Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TD15

n TD14 North Bldg 126C Queues in Services I Sponsored: Manufacturing & Service Oper Mgmt/Service Operations Sponsored Session Chair: Amy R. Ward, University of Southern California, Los Angeles, CA, 90089-0809, United States Co-Chair: Rouba Ibrahim, University College London, London, WC1E 6BT, United Kingdom 1 - Large Deviations Analysis for Non-Markovian Multi-server Queues with Abandonments in the QED Regime David Goldberg, Cornell University, 136 Hoy Road, Ithaca, NY, 14850, United States, Debankur Mukherjee, Yuan Li We consider the FCFS M/H_2/n + M queue in the QED regime. It is known that the normalized sequence of steady-state queue-lengths converges to a limit W, with sub-Gaussian tails. However, the exact tail behavior was left open. Dai and He later conjectured an explicit form for this tail exponent, which was insensitive to higher moments. We explicitly compute the true large deviations exponent, the first such result for non-Markovian queues with abandonments, and resolve the conjecture of Dai and He in the negative. Our stochastic comparison approach sheds light on several novel ways to think about multi-server queues with abandonments, and may be helpful in analyzing other related models. 2 - Workload Management in Telemedical Physician Triage and Other Knowledge-based Service Systems Soroush Saghafian, Harvard Univeristy, Kennedy School of Government, 79 John F. Kennedy Street, Cambridge, MA, 02138, United States, Wallace J. Hopp, Seyed Iravani, Yao Cheng, Daniel Diermeier Telemedical physician triage (TPT) is an example of a hierarchical knowledgebased service system (HKBSS) in which a second level of decision agent (telemedical physician) renders a decision on cases referred to him/her by the primary level agents (triage nurses). Managing the speed-versus-quality trade-off in such systems presents a unique challenge because of the interplay between agent knowledge and flow of work between the two levels. We develop a novel model of agent knowledge based on the beta distribution, and deploy it in a partially observable Markov decision process model to describe the optimal policy for deciding which cases (patients) to refer to the second level for further evaluation. 3 - Operational Issues in Large Jail and Judiciary Systems Russell Charles Hannigan, University of Chicago Booth School of Business, Chicago, IL, 60615, United States, Baris Ata We perform an analysis of a large jail and judiciary system with the goal of simultaneously reducing Length of Stay (LOS) and improving outcomes for detainees. Two primary issues contribute to high LOS: a large number of outstanding warrants and detainees’ trials often lasting longer than their eventual sentences require. We model the jail as a queueing network and consider the benefits of letting low-level warrants expire. We also consider a dynamic discrete choice model of detainees’ behavior to estimate detainees sentence location sensitivity and resulting optimal trial termination time. 4 - The Impact of Customer Impatience on Scheduling in a Multi-class Many-server Queue Amy R. Ward, University of Southern California, Marshall School of Business, Bridge Hall BRI 401H, Los Angeles, CA, 90089-0809, United States, Amber L. Puha Many classic models used to study scheduling problems do not incorporate customer impatience. Furthermore, many of the ones that do assume the time a customer is willing to wait for service is exponentially distributed, which can lead to poor scheduling decisions. We study the interplay between customer impatience and scheduling decisions in a many server queue with reneging (G/G/N+G) and multiple customer classes. On fluid scale, we characterize a closed-form non-linear relationship between the queue-length and server effort allocation that is determined by the reneging distribution. This leads to an optimization problem whose solution provides insight into scheduling decisions.

n TD15 North Bldg 127A Economic Models in OM Sponsored: Manufacturing & Service Oper Mgmt/Service Operations Sponsored Session Chair: Krishnamurthy Iyer, Cornell University, Ithaca, NY, 14850, United States 1 - Diffusion in Random Networks: Impact of Degree Distribution Scott Rodilitz, Yale University, New Haven, CT, United States, Vahideh Manshadi Motivated by viral marketing on social networks, we study the diffusion process of a new product on a network where each agent is connected to a random subset of others. The number of contacts (i.e., degree) varies across agents and the firm knows the degree of each agent. Further, the firm can seed a fraction of the population and compensate them on a per-contact basis. Under any bounded degree distribution and for any target adoption proportion, we compute both the cost and the time it takes to reach the target in the limit of network size. Using our limit results, we conduct comparative statics on degree distribution and I consider information and pricing design in unobservable queues with strategic and heterogenous customers that are privately informed and impatient. I show that 1) if the platform optimizes among all possible pricing and information mechanisms, the optimum is achieved by disclosing all the information about congestion and setting a sequence of posted prices, 2) for any information policy, the steady-state can be replicated by full-information and prices while achieving a higher revenue, 3) I identify conditions under which the platform profits from releasing information strategically and provide a characterization for the optimal rule. 3 - Section 8 Voucher Exchange Nicholas A. Arnosti, Columbia Business School, 3022 Broadway, Uris Hall, New York, NY, 10027, United States We consider a dynamic model motivated by the administration of Section 8 vouchers. In the model, family relocation results in inefficient billing between housing authorities. We study the potential for voucher exchange to reduce billing. We show that if the rate at which families become ineligible is identical across cities, then voucher swapping has no long-term effect on the number of voucher recipients living in each city. This implies that either pairwise or multi- way voucher exchange offer a Pareto improvement over current practice. Additionally, we calibrate our model using data from the Twin Cities metropolitan area, and study the magnitude of the benefit of multi-way exchange. 4 - A Truth Serum for Large-scale Evaluations Vijay Kamble, University of Illinois at Chicago, 405 N. Wabash Ave, Unit 3511, Chicago, IL, 60611, United States, Marn David, Nihar Shah, Abhay Parekh, Kannan Ramchandran A major challenge in obtaining large-scale evaluations of products or services in online platforms is that of eliciting honest responses from agents in the absence of verifiability. We propose a new reward mechanism with strong incentive properties applicable in a wide variety of such settings. This mechanism has a simple and intuitive output agreement structure: an agent gets a reward only if her response for an evaluation matches that of her peer. But instead of the reward being the same across answers, it is inversely proportional to a popularity index of each answer. Rare agreements earn a higher reward than agreements that are relatively more common. uncover a trade-off between cost efficiency and fast growth. 2 - Information and Pricing in Unobservable Queues Guido Martirena, Stanford University, 579 Serra Mall, Stanford, CA, 94305, United States

351