INFORMS 2021 Program Book

INFORMS Anaheim 2021

WB09

2 - Effectiveness of AI Assistant in Live-streaming: A Randomized Field Experiment Yumei He, University of Houston, Houston, TX, United States, Lingli Wang, Jiandong Ding, Ni Huang, Yili Kevin Hong, Xunhua Guo, De Liu, Guoqing Chen Live streaming selling is a transformative ecommerce channel that features real- time interaction. However, streamers’ service capacities cannot fulfill viewers’ needs for social interaction, resulting in a loss of viewer engagement and financial gains. Accordingly, we examine whether the implementation of an AI assistant that tracks, understands, and responds viewers’ requests drives viewers’ purchase decision. In collaboration with Taobao.com, we conducted a randomized field experiment on its live streaming platform. In the experiment, a subject was either assigned to the treatment group with an AI assistant in any live streaming or the control group in which the AI assistant was absent. Our research contributes to the literature on the business value of AI applications and live streaming while implicating AI system designs for live streaming platforms. 3 - Pedaling Our Way to Clean Air: An Empirical Investigation of Bike- sharing Platforms and Local Air Quality Ecem Basak, University of Illinois at Chicago, Chicago, IL, 60640, United States, Ali Tafti, Mary Beth Watson-Manheim Bike-sharing platforms contribute to better allocation and more efficient utilization of resources. However, bike-sharing platforms are not only associated with reduced traffic congestion and flexible mobility but also offer health benefits such as reduced greenhouse gas emissions and air pollution. Environmental sustainability is important for creating healthy societies and eco-friendly cities. The sustainability of cities and their greening is essential to address environmental and health issues. We implement a difference-in-differences analysis to examine the impact of bike-sharing platform entry on the level of PM2.5 concentrations in U.S. cities. We also explore the heterogeneous impacts of the platforms depending on multiple factors such as bike-friendliness, pedestrian-friendliness, environmental policies, and population. WB11 CC Room 304C In Person: Airline Operations Recovery General Session Chair: Vikrant Vaze, Dartmouth College, Hanover, NH, 03755-3560, United States Co-Chair: Lu Dai, University of California, Berkeley, CA 1 - The COVID-19 Pandemic and U.S. Aviation: System Adaptation and Performance Impact Michael O. Ball, University of Maryland-College Park, Silver Spring, MD, 20910, United States, Vivek Ramanathan, Dan Murphy, Mark M. Hansen, Vanessa Li We investigate the impact of the COVID-19 pandemic on the performance of the U.S. domestic air transportation system. We analyze both the changes in the volume and characteristics of flight operations and also changes in system performance metrics. We also discuss various adjustments made by the FAA both to take advantage of reduced system congestion and also to cope with COVID-19 infections among controllers. 2 - Aiding Airlines in a Pandemic for the Benefit of Whom? An Applied Game-Theoretic Approach Gianmarco Andreana, Univerità degli Studi di Bergamo, Bergamo, Italy, Nicole Adler In 2020 the Covid-19 pandemic disrupted aviation industry profitability. Governments and banks stepped in to bailout airlines with substantial financial aid of differing forms. This research models airline competition by developing a single-stage, Nash best-response dynamic game. Given this framework, airlines compete at a strategical level setting airfares and service frequencies across their network. Investigating the European aviation market, characterized by legacy and low-cost carriers, we assess how the form of aid distorts the market equilibrium under different scenarios.

WB09 CC Room 303D In Person: Control of Queueing Systems and Applications General Session Chair: Sai Mali Ananthanarayanan, Columbia University, New York, NY, 10027, United States 1 - Dynamic Allocation of Reusable Resources: Logarithmic Regret in Hierarchical Networks We study network revenue management problem with reusable resources. The resources are to be sequentially allocated to customers with different arrival rates, rewards, and resource requirements. Each accepted customer occupies the requested resources for a random duration after which the resources become available again. The objective is to maximize the long-run average reward under resource constraints. We adapt a queueing loss network framework to solve such problem. The performance of any online policy is bounded by the solution of a corresponding linear program relaxation. We show that when the network has certain hierarchical structure, a simple threshold policy induced by the LP solution achieves logarithmic regret in proper asymptotic regime. We demonstrate through numerical examples that hierarchical structures play a key role in the performance. 2 - Optimal Ergodic Harvesting under Ambiguity Chuhao Sun, University of Michigan, Ann Arbor, MI, United States, Asaf Cohen, Alexandru Hening We consider an ergodic harvesting problem with model ambiguity that arises from biology. To account for the ambiguity, the problem is constructed as a stochastic game with two players: a decision-maker (DM) chooses the best harvesting policy and an adverse player chooses the worst probability measure. The main result is establishing an optimal strategy of the DM and showing that it is a threshold policy. The optimal threshold and payoff are obtained by solving a free-boundary problem based on the HJB equation. As part of the proof, we fix a gap that appeared in the HJB analysis of [Alvarez and Hening, Stochastic Process. Appl., 2019, [5]], a paper analyzed the risk-neutral version of the problem. Finally, we study the dependence of the optimal threshold and payoff on the ambiguity parameter and show that if the ambiguity goes to 0, the problem converges to the risk-neutral problem. 3 - Queuing Safely for Elevator Systems Amidst a Pandemic Sai Mali Ananthanarayanan, Columbia University, New York, NY, 10027, United States, Charles Branas, Adam Elmachtoub, Clifford Stein, Yeqing Zhou Elevator capacity in high rise buildings during a contagious pandemic can be reduced by as much as 90% of the normal amount to allow for social distancing. Such a reduction, combined with the commonly used FCFS queuing policy, can cause large queues to build up in lobbies. Using mathematical modeling, epidemiological principles, and simulation, we propose simple interventions requiring no programming of the elevators for safely managing the elevator queues. The key idea is to explicitly or implicitly group passengers going to the same floor into the same elevator as much as possible. Based on simulation and analytical findings, our proposed interventions can significantly reduce queue length and wait time, while also maintaining safety from viral transmission in otherwise crowded elevators, building lobbies, and entrances. WB10 CC Room 304B In Person: Application of AI on Digital Platforms General Session Chair: Ecem Basak, University of Illinois at Chicago, Chicago, IL, 60640, United States 1 - Impact of AI on Consumer Decision-making and Sales Diversity Yu Kan, University of Washington, Yu Kan, Seattle, WA, 98105, United States, Uttara Ananthakrishnan In this paper, we aim to address the gap in the literature by addressing the following questions: 1) How does AI-based recommendation systems built on large-scale data with hundreds of features per customer impact consumers’ choice, purchase behavior and engagement 2) Can creating taste-based clusters (such as the algorithms deployed by Netflix on their platform) improve customer retention and engagement on the platform? 4) How does advanced machine- learning based personalization impact sales diversity on subscription-based business models? 5) How can advanced, feature-rich recommendation systems match consumers’ expectation with the delivered product thereby reducing returns and the associated negative impact on the environment 6) How does feature-rich recommendation systems impact long-chain and social media engagement? Xinchang Xie, Northwestern University, Kellogg School of Management, Evanston, IL, United States, Itai Gurvich

150