Final one

Article preview

September 30, 2025

journal

Deep Roots, Big Impact

Our origins date back to 1929, when the National Research Council of Canada launched the Canadian Journal of Research. Today, Canadian Science Publishing has grown to become Canada’s largest publisher of international scientific journals. Learn about our history.

We publish 22 journals that cover the spectrum of natural and physical sciences and engineering, including three open access journals that are interdisciplinary and multidisciplinary. Learn more about our journals. As a mission-driven organization, we are dedicated to meeting the needs of the global research community. We are proudly affiliated with over 25 Canadian and international scientific societies and associations. Learn more about our partnerships.

We help researchers enhance the rigour and relevance of their research by making it easy to find, trust and use. We believe that scientific knowledge is powerful. Through our partnerships and agreements, we remove the barriers to vital research making it sustainable and accessible to anyone. Empowering everyone with science.

Our commitments Open Science future

CSP believes that the future is open. We are committed to becoming a premiere open access publisher and champion for open science. We strive to make high-quality, peer-reviewed scientific knowledge discoverable and accessible for everyone.

We are committed to actively removing barriers to build equitable, diverse, inclusive, and accessible participation in research publishing and publications. It’s clear to us that social justice isn’t just what’s right, but also what’s necessary for a full understanding of our world and its science. We pledge to do the work to change culture, policies, and practices within our workforce, our journals, and in the wider research community.

Sustainability

CSP is committed to developing sustainable practices and to act as champions of the SDGs during and beyond the Decade of Action (2020-2030). We acknowledge our important role in increasing awareness and understanding of sustainability issues and will strive to adopt sustainable business practices to address the climate crisis.

CSP is built on a foundation of community

Our People

Abstract

In consensus-based large scale group decision making (LSGDM) problems, some experts often exhibit adverse selection behavior due to asymmetry in information availability. This may lead to results deviating from the optimum, weakening decision making fairness and reducing consensus efficiency. For this reason, this paper proposes a large group consensus decision making method based on managing adverse selection behavior in an asymmetric information environment. Firstly, the directed Louvain algorithm is introduced to achieve the decision making subgroup division based on the directed social network. On this basis, considering the different qualifications and research fields of experts, a new weight allocation method is proposed based on the authority of experts. Next, focusing on the consensus-reaching process, a mechanism for identifying and managing adverse selection behaviors is proposed. A hierarchical recognition framework is designed for behavior identification, incorporating behavioral patterns and underlying motivations. A multidimensional dynamic adjustment strategy based on weight and preference is introduced for behavior management, then a comprehensive large-group consensus decision making method based on adverse selection behavior management is developed. Finally, the feasibility and effectiveness of the proposed method are verified using case studies and parameter discussions.

Introduction

Group decision making (GDM) is a process of gathering the knowledge, experience, and judgment of multiple decision-makers (DMs) to express preferences for solutions (Dong et al., 2018, Hochbaum and Levin, 2006, Pfouts and Arrow, 1951, Zhang and Guo, 2017), aiming to form group preferences by integrating multiple wisdom, to select the best solution, which is especially suitable for dealing with complex or multi-angle decision making problems (Ding et al., 2024, Wallenius et al., 2008, Zhang et al., 2023). The development of social media has facilitated public participation, placing DMs within an interdependent social network (Dong et al., 2018, Gai et al., 2023). GDM in social network environments has become a research focus, especially concerning the aggregation of wisdom in large-scale groups (≥20 people), large scale group decision making (LSGDM) method is proposed (Chao et al., 2021, Lu et al., 2022). This decision emphasizes the complex social relationships among DMs, often employing clustering methods to reduce complexity (Ding et al., 2020, Jiang, 2023, Nie et al., 2024). In the decision making process, achieving consensus is crucial, but conflicts often hinder consensus formation. Therefore, effectively coordinating conflicts and optimizing decision-making has become a significant issue in group decision-making research. Various conflict resolution methods have been proposed in the literature, such as negotiation and preference adjustment strategies (Cao et al., 2024, Ma et al., 2021, Zhang et al., 2024), as well as path optimization methods (Fan et al., 2024, Yuan et al., 2023). For example, Gong et al. (2023) proposed a maximum fairness consensus model that considers decision-makers’ fairness preferences under limited costs. Zhu et al. (2025) developed a minimum-cost conflict mediation path model based on graph theory to calculate the optimal path from the current conflict state to the desired consensus state, further extended under uncertain conditions. Jia et al. (2025) proposed a Q-learning-based distributed particle swarm optimization consensus mechanism for industrial product design evaluation. In LSGDM problems, it is generally necessary to invite experts from multiple fields to make decisions. At this point, it is imperative to determine the weight of experts. In most studies, experts’ social network relationships are the primary factor in determining expert weights, such as in the PageRank algorithm (Song & Gong, 2023), centrality analysis (Cheng et al., 2020, Chu et al., 2023), and influence propagation (Zhang et al., 2023). However, the aforementioned studies that solely rely on social network relationships to determine expert weights may have the following limitations: (1) High subjectivity. Social network influence mainly reflects an expert’s popularity rather than actual professional competence. (2) Susceptible to social activity. Experts with high social activity may have their weights overestimated, while those with low profile but high professional competence may underestimate their weight. (3) Lack of professionalism consideration. Failure to include experts’ academic achievements, practical experience, etc., in the calculation may lead to expert weights deviating from their actual contributions. This paper proposes a weight calculation method based on expert authority to address the above issues. Unlike existing methods that rely only on social network relationships, this paper quantifies the authority of experts from multiple dimensions, such as academic achievements, years of relevant work experience, and familiarity with the domain, to ensure that the expert weights can objectively reflect their actual contributions (Li et al., 1995). The method reduces the subjectivity bias of the social relationship factor and improves the scientificity of decision making and the reliability of consensus building.

The field of GDM is continuously evolving and deepening with the progress of information technology. In today’s highly informationized society, information acquisition and transmission play a crucial role in the decision making process. However, the distribution of information is often uneven, and this asymmetry profoundly impacts GDM. This impact is especially evident in LSGDM within a social network environment. Asymmetric information refers to the differences in the amount, quality, or interpretative ability of information held by different participants during the decision making process. Such differences may lead to certain participants taking advantage of the information to make strategic choices, thus affecting the fairness and effectiveness of GDM. This phenomenon was further elaborated by economists like Akerlof, building on Arrow’s initial challenge to the full information hypothesis, and was recognized with the Nobel Prize in Economics, highlighting its negative impact on markets and decision making (Arrow, 1963, Akerlof, 1970, Spence, 1973, Stiglitz, 2000). In GDM problems, information asymmetry may lead to deviation from the optimal result, hinder the reaching of consensus, and even trigger adverse selection; that is, the information dominant party guides the decision to favor private interests and damage collective interests (Arrow, 1963, Akerlof, 1970, Stiglitz, 1976). This undermines group fairness and trust, reduces consensus efficiency, and may lead to conflict. This paper aims to explore the influence of this phenomenon on GDM and the countermeasures.

Current research on decision making under asymmetric information environments mainly focuses on areas such as supply chain coordination and risk management, financial markets pricing and investment, and intelligent decision making with deep learning. In the field of supply chain management, emphasis is placed on information sharing and transparency, strategic alignment among partners, risk identification and decision optimization, etc., which requires DMs to pay more attention to building and maintaining trust-based information-sharing mechanisms, stable cooperative relationship networks, and incorporating risk management into the core consideration of decision making (Avinadav and Shamir, 2021, Cai et al., 2024, Fu and Xing, 2021, Lai et al., 2018). Scholars in financial markets have found that the amount of information can guide the allocation of social resources, and information superiority can bring certain benefits. Strengthening information disclosure, improving market transparency, optimizing regulatory mechanisms, and exploring effective signal transmission methods are crucial for alleviating information asymmetry and improving the pricing and financing efficiency of the financial market (Chen et al., 2024, Lee, 2021, Tian and Hu, 2023, Zhao and Hou, 2024). In the field of intelligent decision making, decision making under an asymmetric information environment is regarded as a process requiring high intelligence and strategic thinking, aiming to overcome the challenges brought by information asymmetry by optimizing information acquisition and utilization methods. This requires the system to use advanced data analysis, machine learning, and other technical means to minimize the negative impact of information asymmetry as much as possible, thereby improving the scientificity, accuracy, and timeliness of decision making (Tao et al., 2021).

However, despite the significant progress made in research on decision making under an asymmetric information environment, the existing framework is still insufficient.

(1)
Most current research focuses on the behavior analysis and model construction of individual DMs, ignoring the unique performance and potential of GDM in this complex environment (Chen et al., 2024, Gim and Jang, 2023, Lin et al., 2023, Zissis et al., 2020, Xue et al., 2020). GDM involves the interaction and cooperation of multiple DMs and contains rich social psychological processes, which are particularly complex and changeable under the condition of information asymmetry. The pioneering work of Brodbeck et al. (2007) provided a new perspective for understanding the impact of distributed knowledge conditions, particularly the influence of information asymmetry on the quality of GDM. They highlighted that when information asymmetry is handled appropriately, GDM can transcend the limitations of individual versus simple voting and achieve better decision making through the integration of unshared information. However, the exploration of this area is still in its early stages. The dynamic changes in the GDM process, such as how information differences stimulate or inhibit communication among members, the interaction patterns among members with different information levels, and how these interactions affect the quality of decision making, all lack in-depth theoretical discussion and systematic empirical verification.
(2)
Information asymmetry in GDM exists not only between DMs and those affected by the decisions but also among members within the group. Existing research often overlooks the impact of information asymmetry within the group on the decision making process. For example, some members may possess more critical information than others, which may lead to information monopolies, power concentration, adverse selection behavior, and decision biases. To date, several group decision making studies have considered the behavior of decision making groups but focused only on behaviors such as cooperation and strategic manipulation. For example, based on the theoretically reasonable optimal feedback model, Wu et al. (2020) proposed a new framework to prevent manipulation in the consensus-reaching process of social network group decision making so that experts with inconsistent opinions could strike a balance between group consensus and adjustment cost, improve the willingness of experts to adopt recommendations and achieve consensus. Li et al. (2021) proposed a method to identify and manage non-cooperative behaviors in an interactive network environment like WeChat. Xiong et al. (2023a) developed a large-scale consensus model based on a clustering method to manage non-cooperative behavior using DMs’ historical preference data. This method will better detect DMs with three definitions of non-cooperative behavior through preference adjustment, and punishment strategies will be more effective. Wu et al. (2023) applied the intuitionistic fuzzy C-means clustering method to cluster DMs and subsequently proposed a non-cooperative behavior identification and management approach based on intuitionistic fuzzy sets. This method has contributed to developing open-pit mine slope stability research and risk assessment. Du et al. (2020) introduced a hybrid consensus reaching model for managing non-cooperative behavior, combining independent consensus models with traditional ones. Gong et al. (2024) proposed a maximal fairness consensus model based on the individual’s sense of preventing manipulation to prevent the sense of manipulation and unfairness generated by DMs from destroying the consensus. Wang et al. (2025) proposed a segmented cost consensus mechanism based on a trust relationship to minimize the cost to achieve target classification in multi-criteria sorting through a preference strategy manipulation model and applied it to risk level optimization and resource allocation in supply chain risk management. Sun et al. (2025) introduced a self-esteem-driven feedback mechanism specifically designed to analyze the power structure and prevent strategic manipulation behaviors in group decision making in social networks. In fact, adverse selection is a typical group behavior in social network-based decision-making but differs from non-cooperative behavior and strategic manipulation. The core of adverse selection lies in information asymmetry, where differences in access to or quality of information enable certain individuals to exploit their informational advantage, thereby impacting decision fairness. In contrast, non-cooperative behavior typically stems from individual stances or conflicts of interest, leading participants to refuse preference adjustments and resist integration into the consensus process (Zhang et al., 2018). Strategic behavior, on the other hand, involves individuals employing game-theoretic thinking to deliberately manipulate their choices within known rules to influence the final decision (Xiong et al., 2023b). However, there is currently little research considering the effect of adverse selection on consensus decision making in large scale groups under asymmetric information environments, so this is an important direction of theoretical and practical significance.

Based on the above analysis, this paper focuses on the following three aspects:

(1)
Aiming at the shortcomings of traditional weight calculation entirely relying on social networks, we propose an expert weight allocation method based on expert authority and social network relationships. On this basis, the directed Louvain algorithm divides the subgroups of the directed social network and calculates the subgroup weights.
(2)
According to the influence of asymmetric information on LSGDM and the characteristics of social networks, we design a reasonable identification and management mechanism to address potential adverse selection behavior that may exist in the consensus-reaching process.
(3)
The consensus model of LSGDM in social networks considering adverse selection under a complete asymmetric information environment will be applied to specific cases to discuss its feasibility.

Through the above research, this paper hopes to provide new perspectives and ideas for the research and practice of LSGDM and promote the integration of decision science and social network theory.

The rest of this paper is arranged as follows: Section 2 is the basic introduction of relevant theories. In section 3, we analyze the characteristics of adverse selection to discuss the influence that adverse selection may have on the process of reaching group consensus and then construct a consensus model of LSGDM in social networks considering asymmetric information. Section 4 provides a concrete application to demonstrate the effectiveness of the proposed algorithm and analyzes the relevant parameters. Finally, section 5 offers a conclusion to the paper.