Addressing CO2 through artificial intelligence enabled technologies is an emerging priority in sustainable supply chains. This study investigates the joint adoption of Smart Emission Control Systems (SECS) and Energy Security (ES) within a tri-tier supply chain of 300 firms. We develop an integrated framework combining game theory, simulation-based optimization, reinforcement learning, and decision modeling to capture adaptive learning and profit-driven behavior. Four adoption scenarios are evaluated. Results show that ES alone provides minor efficiency gains, while SECS alone significantly reduces carbon costs and improves profitability across all tiers. The combined adoption of SECS and ES yields the strongest environmental and financial performance. Profits are reported separately for manufacturers, agents, and retailers, reflecting their distinct positions within the supply chain. Overall, coordinated adoption produces the highest gains and demonstrates that emission-reduction technologies and efficiency improvements are most effective when implemented together. The framework illustrates how multi-tier, simultaneous adoption decisions can be supported and highlights the need for policy tools-such as subsidies and carbon pricing-to enhance the economic viability of sustainable practices.

Many Chinese e-commerce platforms, including Pinduoduo and Taobao, have implemented returnless-refund policies in recent years to enhance their after-sales service systems. A "returnless-refund" policy refers to a mechanism where consumers are entitled to a refund for a defective product without returning it. However, this seemingly consumer-friendly policy introduces hidden trade-offs that complicate the joint optimization of pricing and shipping insurance decisions. This paper employs a game-theoretic framework to investigate the optimal pricing and the return shipping insurance strategies of e-retailers under the returnless-refund policy. To ensure the practical relevance of our analytical insights, we further validate the theoretical results using empirical data from a prominent Chinese e-commerce platform. Our analysis reveals that adopting a returnless-refund policy does not always hurt e-retailers’ profits, instead, the impact is determined by product cost and the proportion of returnless-refund transactions. However, the returnless-refund policy consistently suppresses the effectiveness of the shipping insurance strategy for e-retailers. We also demonstrate the impacts of the returnless-refund policy on platforms and consumers in different market scenarios. The findings challenge conventional wisdom by revealing that returnless-refund policies can reshape, rather than simply erode, e-retailers’ profit logic and risk management strategies.

The high risk associated with industrial water use and the asymmetry of accident consequences shape decision-makers’ strong aversion to losses. However, existing water resource vulnerability (WRV) assessments often overlook decision-makers’ risk aversion and system resilience. To address this, this study develops an industrial-driven regional water vulnerability (IDRWV) framework integrating Driving Force-Pressure-State-Impact-Response-Resilience (DPSIRR) diagnosis, stochastic Deck of Cards (DoC)-CRITIC weighting, and prospect-theory-based K-means-TODIMSort classification. This framework distinguishes external response from endogenous resilience, quantifies expert cognitive divergence, and captures loss-averse vulnerability sorting. Anhui Province in China, a typical industrial-agricultural composite region, serves as the case study. Results indicate that the provincial IDRWV score increases from 0.5013 in 2012 to 0.8086 in 2023, while the number of cities reaching Level V or above rises from 1 to 13. The 2020 peak score of 0.8312 reveals the short-term effect of end-of-pipe treatment, but source reduction remains a bottleneck. Industrial wastewater environmental load ratio (q=0.6868) and total industrial wastewater discharge (q=0.6015) are the dominant drivers, indicating a shift from discharge-volume control to load-capacity mismatch management. The rigid industrial structure of heavy industrial cities along the Yangtze River causes their IDRWV improvement to lag behind, making resilience building the key to overcoming their developmental challenges. This study reveals that the key to the transformation of industrialized cities lies in establishing a water resource resilience system that aligns with industrial structures through adaptive management.

Benefit of the doubt (BoD) models are widely used for multicriteria analysis. A BoD model allows the weights of criteria to vary across observations based on the observed data; however, this can reduce the discriminatory power and induce rank reversal. The common weight BoD was developed to mitigate these issues at the expense of the “benefit of the doubt” itself because a degree of weight flexibility is lost. This paper proposes the graph subset dominance BoD (GSD-BoD) method to reduce the number of observations within groups of decision-making units with similar performance and allow BoD weights to be unrestricted within those groups. As a result, the proposed GSD-BoD method overcomes the major limitations of the conventional BoD method and its extensions without increasing the occurrence of the rank reversal. The proposed method is tested through illustrative examples and simulations.

Coordinated water allocation is vital for urban agglomerations where high inter-city dependence amplifies systemic risks. However, traditional management faces strains from climate change and urbanization alongside conflicting interests across management levels. To this end, this study develops a distributionally robust bi-level multi-objective water allocation (DRBMWA) framework to facilitate coordination by integrating a Wasserstein-based distributionally robust representation of joint supply-demand uncertainty. The model features a bi-level multi-objective structure balancing economic cost, environmental load (COD), equity between cities (Gini coefficient), and systemic coordination, with lexicographic priorities (Living > Ecology > Production) reflecting social mandates. The original bi-level model is solved through a tractable single-level reformulation that embeds the lower-level lexicographic responses while preserving the hierarchical decision logic. Applied to the Shandong Peninsula urban agglomeration, results show that: (1) The DRO results reveal a robustness premium, as increasing the conservatism parameter α, which reflects the decision maker’s preference for uncertainty protection, from 1.0 to 1.5 raises the compromise cost by 34.88% while reducing the Gini coefficient from 0.034 to 0.018. (2) Under joint supply-demand stress, the deterministic compromise solution reaches a maximum city–sector shortage rate of 85.64%, whereas the DRBMWA solution maintains sectoral shortage levels within a more controllable range. (3) Compared with the single-level model, the bi-level model reduces the maximum city shortage rate from 21.16% to 14.19% and strengthens the protection of living and ecological water, although at higher transfer cost and COD discharge. (4) Scenario analysis shows that worsening supply-demand conditions substantially increase transfer costs and weaken coordination performance, indicating a greater need for earlier transfer activation and larger fiscal reserves. These findings support annual quota setting, transfer planning, sectoral allocation, and contingency preparation under uncertain water availability.

As the retail landscape evolves with the rise of digital technologies, the fusion of online and offline experiences is becoming a key strategy for businesses seeking to enhance customer engagement. This study examines the evolving dynamics of Online-to-Offline (O2O) retail, focusing on the integration of offline showroom experience services and online Augmented Reality (AR) endorsements. Offline showrooms allow consumers to evaluate products physically, while AR services provide immersive online experiences. Our findings highlight several interesting yet challenging insights. First, suppliers may shift from single-channel strategies to integrated interaction-based services when both showroom and AR effects are strong; however, when AR effects are weaker, offline showrooms become more dominant. Second, both offline and AR services stimulate respective service investments and improve supplier profitability, revealing the dual role of these channels as both substitutes and complements. Third, suppliers are more likely to pursue integration strategies when integration costs are low, but intriguingly, they may still integrate even under high costs if the interactive benefits are sufficiently large. Finally, pricing and service quality differentiation across channels illustrate the strategic difficulty of maintaining consistency in omnichannel environments. This research underscores the importance of balancing AR and showroom services, with integration offering strategic advantages for businesses like Apple Inc., especially when backed by strong financial resources to support higher integration costs.

This paper is about N-person cooperative games embedded in a set of N linear programs. In the 1975 seminal paper about linear production, Owen showed that an optimal dual vector of the Grand Coalition (of the Producers) provided the key for determining a solution in the Core. This Core solution depends entirely on the resources (right-hand sides) of the N linear programs; the technology matrices and objective coefficients have no role. We provide a Shapley inspired alternative Core solution which includes the impact of technology and prices missing in the Owen Core solution.

People commonly perceive the Yangtze and Yellow River basins as important water systems in China and rich in cultural and ecological values. However, they are also frequently threatened by floods, earthquakes, and other disasters. The traditional top-down disaster management model has always neglected the public's needs, making it difficult to enhance community resilience effectively. To improve such a situation, this study combines the analysis of public opinion in social media to identify the public's demand for “cultural construction” and “community participation” in disaster response. Moreover, it also proposes a resilience assessment framework centered on cultural embeddedness, which embodies the concept of “people-centered” governance. More precisely, the framework covering disaster response is constructed based on provincial panel data. According to the data, a three-stage network DEA model covering the entire process of pre-disaster, disaster, and post-disaster is constructed to quantify the level of Community resilience. So that the indicator system is dynamically modified to enhance the adaptability of the national situation. Finally, through empirical research, the results show that cultural factors contribute significantly to the resilience of post-disaster recovery, indicating that cultural inputs and community identity play a key role in disaster governance. To sum up, the study provides theoretical support and practical paths for disaster management in watersheds. Moreover, it emphasizes the significance of multiple synergistic mechanisms in enhancing regional sustainable development and social resilience.

Driven by the increasing complexity of modern data environments, the challenge of making accurate and efficient decisions in uncertain contexts has become a prominent research topic. Inspired by this, a sequential three-way decision model is proposed to dynamically fuse incomplete mixed data from both static and dynamic perspectives in this paper. Specifically, complete datasets are firstly utilized in the static perspective, employing an attention-based method to evaluate the importance of attributes, partition attribute sets, and systematically determine the order of attribute analysis. In the dynamic perspective, data processing in incremental environments is focused on, and adaptive threshold pairs are introduced. Subsequently, two distinct metrics, namely matching coefficients and Gaussian functions, are adopted to compute similarities for categorical and numerical data respectively. Moreover, four T-norm fusion methods are integrated, which are Minimum T-norm, Product T-norm, Lukasiewicz T-norm, and Cosine T-norm. In addition, a crucial cost function is constructed to incorporate hierarchical importance and misclassification penalties. Finally, comparative experiments are conducted on six datasets to evaluate our model against existing methods. Experimental results show that our method can effectively reduce decision-making costs while maintaining decision accuracy. In conclusion, this study provides effective solutions for dynamic data fusion and multi-stage decision-making in complex environments, offering significant theoretical and practical importance.

Large-Scale Group Decision-Making (LSGDM) requires integrating diverse stakeholder opinions in complex, dynamic environments. Traditional GDM methods often lack adaptability to evolving social dynamics and noncooperative behaviors. We propose a Multi-Agent Reinforcement Learning (MARL) framework using Multi-Agent Deep Q-Networks (MADQN) integrated with Social Network Analysis (SNA). Decision-makers (DMs) are clustered into communities, each managed by an agent that autonomously adjusts preferences. A weight penalty mechanism mitigates noncooperative behaviors by reducing the influence of resistant clusters, enhancing consensus efficiency. Simulation results show that MADQN outperforms traditional GDM and other RL-based methods in consensus quality, decision efficiency, and adaptability, reducing iterations and preference adjustments, especially in time-sensitive scenarios.

The rise of low-altitude economies and the rapid development of unmanned technologies underscore the strategic role of Unmanned Aerial Vehicle (UAV)-based delivery. This study examines a two-tier food delivery supply chain with a dominant platform (e.g., Meituan) and two types of providers: riders and UAV service suppliers. We analyze stakeholder strategies under varying UAV technological value, government subsidies, and platform cooperation models. Findings show that UAV technological value strongly influences decisions, enhancing rider orders and delivery fees while temporarily suppressing UAV orders until clear time advantages emerge. Government subsidies enhance UAV competitiveness, adjust fees, and increase profits for both platforms and UAV suppliers. Platform cooperation models and subsidies interact synergistically: revenue-sharing with subsidies is optimal under high technological value, while per-order payments with subsidies are preferred under medium value or extreme time gaps. Furthermore, riders act as flexible incumbents, differentiating when UAV technological value is high, but time advantages are limited, collaborating through revenue-sharing when UAV value is moderate, and adjusting services (e.g., night or niche delivery) when UAV subsidies are high but technological value is low, with some conditions creating a “strategy vacuum.” UAV service suppliers behave as agile entrants, favoring per-order payment with subsidy under moderate technological value and extreme time gaps, and revenue-sharing with subsidy when technological value or combined subsidy conditions support shared-risk growth. Platforms act as strategic coordinators, dynamically selecting between revenue-sharing and per-order payment with subsidy based on UAV value, time gaps, and subsidies to balance risk, secure profits, and align incentives.

To address the critical limitation of traditional data envelopment analysis (DEA) in discriminating among efficient decision-making units (DMUs), this study introduces a novel contribution-driven ranking framework integrating Shapley values and hierarchical frontier analysis. An interactive contribution analysis framework is developed to systematically quantify the marginal impacts of efficient DMUs on the efficiency fluctuations of inefficient ones via Shapley values, overcoming the single-unit removal bias and modeling multivariate combinatorial interactions. A dynamic hierarchical frontier stripping system is designed to reconstruct efficiency frontiers by iteratively excluding evaluated DMUs, incorporating potential benchmark relationships to enable comprehensive discrimination of all DMUs. The proposed method establishes a bidirectional interaction mechanism between efficient and inefficient DMUs, inferring the structural contributions of efficient DMUs to the frontier through the efficiency fluctuations of inefficient DMUs. This approach significantly enhances the discrimination capability for both unreferenced efficient DMUs and inefficient DMUs. Empirical validations across two cases demonstrate that the framework provides a data-driven methodological advancement for DEA ranking, seamlessly integrating with traditional efficiency evaluations to offer scientific decision support for resource optimization and benchmark analysis in DEA contexts.

This paper proposes an integrated optimization framework for electric bus systems, jointly addressing strategic charging facility planning and operational charging scheduling. The model accommodates heterogeneous fleets with varying energy consumption rates, explicitly models deadhead movements between routes and chargers, and enforces tight temporal tracking of charging start times and durations to avoid charger conflicts and minimize idle charger usage. To address computational challenges on large-scale transit networks, a decomposition-based heuristic algorithm (DHA) was developed. The framework is validated on a real-world transit instance in Kingston, Canada. Results show that the integrated approach yields cost-efficient combinations of fleet composition, charger deployment, and charging schedules; a heterogeneous fleet matched to route requirements outperforms homogeneous alternatives. DHA provides substantial reductions in computation time with only minor losses in solution quality, making it suitable for large-scale exploratory analysis and preliminary design, while CPLEX can be used to refine high-quality candidates. Sensitivity analyses reveal that restricting en-route or overnight depot charging notably increases fleet sizes and system costs, while total costs exhibit a U-shaped relationship with charging power and duration. These findings offer actionable guidance for transit planners prioritizing investments and designing resilient charging strategies during staged electrification.

While coping with the competition from autonomous ride-hailing (AR) platforms is a major challenge for manual ride-hailing (MR) platforms, existing literature rarely investigates their responses through horizontal cooperation. We build a game-theoretic model involving one AR platform and two MR platforms to analyze how MR platforms, with one MR platform as the initiator of the cooperation, choose the team-up mode under the influence of passenger structure and aggregation cost sharing. In equilibrium, when the increase in loyal passengers is small, the team-up mode enlarges the loyal passenger base without incurring excessive aggregation costs and strengthens the competitiveness of the MR platforms in the non-loyal passenger market, thereby increasing their profits. In this case, both MR platforms choose the team-up mode. When the increase is moderate, the aggregation cost-sharing structure becomes the decisive factor: if MR platform a bears a lower cost, it chooses cooperation, while MR platform b chooses the go-solo mode, and vice versa, leading to failed cooperation. When the increase is large, the profits under the team-up mode are lower than those under the go-solo mode, so both MR platforms choose the go-solo mode. The results provide a theoretical basis for MR platforms to respond to the impact of AR platforms.

Firms in the commodity industry are not only faced with increasingly fierce competition, but are also exposed to price risks on both supply and demand sides. This paper examines operational and futures trading decisions for raw materials and finished products by classifying bilateral spot-futures price correlations into three types (same-side, parallel, and cross) and proposing a two-stage stochastic supply chain network equilibrium model. The analytical results under duopolistic competition show that the bilateral hedging strategy can provide a firm with an advantage in increasing market share and profits when integrated with operational decisions. Further numerical analysis reveals that this advantage can either amplify or diminish depending on price correlations. Moreover, to analyze multi-competitor scenarios, we conduct simulations across varying market structures using realistic data. Our results show that widespread adoption of bilateral hedging intensifies competition and erodes its advantage. When a firm enhances the completeness of its hedging strategy, this adversely affects all other market participants. This paper presents the first systematic study of how competitive operational and financial decisions interact under bilateral risk hedging.

This study develops a dynamic pricing model for experiential products by explicitly integrating product perceived quality and AI-driven consumer learning. Several interesting and key conclusions are obtained. First, the results demonstrate that companies strategically adjust pricing in response to the strength of AI-enabled information cascades and the perceived quality signals after AI learning. Moreover, AI-driven consumer learning significantly shapes both optimal pricing and revenue. Specifically, consumer AI learning has a significant impact on both pricing and revenue gains. When the information cascade is in a positive state and the perceived quality indicator after AI learning is positive, it indicates strong market demand, and consumers are willing to accept higher-priced products, allowing companies to adjust prices upwards to increase revenue. In contrast, when the information cascade is in a positive state and the perceived quality indicator after AI learning is negative, it suggests weak market demand, and consumers are unwilling to pay higher prices for products, prompting companies to lower prices to stimulate demand. Furthermore, this study explores when demand fluctuations under consumer AI learning for various products increase, prices should be adjusted downward, highlighting the significant effect of demand changes under AI learning on pricing strategies when facing competition.