Yuxiang Sun

Role-based collaboration (RBC) is a role-centered computational paradigm for solving collaborative problems, where group multirole assignment (GMRA) is an important component. This article focuses on leader–member teams, a common organizational structure in project management, and extends the GMRA framework to address two critical challenges. First, evaluating the qualifications of leaders and members is nontrivial due to their distinct responsibilities. To address this, we propose a capability–requirement matching evaluation (CRME) method that applies differentiated mechanisms to assess leaders and members. Second, existing studies mainly maximize overall performance while neglecting task performance balance, which is vital for synchronized progress. To overcome this limitation, we develop a group multirole assignment with balanced task performance (GMRABP) model that incorporates a penalty-augmented objective to maximize team performance while reducing disparities across tasks. Furthermore, two linearized variants, GMRABP-A and GMRABP-B, are introduced to enhance computational efficiency. Extensive experiments and comparative analyses validate the effectiveness of the proposed methods, offering practical strategies for managing projects where both performance maximization and progress coordination are essential.

In competitive wargames, human players leverage visual attention mechanisms to rapidly identify mission-critical information and make real-time decisions. However, conventional deep reinforcement learning (DRL) methods often lack this ability to focus on critical cues, limiting their efficiency and decision quality. To address this gap, we propose a novel framework that enhances agent decision-making by incorporating human visual attention (HVA) into DRL through two key innovations. First, we develop an improved convolutional neural network for visual attention prediction, which outperforms existing models across multiple evaluation metrics. Second, we propose HVA proximal policy optimization (PPO), a novel DRL framework that integrates HVA through two mechanisms: 1) attention-guided sampling; and 2) multimodal feature fusion for comprehensive state representation. Experimental results demonstrate that our framework achieves an 84.4% winning rate (15.6% higher than standard PPO) with significantly faster convergence. Ablation studies further confirm that both the attention prediction model and its integration mechanism contribute substantially to performance gains. These findings highlight the effectiveness of incorporating human cognitive patterns into agent learning, offering new insights for human-agent hybrid strategies in dynamic tasks.

With the widespread application of AI in workplace scenarios, integrating AI into workflows has become a significant trend. However, most existing studies treat AI as independent agents operating in parallel with humans, assigning tasks in isolation, and failing to fully exploit AI’s impact on human capabilities. This article goes beyond the simplistic division of labor and proposes an AI-augmented collaborative task allocation method, emphasizing AI’s role in supporting human performance. By systematically modeling factors, including individual differences, interpersonal conflicts, technical constraints, and AI’s dynamic impact on human capabilities, we establish a multidimensional AI-augmented capability model to quantify capability impacts. Fuzzy interval numbers and cloud models are employed to address measurement instability and the heterogeneity of individual capabilities. Real-world case studies and numerical experiments validate the method’s effectiveness in scenarios that reflect realistic office characteristics and scales. Furthermore, experimental analyses identify transition patterns in AI-augmented environments, and verify the method’s adaptability to different AI development stages and diverse business contexts. The results provide a new theoretical perspective for understanding organizational resource reallocation driven by emerging technologies.

The human–machine collaboration task-allocation problem involves three major challenges: role diversity, capability heterogeneity, and task dynamics. Most existing studies treat humans and machines as parallel units through a static “Human + Machine” additive paradigm, which neglects the evolution of capability during collaboration. Some works “deeply couple” relatively low-autonomy machines with humans, thereby limiting the system’s flexibility in resource scheduling and dynamic reconfiguration. This study analyzes the problem from a multiagent system perspective and classifies execution units into three types: human agents, machine agents, and human–machine collaborative agents, and distinguishes between their independent and collaborative capabilities. Next, we propose the dynamic short-board balance synergy assessment method, which integrates the “short-board” concept to quantify collaboration performance and leverages agents that have low independent but high collaborative capabilities. By incorporating multiple constraints, we establish the role-based human–machine collaboration (RBHMC) model, prove its NP-hardness, and design a multi-level solving approach to handle small-scale and medium-to-large-scale data separately. The experimental results indicate that, compared with “Human + Machine” and “Deep Coupling” models, RBHMC outperforms in task completion rate, resource utilization, and system robustness. An industrial case study further validates its applicability and superiority in real-world settings. Finally, RBHMC’s transferability is validated through vertical technology adaptation and horizontal scenario migration, providing a scalable solution for multidomain human–machine collaboration in complex scenarios.

Recent advances in mathematical reasoning and the long-term planning capabilities of large language models (LLMs) have precipitated the development of agents, which are being increasingly leveraged in business operations processes. Decision models to optimize inventory levels are one of the core elements of operations management. However, the capabilities of the LLM agent in making inventory decisions in uncertain contexts, as well as the decision-making biases (e.g. framing effect, etc.) of the agent, remain largely unexplored. This prompts concerns regarding the capacity of LLM agents to effectively address real-world problems, as well as the potential implications of biases that may be present. To address this gap, we introduce AIM-Bench, a novel benchmark designed to assess the decision-making behaviour of LLM agents in uncertain supply chain management scenarios through a diverse series of inventory replenishment experiments. Our results reveal that different LLMs typically exhibit varying degrees of decision bias that are similar to those observed in human beings. In addition, we explored strategies to mitigate the pull-to-centre effect and the bullwhip effect, namely cognitive reflection and implementation of information sharing. These findings underscore the need for careful consideration of the potential biases in deploying LLMs in Inventory decision-making scenarios. We hope that these insights will pave the way for mitigating human decision bias and developing human-centred decision support systems for supply chains.