Artificial Intelligence (AI), a term coined by Emeritus Stanford Professor John McCarthy in 1955 [10], was defined by him as “the science and engineering of making intelligent machines”. Much research has created human programmed machines to behave in a clever way, such as playing chess, but, today, machines can learn, at least partly like human beings do [9]. Machine Learning (ML) is the part of AI studying how computer agents can improve their perception, knowledge, thinking, or actions based on experience or data. For this, ML draws from computer science, statistics, psychology, neuroscience, economics, and control theory [9]. ML learns through examples and benefits from human data contributions, while the process and outcome of the machine remain partially opaque. No wonder users without experts in this field hesitate to trust such a machine.

Reliable Blockchain Transaction Fee Mechanisms with Revenue Guarantees In the paper “Bayesian Mechanism Design for Blockchain Transaction Fee Allocation” by Xi Chen, David Simchi-Levi, Zishuo Zhao, and Yuan Zhou, the authors address the critical challenge of designing reliable blockchain transaction fee mechanisms (TFMs) that ensure truthfulness, collusion-proofness, and desirable miner revenues. By proposing an innovative “auxiliary mechanism method,” the authors establish connections between Bayesian Nash Incentive Compatible (BNIC) and Dominant Strategy Incentive Compatible (DSIC) mechanisms through an “auxiliary-variation decomposition.” This breakthrough allows the proposed mechanism to bypass previous limitations in TFM design via the Bayesian game setting, achieving constant-ratio approximations of optimal miner revenues while providing strong incentive guarantees. The research not only advances the theoretical understanding of blockchain fee mechanisms but also demonstrates the potential versatility of the auxiliary mechanism method for designing and optimizing BNIC mechanisms across broader applications. This work represents a significant contribution to both blockchain economics and mechanism design, offering a pathway toward more reliable and efficient blockchain ecosystems.

Dual-Sourcing Inventory Management Under Incomplete Information The class of dual-sourcing problems is one of the most fundamental challenges in inventory and supply chain management and is a common practice in global supply chain management. Under a dual-sourcing strategy, a firm procures items from a regular source with lower cost and longer lead time and an expedited source with higher cost and shorter lead time. We focus on Tailored Base-Surge (TBS) policies, where a constant order is placed with the regular source each period, and the expedited source follows an order-up-to rule. Although existing literature assumes that stochastic demand processes are known inputs to optimization models, decision-makers often lack this information in practice and must learn it in real time to maximize expected revenue. In “Tailored Base-Surge Policies in Dual-Sourcing Inventory Systems with Demand Learning”, B. Chen and C. Shi develop online learning algorithms for finding the optimal TBS policy, demonstrating strong theoretical and numerical performance.

Advancing Column Generation by a Novel Variable Fixing Method In the paper titled “DeLuxing: Deep Lagrangian Underestimate Fixing for Column-Generation-Based Exact Methods,” Dr. Yu Yang introduces DeLuxing—an innovative variable-fixing technique that significantly advances column-generation-based exact methods for solving large-scale optimization problems, particularly vehicle routing problems (VRPs). DeLuxing leverages a novel linear programming formulation with a small subset of the enumerated variables, which is theoretically guaranteed to yield qualified dual solutions for computing Lagrangian underestimates. By eliminating over 75% of the unnecessary variables, DeLuxing drastically boosts computational efficiency, doubling the size of CMTVRPTW (capacitated multitrip vehicle routing problem with time windows) instances that can be solved optimally. Additionally, this breakthrough accelerates top VRP solvers like RouteOpt by up to 71%. The core concept underpinning DeLuxing extends to broader contexts such as variable type relaxation and cutting plane addition, achieving an additional 25% speedup for difficult instances.

How to Learn Which Customer Class to Serve Next? In “Learning to Schedule in Multiclass Many-Server Queues with Abandonment”, Zhong, Birge, and Ward tackle the challenge of scheduling (that is, how to choose the customer that a newly available server will serve) in a multiclass many-server queueing system where customers may abandon the queue. The goal is to develop a scheduling policy that performs nearly as well as a benchmark policy under full knowledge of the model primitives despite these primitives being unknown and needing to be learned. They propose a Learn-then-Schedule policy that first estimates the unknown model primitives empirically and then schedules according to the benchmark policy structure using these estimates. Such a policy achieves an optimal regret rate of order logT (where T is the system time), meaning that the performance gap between the proposed policy and the benchmark policy grows logarithmically over time.

Many cities struggle with financing their infrastructure projects. When decision makers cannot fully capture the benefits of their investments, there is a risk of underinvestment. Hong Kong’s transit operator created a model where it not only collects fare revenues but also engages in property management, leveraging the positive effects of public transport on nearby property values. In the article titled “Monetizing Positive Externalities to Mitigate the Infrastructure Underinvestment Problem,” the authors present a stochastic Stackelberg game of timing to examine the reasoning behind this approach. The issue is complex because the operator faces a two-dimensional optimal stopping problem that cannot be simplified by changing the numéraire. The authors determine the operator’s optimal investment strategy through the use of a “penalized problem” and provide comparative statics. They also identify the conditions in which capitalizing on positive externalities can encourage infrastructure investment. Other management challenges share similar structures.

A Robust Time-Series Approach to the Budget-Driven Multiperiod Hub Location Problem The (un)capacitated multiperiod hub location problem involves uncertain time-series demands, the distribution of which is unobservable and correlated over periods. In “Budget-Driven Multiperiod Hub Location: A Robust Time-Series Approach,” Hu, Chen, and Wang develop a nested ambiguity set that characterizes uncertain periodic demands via a general multivariate time-series model and propose a budget-driven model that constrains each expected periodic cost within a budget while optimizing the robustness level by maximizing the size of the nested ambiguity set. Under certain regularity conditions on the underlying VAR(p) or VARMA(p,q) process of the stochastic demand, the developed ambiguity set enjoys the desirable property of measure concentration that translates to finite-sample performance guarantees of optimization problems’ decisions. In the uncapacitated case, the proposed budget-driven model essentially optimizes a Sharpe ratio–type criterion over the worst case among all periods. This research contributes to time-series predictive distributionally robust optimization.

Breakthrough in Measuring Efficiency with Respect to Nonconvex Technology The variable returns to scale (VRS) frontier, commonly used in data envelopment analysis, has a convex technology set and follows a specific “regular VRS” structure in economics. This structure demonstrates increasing, constant, and then, decreasing returns to scale, and it is the standard in data envelopment analysis. When the convexity assumption is relaxed, modeling regular VRS becomes challenging, and no satisfactory solution currently exists for multioutput production. In “Regular Variable Returns to Scale Production Frontier and Efficiency Measurement,” Li, Tsang, Lee, and He introduce a new framework for analyzing regular variable returns to scale and propose an innovative empirical production frontier. This new frontier can more accurately measure technical efficiency without the convexity assumption. The implications of this research are extensive, impacting fields like manufacturing, agriculture, healthcare, banking, etc., with crucial findings for informed decision making and effective policy implementation.

We propose a data-driven portfolio selection model that integrates side information, conditional estimation, and robustness using the framework of distributionally robust optimization. Conditioning on the observed side information, the portfolio manager solves an allocation problem that minimizes the worst-case conditional risk-return tradeoff, subject to all possible perturbations of the covariate-return probability distribution in an optimal transport ambiguity set. Despite the nonlinearity of the objective function in the probability measure, we show that the distributionally robust portfolio allocation with a side information problem can be reformulated as a finite-dimensional optimization problem. If portfolio decisions are made based on either the mean-variance or the mean-conditional value-at-risk criterion, the reformulation can be further simplified to second-order or semidefinite cone programs. Empirical studies in the U.S. equity market demonstrate the advantage of our integrative framework against other benchmarks. Funding: The material in this paper is based on work supported by the Air Force Office of Scientific Research [Award FA9550-20-1-0397]. Additional support is gratefully acknowledged from the National Science Foundation [Grants 1915967, 1820942, and 1838676], the Natural Sciences and Engineering Research Council of Canada [Grant RGPIN-2016-05208], and the China Merchant Bank. V. A. Nguyen gratefully acknowledges the generous support from the Chinese University of Hong Kong [Improvement on Competitiveness in Hiring New Faculties Funding Scheme] and the Chinese University of Hong Kong [Direct Grant 4055191]. S. Wang is partially supported by the National Natural Science Foundation of China [Grant 72371022]. Finally, this research was enabled in part by support provided by Compute Canada. Supplemental Material: The computer code and data that support the findings of this study and the online appendix are available within this article’s supplemental material at https://doi.org/10.1287/opre.2021.0243 .