrastructure. If followed, these recommendations could lead to a significant improvement in air travel reliability.

Project Management and Keeping Adversaries in the Dark

Minimizing project duration is the most frequently assumed objective in the project management literature. However, if the project manager wishes to keep the project secret, such as a product innovation in a competitive environment, simply minimizing duration may not be optimal. Rather, minimizing the time between when an adversary has enough information to act and when the project is completed may be preferable. In “Managing a Secret Project,” E. Pinker, J. Szmerekovsy, and V. Tilson formulate a project management model using this alternative objective within a Stackelberg game setting. With a model that allows for deception activities, they demonstrate the interconnectedness of deception, scheduling, and crashing in reducing the exposure of a secret project to an adversary's response and how these activities influence an adversary's behavior. They apply their approach to the example of nuclear weapons proliferation and relate it to the current concerns about Iranian nuclear activities.

Designing a Fair, Efficient, and Flexible Organ Transplantation Allocation System

In the past 10 years, U.S. policy makers have made a nationwide effort to redesign the national allocation system of deceased donor kidneys for transplantation to make it more efficient and equitable. This challenging task requires expertise not only in identifying the right balance between equity and efficiency but also in designing the actual allocation policy. In “Fairness, Efficiency, and Flexibility in Organ Allocation for Kidney Transplantation,” D. Bertsimas, V. F. Farias, and N. Trichakis propose an outcome-driven method to make the design phase systematic. In particular, the proposed method utilizes historical data to design policies that conform to implementation requirements, approximately maximize efficiency, and achieve user-specified outcomes. The authors demonstrate via simulation studies the method's flexibility, by designing different policies tailored to different outcome requirements, and its efficiency, by designing a policy that improves life year gains by 8% compared to proposals considered by policy makers, without being less fair.

Dual Sourcing Involving Fixed Costs

Traditional procurement models in the operations management literature assume constant or known procurement prices. On the other hand, the spot market with fluctuating prices provides a more flexible channel for procurement. The combined use of the two channels poses a challenging problem. In “Optimal Inventory Policy in the Presence of a Long-Term Supplier and a Spot Market,” Y. Chen, W. Xue, and J. Yang analyze a procurement model that incorporates traditional contract procurement and spot market purchases. Emphasis has been placed on the case in which a fixed cost is incurred when order from the spot market is made. An optimal policy is shown to possess three different forms, with the realization of each form depending on the buyer's inventory level and the prevalent spot price. The authors also identify conditions that would enable monotone trends between policy parameters and spot prices.

Dynamic Second-Price Auctions

The second-price auction has become the prevalent pricing scheme in online advertising. There is a comprehensive literature on the second-price auction in static environments; however, the properties of this mechanism are not yet well understood in dynamic environments where the buyers are uncertain about their valuations. In “Dynamic Pay-Per-Action Mechanisms and Applications to Online Advertising,” H. Nazerzadeh, A. Saberi, and R. Vohra show that the second-price auction, combined with simple learning algorithms, can be implemented in dynamic environments while essentially maintaining the desired properties of this auction mechanism in static environments. The proposed dynamic mechanism is applicable to online advertising markets and would address some of the challenges of this market including click-fraud.

Calculating Cost-To-Serve in Complex Logistics Problems

One of a service provider's most important tasks is calculating each customer's “cost-to-serve.” Most simple methods for allocating the total service costs among customers distribute the costs proportional to a set of customer attributes such as location and consumption rate but do not account for the synergies between customers and hence do not represent the true cost-to-serve. Inventory routing describes a set of vendor-managed inventory replenishment problems in which the service provider decides when and how much to serve to each customer, exploiting synergies between them due to their locations, usage rates, and storage capacities, to reduce distribution costs by serving nearby customers on the same route at the same time. However, these synergies complicate true cost-to-serve calculations. In “Allocating Cost of Service to Customers in Inventory Routing,” O. Ö. Özener, Ö. Ergun, and M. Savelsbergh present algorithms for estimating cost-to-serve and provide computational evidence on the quality of these estimates.

Portfolio Optimization and Gambling

Generally speaking, risk-averse agents do not accept fair gambles because doing so reduces their expected utility. In “Risk Aversion, Indivisible Timing Options, and Gambling,” V. Henderson and D. Hobson show that this conclusion can change if an agent has an indivisible asset to sell—such as real estate, their business, or restricted stock or options. Surprisingly, it can then be optimal for the risk-averse agent to take a position in an asset that has zero expected return and is uncorrelated with the indivisible asset—essentially a risk increasing fair gamble. Informally, the agent should take a trip to the casino before selling his or her house, company, or stock. From the perspective of portfolio choice modeling, this observation demonstrates that seemingly extraneous assets can form part of an agent's optimal portfolio. It also offers a rational explanation for the occurrence of gambling by risk-averse agents.

Dynamic Routing of Multivehicle Fleets

Vehicle routing problems with stochastic demand underlie operational challenges in logistics (for example, LTL trucking and vendor-managed distribution), where routes must be planned without full knowledge of customer demand levels. In “Rollout Policies for Dynamic Solutions to the Multivehicle Routing Problem with Stochastic Demand and Duration Limits,” J. C. Goodson, J. W. Ohlmann, and B. W. Thomas develop rollout-based procedures to dynamically adjust vehicle routes based on observed customer demands. The authors' dynamic decomposition scheme enables dynamic management of large vehicle fleets servicing many customers. Computational experiments demonstrate the rollout-based methods improve upon the performance of a rolling horizon procedure and commonly employed fixed-route policies.

Utility Design for Distributed Engineering Systems

Recently there has been a surge of research attention in applying game theoretic methods to the control of distributed engineering systems. Here, the goal is to establish a dynamical process that ensures that the collective behavior is desirable with respect to a given system-level objective. A pivotal choice in this design process is the assignment of local utility functions to the decision-making entities. In “Distributed Welfare Games,” J. R. Marden and A. Wierman formalize some of the questions and objectives associated with utility design for distributed engineering systems and present some initial results illustrating how classical approaches in cooperative game theory can effectively be utilized in distributed engineering systems.

Classical Portfolio Selection for the Tri-Criterion Situation

Almost everyone is familiar with Markowitz's mean-variance efficient frontier, which was introduced more than 60 years ago. Although mean-variance is still the predominant model in portfolio selection, it has endured many criticisms. One serious criticism is that classical theory cannot scale to additional criteria. The difficulty is that the efficient frontier becomes a surface. In “Computing the Nondominated Surface in Tri-Criterion Portfolio Selection,” M. Hirschberger, R. E. Steuer, S. Utz, M. Wimmer, and Y. Qi, motivated by the desire to extend Markowitz portfolio selection to a third linear criterion (liquidity, momentum, dividend yield, and so forth), provide an exact method for computing a tri-criterion efficient surface in such a situation. With computer graphics, an efficient surface can be shown all at once just as efficient frontiers are commonly shown all at once in traditional portfolio selection. The authors illustrate the usefulness of efficient surface knowledge in an empirical application.

Flow Techniques for Solving the Cheeger, Expander, and Normalized Cut Problems

The Cheeger, expander, and normalized cut problems have traditionally been solved, heuristically, using the “spectral technique.” A unified framework is provided whereby these problems are formulated as a constrained minimization form of a quadratic ratio, called the Rayleigh ratio. In “A Polynomial Time Algorithm for Rayleigh Ratio on Discrete Variables: Replacing Spectral Techniques for Expander Ratio, Normalized Cut, and Cheeger Constant,” D. S. Hochbaum devises an efficient combinatorial algorithm, based on flow techniques, that solves a relaxation of the Rayleigh ratio problem in integers. The author not only shows that the discrete Rayleigh ratio problem is polynomial time solvable, but also that the combinatorial algorithm is more efficient than the spectral algorithm. Furthermore, an experimental study demonstrates that the new algorithm improves the quality of the spectral method's results, in terms of both approximating the true optimum of the Rayleigh ratio problem on the discrete variables and the balance constraint, and the subjective partition quality.

Designing Telecommunication Networks for Survivability

Network survivability is a vital issue in the design of telecommunication networks. A survivable network should continue carrying the traffic it is designed for via alternate paths under any link failure (for example, a fiber cut). The optimal design of such networks has been considered a notoriously hard problem due to numerous variables and constraints. In the past, problems larger than 10 nodes have been considered too big to permit optimal solutions. In “Design of Survivable Networks Using Three- and Four-Partition Facets,” Y. Agarwal presents a polyhedral approach coupled with an efficient implementation of a compact problem formulation called the “capacity formulation,” which succeeds in obtaining optimal solutions for networks with 35 nodes and 80 edges and a fully dense traffic matrix in less than five minutes of computer time.

A Satisficing Paradigm for Multiple Objectives Problems Under Uncertainty

Many real-world problems must deal with multiple conflicting objectives, such as profits and service levels, in which their outcomes are often uncertain. In “Multiple Objective Satisficing Under Uncertainty,” S.-W. Lam, T. S. Ng, M. Sim, and J.-H. Song study the problem of making a decision that has the best prospect of these multiple objectives in attaining some predetermined target levels. It rigorously extends the concept of satisficing to a multiple objective setting and can be applied to situations in which the probability distribution is not fully characterized. The authors also suggest that the method should encourage diversification, which, as a by-product, also leads to more tractable optimization models.

Fairness Toward Employees in Large-Scale Service Organizations

Justice toward employees has been mostly ignored in the operations management literature. However, it has been shown that in the absence of a sense of justice, or fairness, employees' performance degrades and turnover is higher, resulting in higher operational costs. In a previous paper, A. R. Ward and M. Armony focused on how to assign workload to servers in large-scale service systems to obtain fairness in the long run. In “Blind Fair Routing in Large-Scale Service Systems,” the same authors focus on how to assign workload fairly and blindly—that is, without requiring knowledge of system parameter information (such as arrival rates, service times, and server pool sizes). The authors find that the cost of requiring blindness is low; in fact, it is much lower than the cost of requiring fairness.

Designing Redundancy into High-Availability Technical Systems

For many technical systems (for example, medical, manufacturing, and communication systems), users require increasingly high system availability because these systems are essential for primary operational processes. One can meet these elevated availability requirements by increasing the spare parts stock or using faster supply options for the spare parts stock. However, from a total cost of ownership (TCO) perspective, it might be much more efficient to be proactive in the design phase of a technical system and to build in redundancy for some critical components. In “Redundancy Optimization for Critical Components in High-Availability Technical Systems,” K. B. Öner, A. Scheller-Wolf, and G.-J. van Houtum develop a decision support model for this problem. Their model determines how much redundancy should be built in for new systems, under a given target system availability level, and also generates the efficient frontier for the trade-off between system availability and TCO.