INFORMS 2021 Program Book

INFORMS Anaheim 2021

SC43

2 - Welfare-based Fairness Through Optimization in Artificial Intelligence Applications Violet Xinying Chen, Carnegie Mellon University, Pittsburgh, PA, United States, John Hooker We propose optimization as a general paradigm for formalizing welfare-based fairness in AI systems. Optimization models allow formulation of a wide range of fairness criteria as social welfare functions, while enabling AI to take advantage of highly advanced solution technology. We highlight that social welfare optimization supports a broad perspective on fairness motivated by general distributive justice considerations. In addition, we discuss how to integrate optimization with rule-based AI and machine learning, and outline research directions to explore for practical implementation of integrated methods. 3 - A Strongly Polynomial Algorithm for Risk Constrained Problems Alexander Zadorojniy, IBM Research, Haifa, Israel, Takayuki Osogami We consider a Markov Decision Process problem with risk related constraint. The constraint is a linearized variance approximation. We find a policy that maximizes a ratio of the reward expectation to its linearized variance. We show that under monotonicity assumption which is natural for risk related problem the Simplex algorithm with Gass-Saaty shadow-vertex pivoting rule is strongly polynomial for both cost models: discounted and expected average for infinite horizon. We show an application of the algorithm to the problem of maximization of the Sharpe ratio. 4 - Deep Learning-based Cutting Force Prediction with Machining Process Monitoring Data Soomin Lee, Chungnam National University, Daejeon, Korea, Republic of, Wonkeun Jo, Dongil Kim, Hyein Kim, Jeongin Koo We propose a method to predict cutting force with machining process monitoring data. The data were collected from a computer numerical control module and sensors. The proposed method employs an LSTM-based regression model composed of residual and bidirectional structures. In experiments, we compared the performance of the proposed method and the conventional LSTM as baseline. The experimental results showed that the proposed method not only has greater accuracy but also is more proper for various machining process conditions than the baseline. SC43 CC Room 213A In Person: Emerging Topics in Information Systems General Session Chair: Haozhao Zhang, Richardson, TX, 75080-3021, United States 1 - Spectator or Participator? The Optimal Mechanism for Online Advertising Platforms Facing the Transactions Between Advertisers and Third-party Data Sellers Wangsheng Zhu, University of Texas at Dallas, Richardson, TX, United States, Shaojie Tang, Vijay S. Mookerjee Every year, many online advertising slots are sold through auctions which attract a lot of advertisers. Compared with offline advertising, online advertising is more targeted. To better exploit this advantage, advertisers purchase user data and use it to select a better advertisement for each slot. Advertising platforms also realize the importance of user data. They provide technology supports for advertisers to integrate data acquisition into the auction process. Despite this, platforms are spectators of data transactions. In this study, we analyze a context in which the platform actively affects data transactions by either providing a subsidy or charging an extra fee for advertisers purchasing user data. We characterize the optimal subsidy (or fee) that depends on the game between advertisers. We also propose two subsidy mechanisms and compare their performances. 2 - Time and the Value of Data Ehsan Valavi, PhD Candidate, Harvard Business School, Boston, MA, United States, Joel Hestness, Newsha Ardalani, Marco Iansiti This research investigates the effectiveness of time-dependent data in improving the quality of AI-based products and services. We, theoretically, prove several counter-intuitive results. Having these results, we answer questions on how data volume creates a competitive advantage. We complement our theoretical results with an experiment. 3 - A Network Embedding Approach to Measure Competition Jiaying Deng, University of Washington, Seattle, WA, 98105, United States, Yingfei Wang, Zhijie Lin, Yong Tan Competition is a central concept of economic analysis and understanding the competitive market structure is essential for suppliers to derive a good competitive strategy. Leveraging a longitudinal data from a large peer-to-peer food-sharing platform, we construct a dynamic heterogeneous network among kitchens, dishes and customers. Then, a meta-path based random walker is adopted to learn the latent network representation and capture the competitive market structure among suppliers (kitchens). We integrate both the supply-based (i.e., dishes they provide) and demand-based (i.e., customers they serve) perspectives to identify the competitive market structure, and examine the

interplay between competition, reputation and promotion on kitchens’ performance. 4 - Matching Versus Wage Differentiation in Sharing Economy Haozhao Zhang, University of Texas at Dallas, Richardson, TX, 75080-3021, United States, Peng Wang, Zhe Zhang In this study, we examine a ridesharing context where drivers have heterogenous costs of providing service qualities and the platform designs contracts for drivers to reveal their types. In practice, a ridesharing platform can offer contracts that tie a driver’s service quality to either wage, or matching probability (i.e., the probability to match with a rider’s request). We find that the ridesharing platform’s choice of incentive compatible contracts, different wages or different matching probability, depends on factors like the difference in drivers’ cost of providing service quality, the relative size of riders to drivers, and the distribution of driver types.

SC44 CC Room 213B In Person: Supply Chain Management II Contributed Session

Chair: Wen Zhu, Umass Lowell, Lowell, MA, 01851-5173, United States 1 - Supply Chain Management and COVID-19: Scientometric Analysis

Olga Biedova, Assistant Professor, College of Charleston, Charleston, SC, United States, Maryam Mahdikhani

Supply Chain Management (SCM) has matured in a well-researched and highly esteemed field. The ongoing pandemic intensifies interest in this field from the general public as well as various academic groups. In this study, we address the patterns in the SCM publications prior to and after the COVID-19 pandemic. In addition, we propose a novel method that utilizes supervised machine learning algorithms for predicting publication citation scores based on unsupervised latent topic analysis. 2 - Supply Chain Contracting for Network Goods Dawei Jian, University of California-Riverside, Riverside, CA, United States How should manufacturers sell network goods through retail channels? We study this new supply chain contracting problems, where the retailer can privately observe and control the evolving market conditions. The optimal contract resembles the classic second-best in the short run, but converges to the dynamic first-best in the long run. 3 - Order Batching And Driver Routing in an Uber Style Restaurant Delivery Operation Wen Zhu, New Jersey Institute of Technology, Newark, NJ, United States, Marena Marco, Sanchoy Das The time variant state of a restaurant delivery model is described by a set of customer orders and a set of available drivers. Order attributes are the associated restaurant, promised delivery time, and delivery location. Driver attributes are the available time and current location. We model a fixed cycle scheduling model with M orders, N drivers, and R restaurants. The batching and routing objectives are to minimize driver travel distance and order delivery tardiness. Drivers are capacitated but can pick orders from multiple restaurants. SC45 CC Room 213C In Person: Inventory Management I Contributed Session Chair: Antonio Arreola-Risa, Texas A&M University, College Station, TX, 77843-4217, United States 1 - Dynamic Stochastic Inventory Management in E-grocery Retailing David Winkelmann, Universitaet Bielefeld, Bielefeld, Germany, Hermann Jahnke, Roland Langrock, Michael Roemer, Matthias Ulrich Inventory management optimisation requires the determination of replenishment order quantities in a dynamic stochastic environment. Retailers are faced with stochastic determinants such as demand, supply shortages and spoilage. We develop a general model and investigate the importance of accounting for each source of uncertainty when designing inventory management policies. The modelling framework is illustrated in a business case using real-world data from a European e-grocery retailer.

28